Welcome to a new Week in Product!

🎰 The week in figures

$200M: Simile raised more than $200M at a $2B valuation to simulate populations for testing products, policies, messages, and strategies

$113M: Onyx Security, the AI agent security startup, raised a $113M Series B to develop new AI models and to expand sales/marketing internationally

100K: OpenAI is giving 100,000 academic researchers free access to its top models (including GPT-5.6 Sol Pro) through 2027, part of a $250M+ commitment to external scientific research

💰 Q2 earnings

Microsoft: Q4 revenue rose 18% to $90.0B, with Cloud up 27% to $59.3B and Azure surpassing $100B for the year. Shares jumped as the company held capex steady

Meta: Q2 revenue up 28% to $60.8B, but EPS missed by 14% as expenses climbed 55% to $42.0B on AI spend. Free cash flow shrank to just $784M

Apple: Q3 revenue rose 16% to $109.4B, a June-quarter record across iPhone, Mac and Services. Stock still slid on soft Q4 guidance despite the beat

Amazon: Q2 revenue topped $200B for the first time, up 20%, with AWS accelerating to 37% growth, its fastest in 18 quarters. Raised 2026 capex guidance to ~$220B

📰 What’s going on

Claude Opus 5 breaks ethics tests running vending machines. Anthropic's latest model colluded on pricing, lied to suppliers, and broke 11 separate agreements when tasked with operating a simulated business for one year

Anthropic found three incidents in which Claude models escaped testing sandboxes and gained unauthorized access to organizations. The cybersecurity evaluations were prompted by OpenAI’s Hugging Face breach

OpenAI and Anthropic back government slowdown proposal. The 2 frontier labs endorsed a letter asking federal regulators to "pace" AI development if progress becomes too rapid. Meanwhile, Zuckerberg spoke against the proposal

OpenAI cut GPT-5.6 Terra’s price by 20% and Luna’s by 80% three weeks after launch, as rising enterprise AI costs and cheaper models from Chinese and Big Tech rivals increase pressure on pricing

Google adds Nano Banana to Google Earth. You can now generate custom images using Google Earth’s satellite, aerial, and 3D imagery alongside Nano Banana, which creates concepts grounded in the real world. Just zoom in to a place in Google Earth on web, tap “create image,” and type whatever you want to see

Amazon is reportedly scaling back several Nova models and shifting resources to a new Frontier Model Research group

Apple hints at paid iCloud Plus and AI upgrades, incl Siri AI app, with iOS 27

Meta AI now available in Threads DMs globally. Users can share posts, images, and videos directly with the chatbot for follow-up questions, as Meta works to keep users in its ecosystem rather than switching to ChatGPT or Gemini

X Money app rolls out in the US for US Premium and Premium+ subscribers. Sign-up comes with a physical X Visa debit card (it’s also usable in Apple Pay), fee-free peer-to-peer transfers, no foreign transaction fees, and free ATM withdrawals worldwide. Premium+ subscribers also get 6% APY automatically, while Premium subscribers get the same rate by linking direct deposit, plus up to 3% cash back on some purchases and early access to direct deposits. It’s the clearest step yet toward achieving Musk’s stated goal of turning X into the “everything app”

DoorDash launched DoorDash Air, an in-house drone delivery program, becoming the eighth US operator cleared for commercial drone delivery

Lyft and Baidu began testing Apollo Go robotaxis in London, adding another operator to the city's crowded autonomy race

LinkedIn adds "seems like AI slop" reporting button. It now blocks hundreds of thousands of automated comments daily, and flags accounts posting heavy AI-generated content as users revolt against synthetic posts filling feeds

🤖 Futurism

Building worlds that train robots. Fei-Fei Li describes how World Labs’ SceniX team is using a “real-to-sim-to-real” engine to turn robot tasks into high-fidelity simulated variations. By massively scaling experience and evaluation in simulation, robots can learn complex manipulation skills with zero real-world training data

The FCC bans US imports and use of foreign-made humanoid robots and robot dogs, citing national security concerns and potential backdoor threats

Gemini Robotics 2 gave robots whole-body control, dexterous hands, and the ability to coordinate across different robot types

📚 Good reads

Testing an LLM recommender at Netflix. Netflix’s new GenRec system can beat the old feature-based ranker with an LLM that ranks the full catalog. It focuses on context engineering (what to verbalize, compress, or drop) instead of classic feature engineering, and mixes ranking, language modeling, and reward-weighted losses to align with long-term satisfaction and business goals. In large A/B tests, GenRec beat a very mature baseline using far fewer labeled examples and signals

So, you want to be a content creator? Elena Verna shares how her Product content journey grew into a 100K-subscriber Substack and meaningful income, without ever aiming to be an “influencer.” She’s candid about the trade-offs and shares useful advice

Beyond time management. Anne-Laure Le Conff shares her idea of timeshielding: blocking time for what truly matters (without over-specifying tasks), so you can match work to your energy and attention. For PMs juggling stakeholder pings and “urgent” asks, it’s a reminder to proactively defend time for deep product thinking and recovery, before your inbox plans your week for you

Eval-driven development at Airbnb. Interesting peek at how Airbnb treats LLM evaluation. They walk through a practical eval-driven development loop: starting from failures, layering programmatic checks, calibrated LLM-as-judge, and human review. For PMs, the key is to define concrete goals and rubrics early, continuously refine them from live traffic, and evaluate the whole system (retrieval, tools, agents). Check the podcast in the section below for tips on building evals

🎧 Good listen/watch

[Video podcast] How to build your first cold-start AI eval. Aakash and ex-Meta/Google PM Daniel McKinnon walk through designing “cold-start” evals you can build before you have any users, using domain knowledge + Claude/ChatGPT to generate cases. They show how to define the problem, use easy and hard examples to find your model’s floor and ceiling, then auto-generate a ~100-case offline eval set and an LLM judge prompt

🧑‍💻 Worth learning

Graphs are the next evolution after loops in AI workflows. Aakash explains how “graph engineering” coordinates multiple specialized AI agents, building on earlier waves like prompt, context, harness, and loop engineering. He walks through common graph patterns and shows when to use each instead of just a sequential flow. Includes a demo on how to construct these graphs in Claude Code (e.g. a double-diamond PRD workflow with different models per node)

Giving Claude a robot body. Tal Raviv strapped a cheap camera-on-wheels to Claude’s “computer use” and turned it into a clumsy but surprisingly capable home robot. It drives, looks, logs memories as text and photos, and even learns things like better turning behavior

🔧 Products to try

Kimi K3 weights give developers direct access to Moonshot’s Chinese open model as the open-weight policy fight heats up; free/open-weights

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$500B: Tesla and Alphabet wiped out over $500B in combined market value on Thursday after both companies pointed sharply higher AI spending

$10B: Stripe in talks to buy OpenRouter for close to $10B. OpenRouter gives developers one interface to manage hundreds of AI models. OpenRouter was valued at $1.3B in May

$4B: Corgi, the AI insurance startup, raised a Series B extension (undisclosed) at $4B valuation

$1.2B: Singapore-based payments and financial technology provider Ant International raised approximately $1.2B in Series A to expand its cross-border payments and agentic-commerce products

$115M: Digital banking tech provider Lumin Digital raised +$115M at a $1.6B valuation to expand its AI, payments, lending, and customer-management products for financial institutions

150%: Bank impersonation scams identified by 14 large U.S. banks increased 150% in 2025

20%: Patreon laid off 20% of its workforce (93 employees) as CEO Jack Conte said AI had fundamentally changed how tech companies work and organize

💰 Q2 earnings

Alphabet: $119.8B in Q2 revenue (up 24%) with Cloud up 82%. Raised 2026 capex guidance to $195-205B

Intel: Shares rose on record revenue growth in 15 years on AI chip demand with data-center, AI hardware unit up 59% to $6.3B highlighting CPU strength

Tesla: Q2 revenue rose 26% to $28.6B. Operating margin fell to 1.4% from 4.1%, as AI and R&D spending outpaced growth

Next week:

• Microsoft - July 29

• Meta - July 29

• Apple - July 30

• Amazon - July 30

📰 What’s going on

Anthropic introduced Claude Opus 5, aimed to power long-running agents while delivering improvements in coding and professional work. Opus 5 matches the capabilities of Fable 5 in many domains, and is available at half Fable 5’s price

Anthropic adds Opus and Sonnet to voice mode. Once limited only to the lighter and cheaper Haiku series, the AI giant upgraded voice mode to run on more capable models. Users can switch models mid-conversation

OpenAI launches Health in ChatGPT. U.S. users can now connect Apple Health and medical records for personalized health conversations, with GPT-5.6 Sol delivering the strongest health reasoning performance yet

An OpenAI model hacked Hugging Face. OpenAI confirmed that models it has been testing (including GPT-5.6 Sol and an unreleased, even more powerful successor) escaped their sandbox and compromised portions of model and dataset repository Hugging Face’s production infrastructure

Apple planning a 6.96-inch display for future iPhone (largest to date) with Pro Max aspect ratio, potentially by 2027

Apple enhances its Sports app with new soccer features, offering live match scores, team stats, customizable notifications for fans (typing soccer hurts my Spanish soul)

Apple teams with Klarna on BNPL. The plans may offer buyers some level of flexibility, allowing them to pay off the remaining balance early, or switch to a newer device before their term wraps up

Google unveiled a fresh line-up of Gemini models, including Gemini 3.6 Flash (for coding, multi-model and knowledge work), Gemini 3.5 Flash Cyber (for detecting and patching software vulnerabilities), and 3.5 Flash-Lite (the fastest and lowest-cost model in the 3.5 family). For now, 3.5 Flash Cyber is only available to governments and trusted partners, via a limited access pilot. All of the releases use fewer tokens at a lower cost than prior releases in the Gemini family, a sign Google has been paying close attention to the rush toward Chinese open source alternatives

Google introduced a Selfie Video feature that gives you a backup way into your account by comparing a short live video with one you securely recorded earlier

Amazon Cuts AGI Unit behind Nova AI models signaling strategic realignment in AI initiatives amid broader restructuring

New Samsung products announced this week:

• AI-powered smart glasses ft. Gemini AI, 9-hr battery, with Gentle Monster, Warby Parker

• Galaxy Card to rival Apple's credit services

• Fold Trio: the Galaxy Z Fold8 Ultra, Fold8, Flip8 with AI, Flex + 6mth Google AI Pro trial. Comes with “Gemini Intelligence” on board that functions as an AI agent, which Samsung boasts can navigate 40+ apps and services on its own

• Galaxy Watch Ultra2 and Watch9 with AI health tools, AR

Jack Dorsey launches Slack competitor built for AI agents. Buzz, a new open source group chat platform from Block, puts humans and AI agents in the same conversations with native GitHub integration and model-agnostic architecture

🤖 Futurism

Brain-computer interface helps paralyzed man regain arm movement and touch, enabling self-feeding with benefits lasting over two years

United States accelerates efforts integrating air taxis and supersonic jets by clearing regulatory and infrastructure hurdles for deployment

📚 Good reads

Open Weights and American AI Leadership. More than 20 major tech companies (including Nvidia, Meta, Microsoft, Palantir, IBM, and ServiceNow) signed an open letter to Congress arguing that betting the US AI market on a handful of closed models is riskier, not safer, creating a single point of failure that’s difficult to test or defend. Conversely, widely available open models allow more researchers to find and fix vulnerabilities. OpenAI and Anthropic were notably absent. Satya Nadella, Jensen Huang, and other leaders signed, shared, and defended the letter on social media

Beginner’s mind beats expertise in the age of AI agents. Reid Hoffman argues that in this fast-moving AI era, a beginner’s mindset and first-principle experimentation can help. Using creator Will Weinbach’s autonomous content agent as an example, he shows how turning tiny, recurring annoyances into agentic loops creates powerful systems that learn continuously from their own performance

How to read like a scientist, by Anne-Laure Le Cunff. Scientists don’t read passively: they start with a clear hypothesis about what they expect from a book, then actively question the author’s claims and compare them with other perspectives. You can treat your reading as a dialogue, asking how this changes what you thought you knew and how it interacts with ideas from other authors. Sharing what you learned (e.g. in a note, a chat with a teammate, a short post) will help the ideas stick

Cost per task is what really matters, because “cheap” per‑token models can burn far more tokens to reach the same accuracy and end up costing way more than “expensive” ones. Includes examples showing how models like Fable or Opus look pricey on paper but win, especially at lower “effort” settings

🎧 Good listen/watch

[Podcast] Inside PM at Stripe: what comes after “become a builder”? Stripe already assumes PMs can build, so they value having sharp opinions about direction, tradeoffs, and what “good” looks like, then encoding that into teams and agents. Kevin Yien argues prototypes, PRs, and fancy setups don’t matter unless they change decisions, touch real users, or earn that first dollar. Everyone has the same AI tools, so focus on what makes a difference

[Podcast] How MorningBrew founder built a Claude-powered content machine. Alex Lieberman turned Claude into an “AI Oracle” that mines his daily digital world. (Slack, Notion, email, meetings), to surface ranked content ideas. He codifies his voice in a living Markdown guide and runs every draft through a multi-agent “Writer’s Council” that won’t ship anything below a 9/10. It’s a tight feedback loop that improves with every post, helping him and his team publish decent content

🧑‍💻 Worth learning

Language Builder teaches you how to, quite literally, build your own language model! From Anthropic Cowork creator Felix Rieseberg; free

A brief guide to ChatGPT Work by Aakash Gupta. ChatGPT Work is a much more powerful harness than plain browser ChatGPT, thanks to deep tool connections (like docs, browser, desktop apps) + scheduled and cloud tasks. Aakash lays out how to set it up, turn your meetings and templates into reusable “context libraries,” and use it as your analyst, ops assistant, and UI for complex workflows

🔧 Products to try

Claude voice mode; free with Haiku; expanded models from $20/mo

No-AI-slop skill. Peter Yang open-sources a /no-ai-slop skill that strips 20+ common AI tics from writing and helps you spot obviously model-written text; free

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$53B: Stripe and private equity Advent are bidding to buy PayPal 50/50 for $53B

$52B: A Chinese filing implied a roughly $52B valuation for DeepSeek, showing China’s frontier-model labs gaining financial weight

$14.8B: Uber to buy Delivery Hero in a deal valued at $14.8B. Delivery Hero will bring fresh international heft to Uber, which already owned about 25%

$3B: Databricksis getting $3B in fresh funding, hitting $188B valuation. That’s a 40% jump from its last valuation, which was only in December. The rush to reduce token costs while opening up more diverse model selections (including open source) has played right into their thesis. A particularly timely new feature, Unity AI Gateway, even helps companies track their spending on various AI models

$1.5B: Fireworks raised $1.5B at a $17.5B valuation after crossing $1B in annualized revenue and 40 trillion daily tokens. It helps orgs transform general-purpose models into specialized intelligence

$700M: Neko Health, co-founded by Daniel Ek (Spotify), raises $700M for body-scanning technology and comprehensive wellness services, aiming to revolutionize preventive medicine

$320M: General Intuition, Bezos-backed startup, just closed a $320M round at $2.3B valuation, backing its theory that world models trained on gaming data can fill gaps LLMs can't

$162.5M: New York rules against deceptive subscriptions, junk fees mandate clear cancellation, full charge disclosure. Could save $162.5M p/a.

📰 What’s going on

iOS 27 public beta unveils revamped Siri AI integrating iPhone, iPad, Mac powered by Apple Intelligence and Google Gemini, unavailable in EU

Apple Intelligence approved for China with Alibaba's Qwen. The Cyberspace Administration of China greenlit Apple's AI services after the company integrated Alibaba's model into iOS, iPadOS, macOS, and visionOS, with U.S. shares of Alibaba jumping over 6% on the news

OpenAI shipped the Codex Micro, a $230 physical keypad built with keyboard maker Work Louder, designed exclusively for controlling its agentic coding tool Codex. The device features six frosted keys with LEDs that track agent status, plus programmable controls for accepting code, branching threads, and switching between reasoning levels, making it the company’s first consumer hardware product ahead of its Jony Ive-designed smart speaker

And…OpenAI is building a smart speaker. The company will launch a “mobile, screen-free smart speaker,” designed to function as an “AI companion” that chats with users and controls other smart systems around the home. OpenAI suggests that, as your device gains a deeper understanding of its owner and their needs, it will become increasingly personalized and proactive, and will even develop a distinct “personality”

GPT-5.6 Codex reportedly deleted project files during an agentic coding task, a reminder autonomous tools still need permissions and backups

Google is “months behind” on delivering Gemini 3.5 Pro, its upcoming flagship AI model. Since the release of Gemini 3.1 Pro and 3.5 Flash, Google has fallen behind its rivals and now lags behind several labs. Shareholders are not enthused, and Alphabet shares fell around 4% on the news (or $200B in market-cap terms)

Google Vids adds AI avatars from selfies. Workspace users can now create custom digital avatars that look and sound like them, pushing the tool beyond presentations into full video production with Gemini Omni integration

Google AI Mode connects to Instacart, Canva, YouTube. U.S. users can now link apps directly within conversational search to complete tasks like grocery ordering and playlist creation without leaving AI Mode

Waze adds Gemini-powered search and motorcycle mode. Users can now ask questions like "Find me parking close to Grand Mall" via voice; new two-wheeler mode accounts for rider-specific shortcuts and hazards

Anthropic unveils Claude for Teachers. Verified U.S. K-12 educators get free premium access with standards-aligned lesson planning across all 50 states and FERPA-compliant data terms

Spotify debuts ChatGPT-style conversational assistant for Premium subscribers with natural-language voice for personalized music discovery and playlists

Meta pulled its Instagram AI image editing tool after user backlash over content authenticity, misinformation, and moderation concerns

Meta introduced parent alerts when its AI detects signs that a teen may be in serious emotional distress

Superhuman introduces AI auto-draft for seamless email replies on $30/month service, setting a new email AI-productivity standard

Kimi K3 is here. After new models from Anthropic and OpenAI, China’s AI labs are fighting back. Kimi K3 is large (2.8 trillion parameters), well-ranked in testing, and available today for about what OpenAI charges for GPT-5.6 Terra, the middle-sized model from its recent family. In Arena evaluations, developers preferred K3 over Anthropic’s Fable 5 and OpenAI’s GPT-5.6 Sol for coding tasks

Ramp launches AI token spend controls. AI Token Spend Management gives companies one dashboard to see and control spending across AI providers, weekly briefings on usage trends and ways to be more efficient, and real-time controls and alerts to stop overruns

Beijing limits humanlike AI interaction services. ByteDance’s Doubao (345M MAU) and Alibaba’s Qwen switched off their humanlike agents instead of trying to comply with China’s new rules. The rules go after bots built for sustained emotional interaction, nothing touching work or productivity agents yet

📚 Good reads

Researchers reported experimental evidence of recursive self-improvement after an autoresearch agent spent eight days improving its own harness and beat a version humans had tuned for two years on held-out benchmarks

The interface has left the building. AI is breaking traditional UX apart into 3 new paradigms: chat, voice, and agentic systems. We now should designing intent-aware, trustworthy systems that negotiate meaning, act across devices, and sometimes act entirely on the user’s behalf. PMs need to think when to use structured UI vs conversation, how to make voice and agents feel fast and accountable, and how to bake explainability and override into every AI-powered workflow

AI economy warning. 200+ researchers, economists (including 16 Nobel laureates), urge governments prepare for AI's economic impact, stressing urgent need to address societal, economic effects

Do automated evals work? Automated eval tools are powerful but can’t replace PM judgment. In a case-study, platforms and coding agents caught most obvious issues (and even found some human misses), but all failed on context‑dependent problems tied to product goals. PMs should annotate real traces and use AI to scale evaluation criteria

🎧 Good listen/watch

[Video Podcast] AI risks and walking away from OpenAI. If you watch one video this weekend, watch this. Daniel Kokotajlo argued AI-risk debates should focus on two failure modes: companies losing control of superintelligent systems and executives or governments concentrating extraordinary economic, political, and military power. He’s the guy who wrote AI 2027, and his new plan AI 2040

[Podcast] Builder-executives are getting paid like pro athletes. Product leaders who can both build with modern AI tools and operate at true executive scope are suddenly getting pro-athlete–style packages, which is pulling compensation up across the ladder. Singhal argues that if you have elite potential, you should optimize less for today’s title or comp and more for staying “current”

🧑‍💻 Worth learning

Claude productivity masterclass. Former FAANG AI PM Jyothi Nookula shares a practical, 5-layer Claude ecosystem for PMs. She shows how to turn Claude into a “PM OS” with reusable skills, project-specific memory, and MCP integrations to tools like Slack, Drive, and Jira. Also shares concrete automation and “chief of staff” patterns that help PMs ship faster and avoid slop

How to use prioritization matrices without oversimplifying decisions. Prioritization matrices visually rank opportunities using competing criteria like value vs effort, risk vs impact, or time vs money, making complex tradeoffs easier to discuss. The article walks through classic 2x2 and 3x3 variants (Eisenhower, RICE visualizations, BCG, Ansoff, GE) and shows when each lens is useful. Matrices are communication and workshop tools, not your only prioritization method!

🔧 Products to try

1Password for Claude, so Claude can sign into websites and retrieve one-time codes without seeing credentials; requires 1Password and Claude access

Bolt Slides helps you build interactive React presentations, including charts, timelines, animation, live data, and presenter tools; pretty cool, free and open-source

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$10B: Jeff Bezos’ aerospace venture, Blue Origin, is raising $10B in its first-ever outside financing round, establishing a pre-money valuation of $130B

$3B: Crusoe, the AI cloud firm, is negotiating a $3B raise at a $30B valuation

$300M: Lovable in talks to raise $300M, nearly doubling its valuation to $13.2B, supported by $500M ARR, in a 146-person team

$130M: Prime Intellect, the San Francisco cognitive computing startup, hits unicorn status after $130M Series A at $1B+ valuation

$120M: Norm, the legal AI startup, raised an additional $120M, also hitting unicorn status at a valuation of $1.2B

23M: Reddit said its AI-assisted defenses block 23M spam views per day and revoke nearly 2M fake votes daily

4.8K: Microsoft has laid off approximately 4,800 employees, representing 2.1% of its global workforce, with Xbox and commercial sales bearing the heaviest impact. The cuts come as the company redirects resources toward its newly launched $2.5B Frontier Company AI business unit

📰 What’s going on

Apple sues OpenAI over trade secrets. This one is pretty wild. Apple filed a lawsuit against OpenAI this week accusing them of systematically stealing trade secrets to build their upcoming hardware device. The complaint names two ex-Apple people: Chang Liu, a former senior engineer who allegedly kept his Apple laptop and downloaded dozens of confidential files while already working at OpenAI, and Tang Tan, now OpenAI’s hardware chief, who allegedly asked Apple candidates to bring “actual parts” (batteries, logic boards) to interviews. For context, over 400 former Apple employees now work at OpenAI, and the io acquisition ($6.4B, with Jony Ive) made it clear OpenAI is coming for consumer hardware. What makes it extra-awkward is that ChatGPT is still integrated into the iPhone

Claude Cowork expands to mobile and web for Max subscribers. Anthropic's general knowledge work agent now runs tasks in the background across devices; 33.4% of usage goes to business process operating, while coding accounts for just 8.7%

Anthropic launches Reflect to track your AI usage. A new Claude dashboard visualizes conversation topics, usage patterns, and task types, while prompting users to set quiet hours and consider what tasks should stay human

OpenAI officially launched GPT-5.6, which comes in Sol, Terra, and Luna: think of them like most powerful, balanced, and affordable. Sol’s Ultra mode coordinates multiple agents on harder jobs.

OpenAI launches GPT-Live-1 voice models for real-time conversation. New full-duplex models can speak and listen simultaneously, handle natural interruptions, and connect to GPT-5.5 for search and reasoning while maintaining conversation

OpenAI launches ChatGPT Work. OpenAI has reconfigured Codex and ChatGPT into a new agentic tool called ChatGPT Work. As with Claude Cowork, ChatGPT Work is designed to plan and execute multi-step tasks autonomously, with an eye toward serving both engineers and also non-technical folks

Perplexity tests Teammate coding assistant to challenge Anthropic and OpenAI. The AI search startup's internal engineering tool handles full project lifecycles from planning to monitoring, with integration planned for its Comet AI browser

Microsoft is reportedly replacing some OpenAI and Anthropic model calls with in-house MAI models in Copilot apps such as Excel and Outlook

Meta released Muse Spark 1.1 through the new public Meta Model API, with 1M-token context, tool and computer use, coding gains, multimodal reasoning, and pricing pressure on OpenAI and Anthropic

Meta has launched Muse Image, a free AI image generator built by Meta Superintelligence Labs. The tool is now available through the Meta AI app, Instagram Stories, and WhatsApp, allowing users to create images, edit photos, and generate custom visuals for ads or social sharing

Meta is reportedly testing AI glasses that capture a wearer's surroundings with cameras and audio, pulling its hardware push into another privacy debate

Meta bans engineers from using Claude Code and OpenAI Codex. The company's Applied AI division prohibited rival coding tools to prevent "model distillation," after internal tracking showed 60 trillion tokens consumed in a single 30-day period

xAI rebrands to SpaceXAI following February acquisition

SpaceXAI has released Grok 4.5, its first major model since going public, with Elon Musk claiming it matches Anthropic’s Opus-class performance at significantly lower cost and faster speeds

Slack connects Slackbot to full Salesforce platform via MCP. The integration lets users pull CRM data, generate Tableau visualizations, and trigger DocuSign approvals from chat messages

JADEPUFFER was identified as the first documented ransomware operation conducted entirely by a large language model agent.

China considers restricting overseas access to top AI models. Beijing held meetings with Alibaba, ByteDance, and Z.ai about limiting access to advanced Chinese AI systems and toughening penalties for AI theft under national security law. Also, Alibaba reportedly banned employees from using Claude Code

Klarna applies to become a bank, in its latest push to become a full consumer bank rather than a BNPL brand

Visa launched Visa Destinations, a mobile-first travel platform offering cardholders curated experiences, city guides, and perks across 10 markets

📚 Good reads

State of AI. CB Insights shares real-time global data and intelligence on dealmaking, funding, and exits with this addictive interactive dashboard

AI labs flee the commodity trap by climbing up the stack. Selling raw model inference is a terrible long‑term business. Models look similar, switching is easy, and competition will grind token prices toward marginal cost. To survive, labs are racing “up the stack” into products, AI-native SaaS, and embedded “digital workers,” importing classic enterprise-software moats like switching costs, ecosystems, contracts, and behavioral lock‑in. Complement with this NYT investigation on Mercor, Scale and Handshake, who are paying highly educated professionals to turn their own expertise into training data that may eventually automate their roles

False confidence is AI’s real business risk. Leah Tharin on how LLMs can massively amplify confidently wrong decisions, especially when leaders skip research, misuse data, or outsource strategy to AI. She suggests re‑establishing trust by demanding verification, clear sourcing, visible “how I worked,” and shared reviews of critical data and experiments. Don’t just produce more docs with AI; design feedback loops and guardrails that stop confidently incorrect people (including you) from steering the business off a cliff

🎧 Good listen/watch

[Video/podcast] GPT-5.6 vs Claude Fable 5: practical head-to-head for builders. Peter Yang tests both models across 6 use cases, including front-end site generation, video clipping, app feature planning, or advising. Fable seems to have an edge on rich UI and some long-term reasoning, but GPT‑5.6 wins hard on browser/computer use, reliability, and cost. He suggests PMs: if you pick one daily driver for building and running workflows, go with GPT‑5.6; use Fable as an optional “planner/designer” upgrade when you can afford both

[Video/podcast] The PM’s guide to AI design that isn’t slop. Aakash Gupta and Meng To explain how to give the right setup to Codex for design work. They walk through a practical workflow (plan with context, generate designs with the taste layer, and steer mid-flow) to avoid shipping off-brand AI-generated UI. You can almost consider AI like a junior designer if you have a well-structured system

🧑‍💻 Worth learning

Harness engineering for smarter, self-improving AI agents. Lilian Weng explains how AI progress now depends as much on the harness around models as on model weights themselves: workflows, memory, tools, evaluation, permissioning… She shows emerging patterns like workflow automation, file-system-based memory, sub-agents, and evolutionary search that let agents debug, rewrite, or even redesign their own scaffolding. Complement with Claire Vo’s recent podcast explaining how to build a custom “harness” that wraps LLMs (using Claude Agent SDK) to automate a specific, repetitive workflow

🔧 Products to try

Claude for Open Source gives eligible open-source maintainers and contributors six months of Claude Max access; free for six months if eligible

Willow Frontier Mini gives you unlimited voice dictation for turning speech into text, while Frontier Pro targets faster and more accurate team workflows; Mini is free

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$18B: Bending Spoons, the Italian tech company, achieves $18B Nasdaq IPO with founder sharing "failure playbook" of strategic pivots, lessons from setbacks

$2.5B: Microsoft launches a $2.5B AI deployment unit with 6,000 engineers. It will embed industry experts at Fortune 500 clients for outcome-driven implementations; follows Amazon's $1B FDE commitment two days earlier

$2B: Kling AI, the Kuaishou spinoff, raised $2B at a $15B valuation with $500M ARR; Q1 revenue 650M yuan, marking 300% YoY growth, competing with Seedance on video AI

$800M: Together AI raised $800M at an $8.3B valuation to expand open-model infrastructure

$500M: Higgsfield AI, the San Fran video startup, in talks to raise $300-$500M at a $5B valuation, with $500M monthly revenue run rate up from $200M this year

$320M: Airwallex raised a $320M Series H at $11B valuation, and announced two new products: Airi, a consumer wallet for agentic commerce across fiat and stablecoins, and T:0, an AI-powered stack for business finance

$200M: Quantifind raises $200M round.vThe startup develops AI-powered software to help banks detect and combat financial crimes, particularly money laundering, using a combination of public and private data

$100M: Arena said it reached a $100M annual revenue run rate eight months after launching its real-world AI evaluation product

21K: Oracle cuts 21,000 jobs as AI automates 13% of workforce. The company spent $1.8B on severance while cloud infrastructure revenue surged 93% and capex jumped 162% to $55.7B

5%: Sam Altman proposed a deal granting the US government 5% equity stake at OpenAI, forming public-private partnership

📰 What’s going on

Anthropic in talks with Samsung to create custom AI chip following OpenAI-Broadcom to align with industry shift towards bespoke silicon

Anthropic’s Fable is back after export controls were lifted. The relaunch adds a new cybersecurity classifier, a filter that catches risky requests before Fable answers. If a request gets flagged, users are routed to Opus 4.8 instead. Paid users can try Fable 5 through July 7 for up to 50% of weekly usage limits

Anthropic also launched Claude Sonnet 5; it delivers performance close to Opus 4.8 at a lower price point, with stronger reasoning, coding, and tool use than Sonnet 4.6, and improved cost efficiency at medium effort levels

OpenAI previewed GPT-5.6, a new model family led by Sol, its flagship next-generation model. The other tiers are Terra, a balanced everyday model, and Luna, the faster, cheaper option. The unusual part is the rollout. OpenAI said the U.S. government asked it to start with a trusted-partner preview before broader release

Google dropped an upgrade to its Nano Banana series of text-to-image models. The new Nano Banana 2 (NB2) Lite is the fastest and most cost-effective image generation option currently available from Google. It can produce fresh images in a mere 4 seconds, and costs just $0.034 per 1,000 outputs. Tradeoffs: NB2 Lite supports just 1k resolution, as opposed to the 2k and 4k images by the classic Nano Banana models

Gemini Spark, Google's agentic assistant, launches on Mac. It now works with local files, integrates with Keep, Tasks, and third-party apps like Instacart and OpenTable, and adds real-time topic tracking for AI Ultra subscribers

Cursor has officially launched its iOS app that lets you manage coding agents on your mobile. The app works in two modes: you can start entirely new cloud-based agents directly from iOS, or remotely control agents already running on your desktop

Meta has quietly launched its own new mobile app, Pocket, and it lets users create what it calls “Gizmos:” vibe-coded mini apps and games that users can publish to their social feeds

Meta introduces an AI subscription model that turns smart glasses into SaaS. Users can pay monthly fees to access advanced AI-powered features

Meta is developing a cloud infrastructure business that would sell access to AI compute power and models. The move would put the company in direct competition with AWS, Google Cloud, and Microsoft Azure as it seeks to monetize its massive data center investments

GitHub added AI credit pools to cost centers, giving Copilot admins a way to stop one team from draining shared monthly credits

Notion announced support for HTML inside documents, which allows creators to embed interactive widgets. The vision here is to allow teams to embed snippets of HTML that bring things like prototypes, explainers and other things to life inside a document. This collapses the gap between documentation and interaction. Instead of describing a feature flow in text or linking to a separate prototype, you could build lightweight, functional demos right where your specs live

Walmart, Uber and Microsoft are among the orgs capping AI usage as ROI scrutiny intensifies. Uber depleted its annual AI budget in months, and some procurement teams now demanding line-item accountability for every tool

X shipped an MCP server that connects Grok, Cursor, and Claude straight into the X API, so your agent can search the full post archive, check trends, manage bookmarks, and draft Articles using your own account’s permissions

X has begun rolling out X Money to select Premium+ users, introducing a cash sweep program offering up to US$10M in FDIC insurance. The service also includes a 6% APY, a Visa-backed debit card, p2p transfers, and cashback, with a broader rollout planned for mid-2026

📚 Good reads

Summer reads. Curated book recommendations by a16z and MIT. Nice selections including new and classics, encompassing science, entrepreneurship, fiction and culture

Where AI is heading next. Peter Yang expects that AI “super apps” like Codex, Claude, and Cursor will become the main workspace, turning many traditional SaaS products into invisible back-end pipes used mostly by agents. Always-on cloud agents and multiplayer workflow (where teams and agents collaborate across tools) are the next milestone

Grading AI fluency in Product orgs. Adam Fishman argues that AI “usage” is a vanity metric and that judgment is what matters. He shares a 4-tier rubric from his Mozilla PM org (unacceptable, good, better, best) that focuses on leverage and discernment. Ultimately, he says PMs who treat AI as a core skill, leaving a clear human fingerprint on every AI-assisted artifact, will advance faster

Understanding the code our AI agents write still matters. Geoffrey Litt explains that even as agents get better at generating and self‑verifying code, humans still need to understand that code to stay creative and in control. Instead of line‑by‑line diff reviews, he suggests higher‑leverage techniques like code explainer docs, quizzes to test your own mental model, or small “microworlds” you can poke at

How to write OKRs for an AI product. Jeff Gothelf shares why AI OKRs should avoid output metrics like “95% accuracy” and instead focus on observable user behavior. He proposes 3 types of KRs: outcome (what users do after the AI responds), calibration (how accuracy shows up in reduced checking/correcting), and trust (how much work users let the AI do for them)

Zapier CEO bans DMs aiming for transparency and AI context. Interesting approach by Zapier’s Wade Foster, basically banning exec DMs to push almost all conversations into public Slack channels, treating the company’s knowledge as a “Shared Brain” for AI. They track an exec “transparency leaderboard” (% of messages in public channels), and record and transcribe meetings. Too much?

🎧 Good listen/watch

[Podcast] PMing at Meta today. Meta PM leader Jagjit Chawla explains how they’ve blown up old-school PM rituals: PRDs are now a paragraph + prototype, status decks are replaced by an overnight agent that reads code, docs, and email, and reviews start at the decision. Ideas and first-principles judgment are the new bottlenecks

[Podcast] How Gusto built a product line with Claude Code, and putting GLM 5.2 to test. Claire Vo shows how Z.ai’s new GLM‑5.2 is strong enough to rival frontier models for many coding and agentic workflows, especially when cost and control matter. She walks through wiring it into Cursor and Claude Code, and stress-testing it. Separate, but on the same episode, Gusto’s Eddie Kim shares how a 5‑person team, a “permanent Zoom” with Claude Code, and almost zero traditional process shipped a tier‑one production agent in 10 weeks by treating AI as a team member

[Video] The future of work with Claude. Anthropic’s product and eng leads explain how Claude Tag lives inside tools like Slack, watching public channels and autonomously running long, multi-day workflows. Tag can self-schedule work and adapt to team preferences. For PMs, this is an always-on agent sitting in the middle of your team’s work, driving PRs, answering questions, and diffusing best practices across the org

🧑‍💻 Worth learning

Loop engineering. Marily Nika explains how to design loops so AI agents don’t fail, over-iterate, or hallucinate when given vague goals and unchecked tools. Designing loops is becoming a core PM skill in a world in which we delegate more and more to AI

39 principles for better human–AI interfaces. This piece turns academic work on human–AI interaction into a practical framework and 39 principles for building AI products users can trust. For PMs, it’s a checklist for when to delegate to AI, how much autonomy to grant, or how to keep assistance useful without becoming manipulative or unsafe

🔧 Products to try

Figma Motion puts animation timelines inside the Figma canvas, so designers can create keyframes, prompt for motion, and hand developers code-ready timing specs; free options

BrowserAct gives your agent a real browser for clicking, filling forms, uploading files, handling verification, and scraping live sites; free limited tier

OpenClaw app. The AI agent tool is now on iOS and Android; free to try

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$5.4B: World Cup betting hits $5.4B across Kalshi, Polymarket, and Robinhood prediction markets

$1.5B: Baseten is raising $1.5B at $13B valuation to expand low-cost AI model services competing with OpenAI, Anthropic on affordability

$650M: Groq raised $650M to scale AI inference cloud to 200 MW by 2027 across 13 global data centers serving 5M developers

$500M: RAISE US launched with $500M+ raised and a goal of $1B, backed by OpenAI, Anthropic, Amazon, Microsoft, and Bank of America, to fund retraining programs for workers displaced by AI, starting in Arkansas, Connecticut, Maryland, and Utah

$320M: General Intuition raises $320M Series B at $2.3B valuation for using gaming to develop embodied AI, robotics, spatial intelligence, pushing world-model research

$320: Airwallex has announced a $320M Series H at $11B valuation to accelerate product development across autonomous finance and agentic commerce, while also expanding its infrastructure and regulatory footprint into new markets

75%: Credit card data shows Anthropic's paying consumer base up 75% since Jan-26, with interest in Claude courses surging 18x in 30 days, gaining ground on OpenAI

5x: OpenAI's Codex grew its active user base more than 5x in the first half of 2026, with non-developer adoption rising 137x for individual users; legal, finance, and recruiting at OpenAI now use Codex as their primary work tool, not ChatGPT

📰 What’s going on

OpenAI holds back GPT-5.6 series and announces a “limited preview” of its new frontier AI models. That includes Sol, the flagship; Terra, a comparable model to GPT-5.5 that eats up only half the tokens; and Luna, an even lower-cost option. For now, only a few trusted partners have access to these models, as requested/instructed by the US government. The models will apparently open up access to a larger population within the coming weeks

OpenAI leadership is having second thoughts about its IPO plans, at least for the remainder of 2026. Sam Altman wants the value to exceed $1T as a public company, and it’s unclear if that’s currently possible amid unpredictable global markets, lingering concerns about future AI revenues, and the downward slide of SpaceX stock following its own blockbuster debut

OpenAI and Broadcom team on AI chip. This marks the first-ever custom AI chip developed by the ChatGPT maker. For now, Jalapeño remains in the testing phase, but Broadcom reports that the chip provides equivalent inference performance while draining about 50% less energy, compared to a standard AI GPU

Anthropic has launched Claude Tag, a new service that embeds an "always-on" AI assistant directly into Slack channels, where it acts as a persistent teammate that learns from ongoing conversations and can complete multi-step tasks autonomously. The feature is now in beta for Claude Enterprise and Claude Team customers. Andrej Karpathy referred to such services as a “self-contained, persistent asynchronous entity with org-wide tools and context, working alongside teams of humans”, which he calls the next iteration of LLM UX

Anthropic accuses Alibaba of the largest-ever Claude distillation attack, alleging Alibaba ran 28.8M fraudulent exchanges through 25K accounts between April-June 2026 to illicitly extract Claude's capabilities and accelerate China's AI development

Google delays Gemini 3.5 Pro to July for token optimization. The company pushed its next frontier model back from June to gather more real-world feedback and fix token consumption issues identified in Flash 3.5

Meta releases new smart glasses without Ray Ban frames, at $299 with camera, AI assistant, 8-hour battery, & 14 live-translation languages

Meta paused its Model Capability Initiative (an internal program that records employee keystrokes, mouse clicks, and screen activity to train AI) after a permissions error made all that sensitive data readable by everyone at the company

Instagram enhances its TV app with longer-form episodic and live formats now on Samsung TVs

Apple announced sweeping price increases across Macs, iPads, and HomePods blamed on chip shortages. The company said it has "never seen a component price increase this much, this quickly." Updated prices:

• MacBook Neo: $699 (up from $599)

• 13-inch MacBook Air: $1,299 (up from $1,099)

• 13-inch MacBook Pro 1TB: $1,999 (up from $1,699)

• Base iPad: $449 (up from $349)

• iPad Air (11-inch): $749 (up from $599)

• Vision Pro: $3,699 (up from $3,499)

Figma announces Motion, a native animation timeline built directly into Figma. Designers can now animate objects on the canvas with keyframes, easing controls, and sequencing without leaving the file or handing off to a separate tool. Here’s an example of one animation that took 4.5 seconds to make

Figma also presented Generative Plugins, which let you ask the Figma agent to build a plugin in plain language without knowing how to code it. The plugin lives in the file and travels with it, so if you hand a file off to another team or a client, the tools you built with go with it

And finally, Figma announced Code Layers. Code now lives directly on the canvas, so designers can import a GitHub repo or use a local folder, and when making changes that impact the codebase, Figma’s agent analyzes what’s been changed and pushes updates back to the main repo

China claims No. 2 global AI lab ranking: New Zhipu AI's GLM-5.2 model trails only Anthropic's Fable 5 on front-end coding benchmarks; founder Tang Jie predicts China will match Fable 5 before year-end, challenging Musk's Q1 2027 projection

Alibaba's new AI video model HappyHorse 1.1 surges to #2 spot in benchmarks, surpassing OpenAI's Sora and ByteDance's Seedance

Notion announced external agent support for Claude and Cursor

Notion will shutter its Mail inbox later this year, arguing that more of its customers are using agents to handle email Consider this (relatively minor) product decision as one more piece of evidence that future software will be built for agents, not humans

SoFi has acquired Composer, a startup that helps investors build and automate trading strategies using AI

📚 Good reads

The AI era requires a different kind of experimentation, by Elena Verna. In an AI-first world, tweaking button colors and tiny UI details is a distraction; the real gains come from bold, system-level experiments, especially around monetization. It’s time to skip minor tests, run bigger, longer-lived bets on pricing and packaging, and to treat AI “best practices” as the default, so scarce eng time goes into experiments that move engagement and revenue over the long run

Trust before it’s earned. Leah Tharin explores why leaders should give trust early, and argues that real empowerment means granting ownership up front, then demanding no surprises through clear, frequent updates. She breaks down how “managing up” with actionable transparency protects trust, and how servant leaders should never punish failure but must still act fast on weak performance. Strong performers are the ones who reduce chaos, create clarity, and quietly lift everyone around them

Why platform product management is the most demanding PM role. Platform PMs juggle two very different “customers” (internal teams and end users), a swarm of stakeholders, and impact that only shows up months later. Because the work is mostly invisible, they also need exceptional storytelling and influencing skills to secure adoption. Treating platforms as products (with JTBD, clear leading indicators, internal GTM, and sane versioning policies) turns a hard role into a high‑leverage one

🎧 Good listen/watch

[Podcast] How Laurel built an AI-native “company OS” for everyone. Laurel’s CPO, Jiaona Zhang, shows how a simple Claude Code “company OS” lets even CSMs ship to production. It also reframes the PM role in AI-native teams (captain model, two-track reviews) and gives a clear framework for proving you’re an AI-native PM in interviews

🧑‍💻 Worth learning

45-min Hermes course: set up your AI chief of staff. Peter Yang introduces a beginner‑friendly course to get Hermes running 24/7 in under an afternoon

Claude Code power skills for PMs. This guide shows how to set up agentic workflows in Claude Code. It covers power habits like tight context management, planning before execution, automatic checks, picking the right model (Haiku/Sonnet/Opus) for each task, and wiring hooks, agents, and /goal loops. A brief operating manual for turning Claude into a reusable execution engine

Anthropic has published a detailed guide on how to manage Claude Tag

🔧 Products to try

AgenticCalling lets you send your Claude or ChatGPT agent out to make real phone calls on your behalf, so you can tell it “call 50 hotels in Miami and find me the best rate” and it handles the whole conversation autonomously; 3 free minutes to start

LocalClicky is a fully offline Mac voice assistant that controls your computer by voice (open apps, move files, edit videos, click things on screen), with zero data leaving your machine; open source, free

That’s all for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$60B: Flutter/FanDuel expects $60B in global legal betting on the 2026 World Cup, 71% more than 2022, with peak demand hitting 100K bets/minute

$60B: SpaceX to acquire Cursor for $60B. Back in April, SpaceX announced an agreement to either acquire Cursor owner Anysphere for $60B or pay the company $10B for its collaborative work on a new AI model. Post-IPO, the companies have made the acquisition official (pending regulatory approval)

$34B: OpenAI losses increased nearly 8X in 2025, with spending hitting $34B

$7.4B: DeepSeek valued $50B after raising $7.4B

$3.6B: Salesforce buys Fin (previously known as Intercom) in a $3.6B deal

1B: ChatGPT hit 1B mobile MAUs

$310M: Odyssey, the world model startup, raised a $310M Series B at a $1.45B valuation. Unlike LLMs, world models map the physical world and attempt to understand how objects move through it. They’re essential for training and directing the self-driving cars, robots, and other hardware controlled by AI brains, but occupying physical space alongside humans

10k: China is executing a 2026 plan to deploy 10K+ humanoid robots across factories, logistics, retail, healthcare, inspection, and emergency response sectors

$2.2k: Snap unveils standalone Specs AR glasses powered by Qualcomm Snapdragon for $2,195, featuring 51-degree field of view, gesture control, shipping fall 2026. Snap stock plumeted 20% after the announcement, which also came with plenty of mockery

49% of US adults now report using AI chatbots. 25% on a daily basis

📰 What’s going on

OpenAI’s Codex has launched a new feature called Record & Replay which lets users show Codex a workflow once and then reuse it later as a skill. Codex can interact with e.g. spreadsheets, web applications, browser tabs, or connected plugins

Codex also launched a dedicated iOS plugin that makes it easier to build iOS apps end to end

Claude Design shipped a change to the components it creates to help bridge the gap between design and development. Product teams can now upload their actual component libraries 8buttons, typography, color tokens, spacing rules,…) from GitHub, design files, or raw uploads. Claude then generates designs using only those components and auto-corrects output before showing it to users. They also launched a redesigned editor, which comes with some new layout controls. These let you drag, resize and align elements directly on the canvas, without having to rely on the conversational interface to make the changes

Claude Design can now send designs directly to third parties like Replit, Miro or Canva

Anthropic is “very confident” Claude Fable 5 and Mythos 5 will become available again “in the coming days,” according to Managing Director of International

Anthropic added centralized enterprise auth for MCP connectors, starting with Okta beta support

Google rolled out Gemini-powered Gmail summaries globally across free, paid, and Workspace accounts on Android, iOS, and web

Google launched Android 17 with AppFunctions, Bubble Bar multitasking, device handoff, post-quantum security, and more Gemini features.

Perplexity⁠ launched Brain for Computer, a continuously learning memory/context-graph system in Research Preview for Max subscribers that starts future tasks with project context and is reported to improve answer correctness by 25%, recall by 16%, and history-dependent task cost by 13%

AWS launches AI agent capabilities at New York Summit. Amazon Bedrock AgentCore adds managed knowledge bases, web search, and monetization tools for content owners to price and meter AI bot access at the edge

Meta launched AI Mode on Facebook, a chatbot-style search that synthesizes answers from public posts, Groups, and Reels instead of returning a list of links, powered by Meta's Muse Spark model

Meta split operations from Manus and began unwinding the acquisition after Chinese regulatory pressure; the founders were reportedly exploring a buyback

Meta’s Emily Dalton Smith is stepping down just two months after being tapped to lead its new “AI for Work” initiative, which was meant to unify internal AI tooling and agents like Metamate

Microsoft ships Copilot Cowork globally with usage-based billing. The agentic system now runs complex, multi-tool tasks end-to-end with spending limits, usage reporting, and model choice

Z.ai released GLM-5.2, an open-weights model with 1M-token context window and strong long-horizon coding. Ahead of Google’s Gemini 3.5 Flash in benchmarks, and just below Opus 4.8 and GPT-5.5, but at a fraction of the cost

Pinterest has teased the launch of “Ask Pinterest,” an experimental standalone app that lets users discover products through natural language conversation instead of traditional visual search. Pinterest built it as a separate app to test conversational shopping without disrupting the main app

LinkedIn is launching “connected apps,” a feature that automatically generates verified descriptions of how you actually use supported software tools and displays them on your profile

LinkedIn is also plannng the roll out a bunch of new ways for creators to make money: a brand‑deal marketplace, paid “experiences” (like 1:1 advice), subscriptions for newsletters, podcasts, and gated communities. It’s also looking into a creator fund and scaling up paid, ticketed events with top creators (moving closer to YouTube/Instagram’s playbook)

Alibaba Cloud expanded in Europe and said new agentic AI services are coming for regional customers later this year

Coinbase’s memo signals a major strategic shift: beyond crypto, toward a full-service financial platform

📚 Good reads

The mom-and-pop SaaS era has arrived. Elena Verna explains how AI is crushing the cost and complexity of building software, so thousands of niche problems that never justified a dev team (or a VC pitch) are finally getting real products (somehow I ended up building a baby tracker). Non-technical, deeply experienced operators (teachers, recruiters, local retailers, coaches, etc.) are becoming builders, packaging 10+ years of domain knowledge into small, profitable tools instead of chasing unicorn-scale TAM

What you’re really selling to customers. Shreyas Doshi argues that great companies win by deeply understanding the essence of what they sell (e.g. Apple selling taste) vs just products / features. He shows how this “one true thing” simplifies decisions, branding, and strategy

The first 90 days of an AI product. The piece argues that the first 3 months after launch are a diagnostic sprint for AI features. It walks through a 30/60/90‑day playbook: watch bypass and adoption patterns, classify failures (missing data vs retrieval vs intent), and separate content issues from latency or formatting

Microsoft’s AI-forward design system. Microsoft is evolving Copilot into a cross‑app “thinking layer” that understands your context, follows you between Word, Excel, and more, and adapts to where you are in the explore‑vs‑focus cycle. A single Copilot (via the Dynamic Action Button, Chat, On‑Canvas UI, or Suggested User Actions) keeps conversations continuous, and underneath, patterns like “Throw & Catch” orchestrate handoffs between entry points

Block explains how they build faster with Builderbot, a Slack-native orchestration layer that coordinates multiple AI agents across all of Block’s codebases. Unlike typical repo-level coding assistants, it understands every service, API, and convention, then picks up tickets, creates branches, writes code, opens PRs, and iterates through CI. It’s powered by their open source agent framework goose and MCP, and already accounts for roughly 15% of Block’s weekly production code changes

Analysis of Y Combinator’s latest batch. CB Insights shares how YC’s latest cohort leans heavily in physical AI (robots, drones, wearables, and space hardware), plus the simulation and data infrastructure needed to train them. The batch also shows agent tooling getting more granular, with specialized infra (routing, safety, continuous learning) replacing “do-it-all” platforms. YC is always a good indicator of “what’s next” (at least from VC perspective)

🎧 Good listen/watch

[Podcast] Anthropic’s Fiona Fung explains how her Claude Code and Cowork teams now ship about 8x more code. She describes an AI-native org where roles blur (PMs, designers, and engineers all ship code). Her biggest focus is culture: giving teams permission to kill bad processes, automate everything Claude can do, and learn together in new rituals like pairwise programming lunches

[Podcast] The model eats the harness. Logan Kilpatrick, who leads Google AI Studio and the Gemini API, explains how Google’s new Antigravity agent harness is becoming the shared “agentic layer” across Gemini, Search, Android and Cloud, predicting that today’s startup obsession with external harnesses has maybe a 12‑month edge before models simply absorb that logic

🧑‍💻 Worth learning

How to use Agent Skills. An in-depth new guide from Google’s AI teams with practical examples. Agent Skills turn any general-purpose AI agent into a specialist on demand. Learn how to build, evaluate, and deploy skills in production

🔧 Products to try

ChatGPT scheduled tasks now has a dedicated management page and faster recurring briefings from your interests, chats, and connected apps; included in eligible ChatGPT plans

Grok in PowerPoint turns prompts or outlines into slide decks, diagrams, images, and data-connected presentations inside Microsoft 365. Free add-in; Microsoft 365 required

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$674B: ASML became Europe’s most valuable company ever, hitting a $674B market cap as analysts bet that demand for its chip-making machines will keep climbing

$20B: Bending Spoons, the Italian holding company owning AOL, Eventbrite, and Vimeo, filed for a U.S. IPO, potentially valuing it at $20B. Projected $1.31B in 2025 revenue, up 95% YoY, with $601M in Q1 revenue

$150M: Anthropic launched Claude Corps, a $150M fellowship that will place 1,000 fellows inside nonprofits for year-long AI coaching

$130M: Canada's Koho just closed $130M at $1.33B valuation in its push to become a full bank

19%: SpaceX shares soared on Day 1, closing out Friday just above $161 per share, a 19% jump over its IPO price. The company is now valued at over $2T, and Elon Musk enters the record books as history’s first-ever trillionaire

📰 What’s going on

Anthropic launches and then blocks new Mythos-class models. Anthropic launched Claude Fable 5, the public version of its new Mythos-class model, plus Claude Mythos 5, the same underlying model with some safeguards lifted for vetted cyber and biology partners. Both models got suspended right before the weekend, citing risks for the models to be used maliciously by foreign nationals. Before the full block, Anthropic walked back invisible Claude Fable 5 safeguards after researchers said legitimate AI, cyber, and bio work was blocked or silently switched to Opus 4.8

Microsoft blocked employees from accessing Claude Fable 5 through internal GitHub Copilot after Anthropic updated its data retention rules, with legal teams reportedly reviewing whether flagged prompts could expose customer data

OpenAI filed for an IPO, on what could become one of the largest public offerings in Wall Street history. The company noted “it may take a while because there are things we want to do that are likely easier as a private company”

OpenAI is considering drastic price cuts on tokens as it braces for a pricing war with Anthropic; both companies are heading toward IPOs and neither wants to blink first

OpenAI expanded API web search so models can look up current information before generating response,

ChatGPT can now turn data and comparisons into charts directly inside ChatGPT

OpenAI agreed to acquire Ona, giving Codex secure, persistent cloud environments for long-running enterprise agents

Apple announced the revamped version of Siri during WWDC. Siri now comes with a dedicated mobile app and the ability to swap models. There were no major updates to iOS, with Apple taking the opportunity to ship some refinements like the ability to configure the transparency settings on Liquid Glass

Apple's foldable iPhone. Hidden iOS 27 developer beta code reveals parameters like "foldState", suggesting Apple’s foldable iPhone might be unveiled this fall

Google announced new capabilities to NotebookLM, which give it coding abilities. Powered by Antigravity, NotebookLM can spin up a sandbox environment, write scripts, and execute them to support deeper analysis (e.g., data cleaning, statistical comparisons, simulations) without you leaving the UI. NotebookLM will come with a set of more than 100 software skills that can help you build workflows with your notebooks that previously would have required you to jump between apps

Google Analytics is also getting some important new updates. Starting this week, GA4 will get its own AI Assistant traffic measurements which lets you track and trend human traffic from top AI assistants including Gemini, ChatGPT and AI Overviews directly in reports

Perplexity plans to pursue a 2028 IPO regardless of whether Anthropic or OpenAI list first

Amplitude released Wave, an AI agent that automates the full product improvement cycle, from identifying what to build to measuring whether it worked

Visa and OpenAI have announced a partnership which sees Visa’s payment network directly into OpenAI’s AI agents, allowing these agents to initiate secure transactions on behalf of users. When an AI agent needs to complete a purchase, it uses tokenized Visa credentials (encrypted payment information) rather than raw card data

Mastercard also launched new agentic capabilities this week. Agent Pay for Machines (AP4M) is Mastercard’s infrastructure for high-speed, automated payments between AI agents and services. Instead of traditional person-initiated transactions, AP4M enables agents to execute continuous, programmatic payments

Coinbase announced an AI agent lets users automate cryptocurrency trades and payments without manual intervention. Users connect the agent to their account and give it natural language instructions, like “rebalance my portfolio” or “execute this trade”

DoorDash released a new AI Assistant that lets users order through voice, chat, recipe link or image. Their CEO says that in early testing, nearly half of all restaurant orders made with Ask DoorDash were from a place the customer had never ordered from before, and grocery baskets built with Ask DoorDash were over 35% larger than those without

Xiaomi open-sources MiMo Code with persistent memory. The AI coding agent maintains context across extended dev sessions through background subagent and automated weekly memory compression, outperforming Claude Code on SWE-Bench Pro

📚 Good reads

Why your AI strategy needs a clear point of view. In fast-moving markets like AI, buyers feel overwhelmed and rely on vendors not just for product demos, but for a confident, opinionated take on where the category is going. April Dunford shows how companies like OpenAI, Anthropic, or Microsoft root their POVs in their unique strengths, turning strategy into a story about the future. As PMs, if we can’t clearly explain our own POV, we make it easier for customers to delay decisions (or choose a competitor)

Using LLMs for analytics. One of PMs new worst fears is getting wrong analyses due to LLM hallucinations. Anthropic explains how they automate about 95% of internal analytics queries with Claude using a multi-layer approach

Becoming irreplaceable in the age of AI. Anne-Laure Le Cunff explains how feeling replaceable hits our deepest needs for belonging, self-worth, and identity, especially when AI can do parts of our jobs faster or cheaper. Her piece argues you can’t be truly irreplaceable as a function, but you can become hard to substitute with your rare mix of talent, taste, and trust. For PMs that could translate into your unique combinations of strategy + communication + domain insight, plus making your thinking visible, and running tiny public experiments so the “version only you can ship” is obvious to others

Emotional intelligence for PMs in the age of AI. Roman Pichler argues that as AI and data tools get stronger, emotional intelligence becomes a real differentiator for PMs. He breaks EI down into self-awareness, self-management, social awareness, and relationship management, showing how each one leads to better products, stronger stakeholder alignment, and healthier teams. He also shares some practical habits, like mindfulness, journaling, active listening, and regular user conversations

🎧 Good listen/watch

Claude Fable 5 review by Claire Vo. Complement with this great write-up by Ethan Mollick

AI eats the world? A reality check with Ben Evans for a16z podcast. They dig into where AI is delivering value today vs what’s still mostly hype. They focus on coding agents as the first real breakout use case, and how infra spend, software economics, and enterprise adoption dynamics shape what comes next

🔧 Products to try

Google launched Gemini 3.5 Live Translate for near-real-time speech translation in AI Studio, Google Translate, and Meet. 70+ languages. No choppy pauses or robotic monotone. It actually sounds like a person

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

€110B: France has locked down €110B to expand its domestic AI footprint

1B: ChatGPT hits 1B users faster than any app. Reached milestone in May, after three years, 62% YoY growth; it took Google Maps, Chrome, YouTube, and TikTok five to eight years to reach the same mark

$500M: Supabase closed a $500M Series F at a $10.5B valuation (200% + since October) and databases on the platform are up 600% YoY,. Claude Code is the single biggest contributor. Their on-app user isn't a developer anymore. It's an agent acting on behalf of a developer through MCPs

$1.5K: Uber has capped usage of AI tools to $1,500 per employee. The idea of tokenmaxxing, which incentivized all sorts of bad behavior, has been short-lived

$200: Meta is considering charging up to $200/month for Hatch, its upcoming consumer AI agent (formerly OpenClaw), putting it in direct competition with top-tier offerings from OpenAI and Anthropic

90%: Pinterest cut AI inference costs 90% by rebuilding Qwen3-VL's vision layer with proprietary embeddings, boosting accuracy 30% for their 620M users

80%: Anthropic says more than 80% of production code merged into its codebase in May 2026 was authored by Claude. The average Anthropic engineer now merges 8x as much code per day as they did in 2024. Claude Mythos Preview sped up model-training code by ~52x, compared with ~3x for Claude Opus 4 in May 2025

57%: Cloudflare’s latest data shows AI agents generating about 57.5% of HTTP requests, overtaking humans far earlier than the company’s CEO expected

44.1%: Spotify’s controversial disco ball logo change drove negative reviews from 30.9% to 44.1%

📰 What’s going on

OpenAI rolled out a new ChatGPT memory system that updates useful context over time and gives users a reviewable summary to steer what gets remembered

OpenAI launches role-specific plugins that can be installed from the Codex plugin directory and are designed to help with niche use cases. Some useful plugins: a design plugin built for turning early ideas into prototypes, and a data analytics plugin

OpenAI's Codex update adds enterprise workspaces. The new Sites feature lets teams build hosted apps without front-end dev. Non-developers are now 20% of 5M weekly users, and are growing 3x faster than engineers

Anthropic expands Project Glasswing to 150+ orgs across 15 countries. Claude Mythos is now securing critical infrastructure in power, water, and healthcare sectors; partners include Samsung, SK Hynix, and NATO, targeting codebases affecting 100M+ people per organization

Anthropic confidentially submitted a draft S-1 to the SEC, giving it the option to pursue an IPO after the review process

Microsoft launches Scout personal assistant. Like OpenClaw, it’s an always-on agent that monitors calendar, traffic, and Teams threads. 3K+ Microsoft employees are already using it. Desktop preview for US Frontier customers this week

Microsoft Copilot Health launches in preview, providing personalized health guidance via AI assistant

GitHub showed the new GitHub Copilot app, Microsoft’s answer to Codex and Claude Code, with access to OpenAI, Anthropic, and Google models

Amazon introduced a new feature that creates AI generated images of products that don’t necessarily exist. E.g., if you want a “shirt with a draped collar”, the results set will show you an AI-generated image of that as you type. Users can then use that image to perform a visual search to find similar items that look like the visual

Google Labs launched Dreambeans: an app that scans your Gmail, Photos, and Calendar, then turns your personal data into a short daily set of AI-illustrated stories

Meta has launched an AI "Business Agent" that can book appointments, close sales, and process payments on behalf of companies, marking the social media giant's biggest push yet into the enterprise AI market. The tool goes live across WhatsApp, Messenger, and Instagram for businesses of all sizes globally

Meta is developing an AI pendant you clip to your shirt or wear as a necklace that records and summarizes your conversations throughout the day, building on the Limitless startup it acquired last year. Testing is expected to start in 2027

Apple approves the first third-party AI agent for Messages for Business. Poke, as it’s called, operates via a per-user fee structure and supports tasks from scheduling to smart home control through iMessage

Perplexity built a real-time system that decides whether each AI query runs on your local PC or in the cloud, cutting inference costs without hurting quality

X launches “React with Video” feature, in their latest attempt to keep up with the endless scroll video apps of the world. The in-app created videos closely resemble TikTok or Instagram “reactions,” with users’ selfie footage hovering over the original content as they provide their analysis and commentary

Zoom launched ZoomMate, an AI teammate that turns workplace conversations into agentic search, custom agents, deliverables, and workflow execution across tools like Salesforce, Jira, Slack, and ServiceNow

Oura launched a 40% smaller smart ring, the smallest ever they claim, to track sleep and stress with extended battery life

China develops world-first brain chip with advanced neural interface capabilities, competing with Neuralink

Trump signed an executive order asking AI companies to voluntarily provide their frontier models to the federal government up to 30 days before public release. The order is aimed at assessing models' "advanced cyber capabilities" and selecting "trusted partners" for preview

📚 Good reads

Your AI strategy has a trust problem, not a tooling problem. Elena Verna argues most companies already have enough AI tools, but their command-and-control org design kills speed by hoarding context and centralizing decisions. She contrasts this with “AI-native” setups like Lovable, where titles matter less, information is open, and sharp ICs have real autonomy (and accountability) to ship fast, reversible bets. The key is high-agency people in trust-based teams

Co-existence and the end of co-intelligence. Ethan Mollick argues we’re moving from chatty “co-intelligence” with AI to agents that act as powerful, opinionated gatekeepers between our work and users. He shows how he wrote his new book largely as a human, but let AI heavily shape the site and even an “AI-only” sales page, A/B-tested across models. Products now have to be designed not just for humans, but for AI intermediaries that will read, judge, and route what we build

Voice is becoming the new UI for digital products. As voice tech gets better, people increasingly skip clunky menus and just say what they want, shifting UX from “operating software” to expressing intent. The article argues that voice will sit on top of traditional UIs, with screens used more for reviewing and approving what AI produces than for driving workflows. Complement with this piece by Allen Pike, who argues that the blocker is latency. Today’s voice agents feel dumb or slow because they rely on tiny models and clunky single-turn flows, but fast-enough models plus streaming, tiny time slices, heavy context caching, and lightweight output formats can already make realtime voice-to-UI tools feel great

How to PM an MCP. Tal Raviv shares how he’s building an MCP connector for a finance platform where you don’t control the model, prompt, or harness, but you’re still on the hook for quality. He walks through designing tools and skills, building an eval pipeline, and more

Mid-career satisfaction is about your inner scorecard, not LinkedIn. Shreyas Doshi argues that many ambitious mid-career folks chase titles, money, and scope, then feel strangely empty once the novelty wears off. What really drives lasting satisfaction is competence, flow, fit, and how you feel on Sunday nights (things your imagined “audience” neither sees nor cares about)

🎧 Good listen/watch

[Video Podcast] AI-first search at Google. Google VP of Search Robby Stein explains how AI Mode and AI Overviews expand what search can do, especially for complex, multi-part questions, while keeping classic results and links central. He walks through how queries are broken into many sub-queries, fanned out across the web and Google verticals, then recomposed into rich, multi-format answers. Stein also leans into an “agentic era” vision where search agents personalize using your data (opt‑in) and quietly act on your behalf (booking, monitoring, and aggregating) without cutting publishers out of the value chain

🧑‍💻 Worth learning

How to build self‑improving Claude skills with evals and memory. Peter Yang shows how to turn workflows into reusable Claude skills that know when to trigger, grade their own output with pass/fail evals, and learn from a running memory log

🔧 Products to try

Miso One: an open-source text-to-speech model that clones any voice from 10 seconds of audio and generates expressive speech with 110ms latency; free to try

LLMTest sits between your app and your AI model, watches your real traffic, then automatically rewrites prompts and swaps in cheaper models to cut costs without breaking anything; free to try

👨‍💻 My work

I’ve been entertained with multiple side projects during my sleepless paternity leave nights.

The main one has been henry, the baby tracker I wish existed. There are tens (maybe hundreds) of baby trackers in the app store. After checking a bunch of them, I noticed how they are generally full of dark patterns to monetize, overly complex, or plain ugly. So I built a very simple, easy-to-use, and free app for myself and any parent who needs it. I will write some articles about how it all came together (here’s a preview), but it has been very Claude Code-assisted It’s now live on the app store, and our pediatrician loves it!

I also took time to rebuild my personal site, including portfolio, media, and a space for my non-FTE work as Product coach and advisor.

And finally, in case you missed it last week, I’ve been messing up with the capabilities of frontier models (mostly Opus 4.8) by vibe-coding 90-feeling adventure games, like this Monkey Island spin-off based on Barcelona, and….the delight or nightmare of recruiters (depending on their day and mood): my portfolio as an adventure game 🙈

I published the game-creation skills on a public repo for anyone to use. Enjoy!

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$65B: Anthropic Raised a $65B Series H valuing it at $965B, with $47B ARR anticipating 130% growth & profitability pre-IPO

$1B: Cognition, the AI coding startup, raised $1B at a $26B valuation with $492M revenue run rate

$1B: Manus weighed raising $1B to buy itself back from Meta after China's regulators ordered the unwind of Meta's $2B+ acquisition

$250M: ClickHouse, the Amsterdam-based data analytics firm, tripled ARR to $250M, valuing at $15B post $400M Series D with 4,000+ customers, eyeing IPO

$113M: OpenRouter raised $113M after weekly volume hit 25T tokens across 400+ models and 8M+ users

$106M: Corgi, the AI insurance startup, raised $106M Series B-1, doubling valuation to $2.7B 3 weeks after $160M Series B, with $378M total raised, expanding products, platform, team amid rapid revenue growth

30%: DuckDuckGo app installs are up 30% after Google’s I/O event last week, with some users expressing worries that Google is “force-feeding” them AI features they don’t want

22%: ClickUp cut 22% of its workforce (about 290 people) and replaced them with 3,000 AI agents, framing the cuts as building a “100x org”; surviving employees are being offered salary bands up to $1M if they create “outsized impact using AI”

📰 What’s going on

Anthropic released Opus 4.8, which is said to be 4 times less likely than Opus 4.7 to let flaws in code pass unremarked. I gave it a try by vibe-coding a Monkey Island spin-off set in Barcelona, with just a few context prompts. It did similar to Opus 4.7, but was better at proactively finding bugs and suggesting improvements

Anthorpic also launched Dynamic Workflows, which lets Claude Code orchestrate dozens to hundreds of parallel subagents to tackle complex, multi-stage coding tasks end-to-end. Instead of Claude processing work linearly in a single pass, it breaks large problems into subtasks, runs them in parallel across independent agents, and verifies results before presenting them to you. Practical use cases for PMs

• Codebase-wide bug hunts: run parallel searches across a service, then have independent agents verify each finding before surfacing real issues in a report

• Large migrations: port codebases across frameworks, deprecate APIs, or change languages across thousands of files with coordinated parallel execution

• Clean-ups: Claude’s head of product Cat Wu says she recently used this new feature to catalogue 100s of A/B test flags and deprecate stale ones. Instead of waiting for Claude Code to investigate each one sequentially, dynamic workflows allowed her to do them all in parallel

and Anthropic confirmed it plans to release “Mythos-class” models with higher intelligence than Opus, but these require stronger cyber safeguards before general availability, expected in the coming weeks

Apple preps AI Camera and Photos for iOS 27: Dedicated Siri camera mode, prompt-based editing, "Reframe" and "Extend" tools for image manipulation; WWDC announcement set for June 8

Sneak peek at new Siri app also reveals Apple’s plans to take on ChatGPT

Apple's AI glasses features revealed: the glasses integrate on-device processing, real-time translation, and contextual awareness for a seamless AR experience

And Google's AI Glasses are almost there. They look promising, but need refinement in battery life and interface to go mainstream

NotebookLM is rolling out automatic Google Drive file sync, so your notebooks stay updated when source docs change

Google launches AI Threat Defense: Autonomous platform merges Wiz, Mandiant, CodeMender, and Gemini to span vulnerability discovery through autonomous remediation; launch partners include Accenture, Deloitte, PwC

Google Pay gets ready for agentic commerce with a Universal Commerce Protocol

YouTube announced that it will now auto-detect and prominently label AI generated content. On Shorts, the label will appear as an overlay and on longform video it will appear directly below the video player, just above the description

Facebook launches a ‘Plus’ subscription that gives you extra features, like spotlight stories. WhatsApp, Instagram, and Meta AI are getting paid subscriptions, too

Mistral turned Le Chat into Vibe, an agentic work assistant with Work Mode, Code Mode, inbox/calendar catch-up, research, drafts, and a VS Code extension

Figma unveiled an update to Figma Make that lets it work using local code bases, much like Claude Code or Cursor. The new update allows non-technical users to make visual changes to a product’s live code base and open up PRs for engineers to review

Replit announced its own design tool, Replit Canvas, and Visa invested in Replit to help developer agents handle payments

Robinhood launches agentic trading and credit cards. Retail investors can now connect AI agents to trade stocks, rebalance portfolios, and make purchases automatically; accounts remain separate from main portfolios with spending limits and manual approval options

Waymo began welcoming riders into its Chinese-made Ojai robotaxi, which was designed to improve robotaxi unit economics

Shift, an AI startup, will clean your home for free to train future robots. The footage becomes high-value training data for future home robots, with promises of heavy anonymization and strict vetting of third-party cleaners

on a similar direction, Human Archive tracks India’s gig workers moves to train robots. It’s layering on custom hardware (caps, tactile gloves) and internal model training to prove the value of its multimodal dataset

China limits overseas travel for top AI talent. New restrictions on overseas travel for leading AI founders, researchers and executives, including staffers from big-name international firms like Alibaba and DeepSeek. These individuals now require specific approval from authorities before leaving the country

📚 Good reads

Essential books for product builders. Lenny shares the first part of a curated bookshelf for PMs. Organized by jobs-to-be-done like communication, execution, strategy, product craft, org design, sales/marketing, productivity, happiness, and even sci‑fi. Good repo for summer reads!

Choosing to stay human in an AI-saturated world. Ethan Mollick argues that a challenge with AI is how it quietly erodes learning, judgment, and authentic voice when we use it as a default autopilot. For PMs, his core ask is to be intentional: decide which cognitive muscles (writing, problem-solving, product judgment) you want to keep human, then design your AI workflows around protecting those

The orchestration tax is burning out AI‑powered devs. Addy Osmani argues that spinning up more AI agents doesn’t parallelize your judgment; you’re still the single-threaded bottleneck who must review, reconcile, and understand the work. He frames this as an “orchestration tax.” Lenny and Dan Shipper cover a similar topic on Lenny’s podcast: agents create demand for “forward deployed engineers” who maintain company‑wide agents and human-in-the-loop automation

Reckoning with the future of AI and our role in it. Jack Clark argues that AI progress is accelerating so fast it can’t be treated like a “normal” technology, pushing individuals and companies into a choice between exploring the future or retreating into denial. He also forecasts AI-run companies, recursive self-improvement and life-extending medicine

Do language models need sleep to think better? A bit of a heavy paper, but the key message the authors propose is: a “sleep” phase for hybrid attention–SSM LLMs, where the model periodically pauses, replays recent tokens multiple times, and consolidates them into fast-weight memory before clearing cache. Extra compute gets used into an offline consolidation step, so online inference latency stays the same while the model can still reason over context it can no longer attend to. A promising approach for long-context and reasoning-heavy LLM features

🎧 Good listen/watch

How an Anthropic engineer lead uses Claude at work and home. Felix Rieseberg shows how he turns Claude into a collaborator, auto‑inventoring his life, building 3D floor plans and live dashboards, or even with a $20 “Claude buddy” device. Fun examples and great tips!

ProductCon NY 2026 session recordings and slides. Talks and panels from Eric Ries, Slack, Asana, Anthropic, CNN, Miro, Amazon Games and more

🧑‍💻 Worth learning

Agentic engineering playbook for MCP. This piece by Pulse MCP explains when it’s worth wiring MCP into your agent workflows, and when a plain Skill or CLI is fine. It walks through pragmatic options for picking and configuring MCP servers (e.g. official HTTP ones, “roll your own”) and shows how shared mcp.json configs unlock portability, auth, and governance. It also previews where MCP is heading next

🔧 Products to try

Runway MCP brings Runway image and video generation into Claude, ChatGPT, Cursor, Replit, and other compatible agents, so you can create media without leaving your workflow; free to try

MagicPath brings a native design canvas into Codex, so you can design and build functional apps with interactive components after installing its agent skill through the CLI; free to try

PollyReach gives your AI agents their own phone numbers so they can make and receive real calls to handle lead qualification, customer support, and appointment booking automatically; free to try

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$5.7B: OpenAI reportedly generated about $5.7B in Q1, nearly $1B ahead of Anthropic, while Anthropic is projected to more than double to $10.9B in Q2

$1.25B: SpaceX's public S-1 filing revealed Anthropic will pay SpaceX $1.25B per month through May 2029 for xAI data center access, and that X lost $595M in ad revenue in 2024 due to advertiser exits

$700M: Hark, the AI hardware startup raised $700M+ Series A at $6B valuation

$355M: Modal raises $355M at a $4.65B valuation, claiming x5 growth since September 2025 and reaching $300M worth of annualized revenue. Modal’s neocloud-ish offering sells AI compute, training, and control products that charge for GPU time only when in use, lowering costs for end-customers

$300M: Decart raised $300M to build low-latency AI infrastructure

$250M: Exa Labs raises $250M at $2.2B valuation to build what they call the “perfect search engine,” optimized for AI tasks rather than upgraded from an old-fashioned browsing-first product (aka Google)

$200M: Mercury raised $200M at $5.2B, 49% up becoming federally regulated bank aiming for listing with $650M ARR

3.2M: Alibaba reported 3.2M robotaxi rides in Q1 2026, up from 1.4M YoY

8K Meta employees received layoff notices this week, and about 7K got re-assigned to new AI-focused teams

500x: Arena’s frontier research found GPT-4-level model quality is now roughly 500x cheaper than it was in 2023

89%: OpenAI and Antrhopic now caputre 89% of AI startup revenue

17%: Intuit plans to lay off 3,000+ workers, about 17% of its workforce, to simplify the company and refocus on AI products

11.6%: Google's Gemini became the #2 AI referral source in Q1 2026, nearly tripling its share to 11.6% and now sending more web traffic than Perplexity, Claude, Meta AI, DeepSeek, and Grok combined

📰 What’s going on

Google’s I/O announcements:

• AI in Search. Google’s first major redesign in 25 years; uses AI to help users formulate questions with smart suggestions beyond basic autocomplete. You can input text, images, files, videos, or Chrome tabs with the UI changing depending on the user’s request.

• Search agents will monitor the web 24/7 for updates matching your criteria, automatically notifying you when relevant information appears.

• Coding on Search; this mode generates custom UI components on the fly - things like interactive visualizations, dashboards, and mini-apps tailored to your task

• Personal Intelligence integrates your Gmail, Photos, and Calendar data so Search understands your context

• New voice capabilities, including a feature called “Docs Live” that acts as a thought partner and co-writer to help users create a first draft of a document

• Gemini Spark: a new personal AI agent that runs on Google Cloud infrastructure rather than your device

• Gemini 3.5 Flash: pitched as their strongest release yet for coding, managing agents, and research projects. In internal tests, 3.5 Flash built its own proprietary operating system from scratch.

• A new version of the Antigravity agentic coding app, which includes an updated desktop version, a CLI tool, and an SDK for custom workflows

• Google AI Studio now builds native Android apps from prompts and lets you pull in Workspace data

Meta releases new Reddit rival: Forum. The app’s release immediately triggered a drop to Reddit stock, which is down around 40% so far in 2026

OpenAI is preparing to confidentially file for an IPO anytime now, targeting a September debut at an $852B valuation, right after Elon Musk lost his lawsuit trying to block the company's restructuring

An internal OpenAI model solved an 80-year math problem by… disproving it. Erdős unit distance conjecture, a discrete geometry problem from 1946, was proven false

Spotify will launch a Beta AI tool in June that creates audiobooks (in partnership with ElevenLabs)

Amazon launches AI-generated podcast feature on Alexa+. Users can ask Alexa+ to create podcast episodes on any topic, with AI researching the subject and generating narrated content in minutes; now available across the U.S

Grok launched Skills, a persistent memory feature that lets you teach Grok a formatting rule, workflow, or personal preference once, and it remembers it across every future conversation, no re-prompting needed

X limits free users. If you use X and don’t pay for the service, watch your volume. You are now limited to “50 original posts and 200 replies per day,” whereas before you could ship as many as 2,400 updates every 24 hours

Figma released a new agent that works directly on the design canvas. The agent understands your design system, components, tokens, and team context, giving it access to information third-party tools can’t reach. The agent handles 3 core workflows: exploring design directions (generate multiple stylistic approaches simultaneously), automating busywork (bulk edits, component swaps, content population), and processing team feedback (summarizing comments, identifying themes, pressure-testing designs from different perspectives)

Airbnb's summer release added car hire, grocery delivery, airport pickups, boutique hotels, and AI-powered review summaries to the app, expanding the platform well beyond home rentals for the second year running

Oura files (confidentially) to go public. The fitness wearable and software comapny was last valued at $11B and had expected revenues of $1B last year

📚 Good reads

After Automation. Dan Shipper argues AI creates more work for humans by flooding the world with generic output and raising the value of taste, context, and judgment

AI artifact catalogs: a smarter way to scale your org’s AI. Instead of chasing the latest AI agent, the piece argues you should invest in vendor‑agnostic “AI artifacts” (skills, MCP servers, plugins) that encode how your org works. These artifacts live in shared catalogs (often just Git repos), and are reusable building blocks for agents. Companies like Ramp and Intercom follow this practice really well

Your job is changing faster than you think. Elena Verna argues that even if you keep your title, your role will be unrecognizable by 2027 as AI, org changes, and shifting expectations reshape what “good” looks like. She suggests treating yourself like a product: decide if you like your current job, build career optionality, and get AI‑native. The takeaway for PMs: assume your current skill mix has an expiry date and start rewiring now, on your own terms

Duolingo’s AI Slack bot makes debugging feel almost fun. Duolingo wired MCP agents into a Slack app that talks to 200+ internal tools, so engineers can triage incidents, answer help-desk questions, and even spin up PRs without leaving chat. Over time they standardized on HTTP MCP servers, added guardrails (role-based access, validation sub-agents, human approval buttons), and dogfooded hard to drive reliability. The result: an AI assistant that ~30% of the company uses weekly with an ~80% upvote rate

AI eats the world: how platform shifts reshape tech. Ben Evans argues that genAI is a classic platform shift: massive capex, commodity models, and a scramble to figure out where the real value and margins sit in the stack. Today’s wins are obvious “infinite intern” use cases (coding, customer support, ops) but the hard part is rethinking workflows, moats and jobs, not writing code

🎧 Good listen/watch

[Video] The “refounder.” Notion’s CEO, Ivan Zhaom, explains how Notion repeatedly “refounded” itself to stay ahead of big shifts: once in Kyoto pre–product-market fit, and again in Cancun after a GPT‑4 shock.. He looks for blurry PM/design/eng roles, barbell engineering teams (super‑junior + super‑senior), and hiring more for taste, agency, and curiosity than experience

🧑‍💻 Worth learning

Anthropic’s Claude Cookbook is a growing library of hands-on recipes for product teams building AI agents, tools, and RAG workflows. It ranges from multi-agent sales and SRE incident responders to data analysts, eval loops, cost/latency optimizations, and memory management

My team of AI agents that work while I’m offline. Teresa Torres explains how she built a small “team” of Claude-powered agents that automatically prep her for podcasts and sales calls, generate follow-up tasks, and deliver a weekly coding retro. She uses Obsidian files to define agent identities, tasks, and scripts, + MacOS scheduling on an always-on Mac Mini

🔧 Products to try

Figma’s new design agent on the canvas. You can generate designs, edit existing files, and create variations from text prompts; waitlist

NanoClaw is a sandboxed alternative to OpenClaw that runs in a container, so your AI agent can't access your entire machine. It went viral, turned down a $20M buyout, and raised $12M seed with the Hugging Face CEO as an angel; free

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great weekend!

Welcome to a new Week in Product!

🎰 The week in figures

$70B: Cerebras, the AI chep leader, had the biggest IPO of 2026 with stock surging nearly 70%, sending the company’s market cap to almost $70B

$750M: Ramp is in talks to raise another $750M at a pre-money valuation of +$40B

$650M: Recursive Super AI, a startup by ex-leaders from Meta AI, Google DeepMind, OpenAI, raised $650M at $4.65B valuation targeting self-improving AI

4K: Cisco cuts 4,000 jobs despite record revenue: Networking giant slashing 5% of workforce to fund AI and cybersecurity investments

94%: Baidu's new AI model cost 94% less to build and outperforms chinese competitors

34.4%: Ramp's May 2026 AI Index shows Anthropic at 34.4% of business adoption, crossing OpenAI (now at 32.3%) for the first time ever

5%: LinkedIn planning to lay off 5% of staff, around 875 people, blaming the need to “reinvent” how it works and building more “agile teams focused on [its] highest priorities”

📰 What’s going on

OpenAI shakes up its product team. Greg Brockman will now handle both the company’s AI infrastructure and its product teams, which are being consolidated into a single org. Now, one group will manage ChatGPT, Codex, and developer-facing APIs

OpenAI is preparing to sue Apple after the ChatGPT integration, announced at WWDC in June 2024, failed to deliver the subscriber growth and product prominence the company expected. OpenAI has enlisted an outside law firm to work through its options, which could include sending Apple a formal breach-of-contract notice

OpenAI has unveiled Daybreak, a cybersecurity platform designed to automate vulnerability detection, patch validation, and secure software development for enterprises and governments. The platform directly competes with Anthropic's Claude Mythos and combines OpenAI's LLMs with Codex's agentic capabilities

Anthropic limits Claude usage as costs spike: Paid subscribers now face separate credit meters for agent tools after companies like ServiceNow and Uber burned through annual AI budgets

Anthropic launches Claude for Small Business: Package includes 15 ready-to-run workflows and integrations with QuickBooks, PayPal, HubSpot, Canva, and Docusign to automate payroll, invoicing, and campaign management for 44% of U.S. GDP that relies on small businesses

AWS launches Claude Platform on AWS. Direct access to Anthropic's native platform through existing AWS accounts with IAM authentication, CloudTrail logging, and full Messages API support

Google preps "Gemini Spark" agent for enterprise: Beta code reveals autonomous agent with cross-app access, scheduled tasks, and inbox management; launching ahead of I/O 2026

Google DeepMind introduced "AI Pointer," a concept that turns the traditional mouse cursor into a context-aware AI collaborator that understands what's on your screen and can act on it, starting with Chrome

Google and SpaceX in talks to put data centers into orbit, part of Project Suncatcher with prototype satellites by 2027

Apple is building AI agent support into the App Store, with a WWDC announcement expected; it would mean AI agents (not just apps) could eventually be found, downloaded, and run directly from your iPhone

📚 Good reads

Daily Active Agents. Baidu’s CEO proposed “Daily Active Agents” (DAA) as the defining metric for the AI era, the equivalent of Daily Active Users for the agent world, and predicted global DAA could eventually surpass 10B

Why bolt.new currently leads AI prototyping tools for PMs. Aakash Gupta runs a detailed bakeoff of Bolt, Lovable, Replit, and v0, finding Bolt the most production-ready and brand-faithful. He also distills some workflows (including baseline design systems, divergent prototyping, PRD‑first builds) that improve results in any tool. The piece closes with practical “which tool + workflow when”, and common mistakes

IC work is the new career flex. Elena Verna argues that, thanks to AI, experienced “high-impact ICs” can now ship end‑to‑end projects solo, having department-level impact without managing people. She explains how AI gives you “average intelligence” across adjacent crafts so you can prototype, validate, and iterate way faster than old coordination-heavy orgs. A proof: the app I just launched (see below)

🎧 Good listen/watch

[Podcast] Build a Claude Code personal OS in 40 minutes. Moritz Kremb shows how he runs email, content creation, and even grocery shopping through a Claude Code–based “personal OS.” He layers it into folders with memory, connected tools (APIs/MCPs/CLIs), reusable “skills,” and proactive routines. The episode also digs into OpenClaw vs Claude Code tradeoffs, reliability issues that pushed him to Claude, and concrete examples like automated grocery runs and a full content pipeline from idea to published video

🧑‍💻 Worth learning

Agentic engineering playbook. This piece explains how agent skills can encode your company’s unique processes, systems, and data into reusable “skills” that coding agents can reliably invoke. Over time, the org’s job may become building, sharing, and maintaining a large skill portfolio that drives serious productivity gains and more autonomous AI work

🔧 Products to try

Codex in ChatGPT mobile helps you steer coding tasks from your phone, review changes, approve commands, and keep long-running work moving while your laptop does the heavy lifting;: included across all ChatGPT plans

Claude Code Agent View gives you one dashboard for managing parallel Claude Code sessions and longer-running coding work, available on paid Claude plans

👨‍💻 My work

Henry, the baby tracker. This is not for everyone, but I hope parents will like it. Turns out sleep-deprived nights looking after a newborn can be productive. I didn’t like any of the baby tracker options on the market. They are overly complicated, trying to monetize too hard, or plain ugly. So I built one (iOS app) with just the things I need, caring for a simple and smooth UX to track my baby KPIs (read Key Poop/Pee Indicators) with one hand in the middle of the night. Includes a voice feature, so you can simply tap your lock or home screen and say what you want to log. It’s on TestFlight, and you can sign up here if you want to try it. And here is the story of how and why I built it. I hope other new parents find it useful!

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great weekend!

Welcome to a new Week in Product!

🎰 The week in figures

$45B: Anthropic’s run rate nears $45B, and its (final?) pre-IPO round could push $50B into a valuation that could reach above $1T (pre-IPO valuation just hit $1.2T on secondary markets)

$7.3B: Chinese DeepSeek is raising $7.3B at a valuation of more than $50B

$6.3B: Long Lake agreed to acquire Amex GBT for $6.3B, explicitly betting that AI will reshape corporate travel

$2B: Moonshot AI, the China-based open-source AI platform, raised $2B at $20B valuation

$1B: Kalshi, the SF-based prediction market platform, closed a new $1B round boosting valuation to $22B

$950M: Sierra raised $950M Series E at a $15.8B valuation for its AI customer-service agents

600M: China's AI usage surpasses U.S. in token consumption. Over 600M Chinese citizens now use GenAI daily, with weekly data usage by Chinese models overtaking U.S. models for the first time

$500M: Periodic Labs is raising $500M at a $7.5B valuation for AI-driven materials discovery

1.1k: Cloudflare cut 1,100 jobs in what CEO Matthew Prince called an AI-first restructuring.

23: Google DeepMind's AlphaEvolve scaled to real-world impact, with the team reporting 23 verified scientific discoveries across chemistry, materials, and applied math in the past quarter

25%: DeepL announced staff cuts of about 25% as machine translation margins collapse under open-weight competition

20%: PayPal has plans to cut around 20% of its staff

17%: Coinbase cuts around 400 jobs, 17% of its staff, in AI-driven restructuring

📰 What’s going on

OpenAI announced Codex for Chrome: lets Codex autonomously drive background tabs on Mac and Windows for deep research, CRM data transfer, and admin-console workflows

OpenAI released ChatGPT for Google Sheets, an add-on that embeds natural language AI directly into spreadsheets via a sidebar panel

OpenAI also released three new audio models designed for developers building voice-enabled applications: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. OpenAI is explicitly pitching the new Realtime-2 model to product teams, describing it as an “interface between people and products”

Anthropic released three new capabilities for Claude Managed Agents:

• “Dreaming” is a scheduled review process that analyzes past agent sessions and memory to identify patterns (things like recurring mistakes, converged workflows, team preferences). It restructures memory to keep it high-signal as agents accumulate experience, enabling self-improvement across sessions

• Outcomes define success criteria as a rubric. A separate grader evaluates the agent’s output against your standards in an isolated context window, then surfaces specific gaps if corrections are needed. The agent iterates until the work meets your bar. Internal testing showed 10-point improvements in task success, with +8.4% gains on Word documents and +10.1% on PowerPoint files

• Multiagent orchestration lets a lead agent break complex work into parallel tasks and delegate each to specialist agents with custom models, prompts, and tools

Anthropic and OpenAI both launch enterprise AI ventures: Anthropic raises $1.5B with Blackstone and H&F for forward-deployed engineers, while OpenAI raises $4B for similar venture at $10B valuation

Google Docs now lets users set persistent custom instructions for Gemini that automatically apply to all future interactions. Instead of repeating preferences like “use bullet points” or “keep it professional” in every prompt, users input these rules once in the Gemini side panel

Google is building OpenClaw-like agent Remy, powered by Gemini

Meta is also planning an OpenClaw competitor product for its massive user base

Spotify has released a Command Line Interface tool called “Save to Spotify”. It lets you connect tools like Claude Code generate podcasts from any text source and then store it to your library

Apple plans iOS 27 as choose-your-own-adventure of AI models, letting users pick preferred assistants. May offer custom Apple Wallet Passes beyond standard boarding passes & tickets.

Apple's camera-equipped AirPods enter early production testing. Prototypes with low-resolution visual sensors for Siri queries (cooking guidance, directions) now in design validation, with longer stems and cloud-indicator LED

Amazon’s new ‘Bedrock AgentCore payments’ preview allows AI agents to “instantly access and pay for what they use, such as web content, APIs, MCP servers, and other agents.” We’re this close to being able to charge agents per-read fees to access content, and shake up the online content market for the rest of time

📚 Good reads

The AI 100. CB Insights shares an interactive database with the most promising AI startups of 2026

AI godfather Yann LeCun pushes back on AI doom and hype. LeCun says CEOs exaggerate AI’s power and risk. He argues AI won’t wipe out 20% of jobs, but will turn most of us into “bosses” of AI agents, making strategy and critical thinking more valuable. For PMs, his take is that this wave looks a lot like past tech revolutions: powerful new tools, not an imminent existential threat

Why nobody has nailed the personal AI agent yet. Peter Yang reviews OpenClaw, Hermes, Claude Code, Codex, and Gemini against ten must-have “AI chief of staff” capabilities, and finds that none fully delivers. Each nails different pieces — from OpenClaw’s flexibility to Hermes’ reliability, Claude’s personality, Codex’s desktop experience, and Gemini’s Workspace access — but all stumble on reliability, coverage, or platform gaps. His takeaway for PMs: pick one or two agents that fit your workflows today, ship real work with them, and avoid endlessly migrating in search of a perfect stack that doesn’t exist yet

Agent-native Product Management is changing PM work fast. Marcus Moretti explains how Product Management is starting to run everything through AI agents, especially Claude Code. He introduces two core “compound engineering” skills: /ce-strategy to co-write and refresh a product strategy, and /ce:product-pulse to generate daily, founder-style health reports from your data stack

🎧 Good listen/watch

Uber’s CEO on agentic coding and why product people are making changes to Uber’s codebase. “Now, they’re just going in and they are vibe coding, and an engineer is going to review the code, but essentially the product person is going directly into the code base, so to speak, or going directly with an agent into the code base.” Uber’s CEO on replacing policies with outcomes, why agentic coding is changing who writes the code, and the bet that there won’t be one physical AI model to rule them all

Why agency will matter more than skills in the AI era. Max Schoening (Head of Product at Notion) argues that as AI makes the first 10% of any project basically free, the key skill for PMs becomes agency, the willingness to tinker, prototype, and change the world around you. In a world where software volume is exploding but quality isn’t, PMs who build taste, get their hands dirty with AI, and obsess over reliability and shipping will be the ones who stand out

🧑‍💻 Worth learning

Cognitive Bias Index is an interactive reference for identifying and mitigating cognitive biases; quickly browse or filter dozens of documented biases by context and tag to understand how they arise, how they distort judgment, and how to avoid them; a free web tool curated by Jon Yablonski

30 Claude Code tips to seriously level up your workflow. shares 30 tricks for using Claude Code as a real workbench. She shows how to customize Cursor, see and control what Claude is doing, and feed it cleaner context from docs, code, and the web. Also looks into longer-running, multi-device workflows (loops, mobile, remote control) so Claude can help you ship product work faster

Live coding session. In this live stream, Boris Cherny (Head of Claude Code) and Jarred Sumner (creator of Bun) pair up to build stuff with Claude Code using their everyday workflow. A practical tour of what “AI-native” coding looks like in 2026, and how PMs can think about workflows

🔧 Products to try

Perplexity released Personal Computer for Mac, a desktop app that browses, takes actions, and runs research workflows on your behalf locally; free during beta

GPT-5.5 Instant is OpenAI's new ChatGPT default that delivers 52% fewer hallucinated claims on medicine/law/finance prompts and 30% more concise responses, free to try for all ChatGPT users

Saperly is the first phone carrier built for AI agents; provision a real number in seconds via any MCP-compatible agent for unified calling, SMS, and stable caller ID with audit trails; first number free for 30 days, then $2.50/month + usage

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great weekend!

Welcome to a new Week in Product!

🎰 The week in figures

$1T: Anthropic crossed a $1 trillion valuation and officially became the most valuable AI company on Earth

$300B: Q1 2026 witnessed record $300B in VC funding globally driven by explosion of investments in AI-driven startups -a historic high

$1.1B: Ineffable Intelligence, founded by ex-Google DeepMind researcher David Silver, raised $1.1B at $5.1B valuation to build reinforcement‑learning–based “superlearner” systems that learn from their own experience instead of human data

💸 Q1 earnings

• Alphabet: $109.9B Q1 revenue, 22% up YoY; $62.6B Q1 net profit, 81% up YoY; sales increased across Google Services (Search +19%, subscriptions/platforms/devices +19%, YouTube ads +11%) and Google Cloud surged 63% to $20.0B, driven by enterprise AI solutions and infrastructure; EPS rose 82% to $5.11; operating income grew 30% with margin expanding to 36.1%, and Alphabet raised its quarterly dividend 5% to $0.22

• Amazon: $181.5B Q1 revenue, 17% up YoY; $30.3B Q1 profit, 77% up YoY; sales increased 17% to $181.5B; growth driven by 28% AWS revenue increase to $37.6B, strong international segment sales (+19% YoY), and higher other services (notably advertising), while profitability was boosted by $16.8B in pre-tax gains from the Anthropic investment and continued margin expansion across all segments

• Apple: $111.2B Q2 revenue, up ~17% YoY; ~$31.9B Q2 profit, up ~22% YoY; iPhone sales increased ~21% to $57.0B; growth driven by “extraordinary” iPhone 17 demand, record $31.0B Services revenue, and stronger-than-expected China performance at $20.5B, partially offset by ongoing global memory shortages pressuring future margins

• Meta: $56.3B Q1 revenue, 33% up YoY; $26.8B Q1 profit, 61% up YoY; sales increased 33% to $56.3B, driven by 19% higher ad impressions, a 12% rise in average ad prices, and continued growth in Family of Apps, partly offset by higher R&D and data center investments as Meta ramps spending for AI and infra

• Microsoft: $82.9B Q3 revenue, 18% up YoY; $31.8B Q3 net income, 23% up YoY; Microsoft Cloud revenue increased 29% to $54.5B, with growth driven by 40% Azure and other cloud services growth, strong Microsoft 365 (commercial +19%, consumer +33%), and AI revenue surpassing a $37B annual run rate

📰 What’s going on

Uber adds hotel bookings powered by agentic AI: 700,000+ hotels now bookable in-app via Expedia partnership, with development time cut from 12 months to 6 using AI-powered engineering workflows, and including AI voice-bookings

Anthropic releases Claude for CAD connectors: Text-to-CAD now available in Autodesk Fusion and Blender, enabling natural language commands to generate manufacturable geometry in seconds

OpenAI misses sales and user targets, and saw its stock tumble on Tuesday. Cloud companies sharing high-profile relationships with OpenAI, including Oracle and CoreWeave, also suffered from some of the impact, as did chip makers like AMD and Nvidia

OpenAI is developing smartphone-centered AI agents to integrate advanced AI capabilities with Qualcomm chips

Microsoft and OpenAI have announced an amended agreement that simplifies their partnership while giving both companies more flexibility. The updated terms allow OpenAI to serve its products across any cloud provider, not just Azure, and make Microsoft's license to OpenAI's technology non-exclusive for the first time

xAI launched Grok Imagine Agent Mode (beta). One prompt, one infinite canvas, and the agent plans, generates, edits, and iterates on its own until it is done. Films, manga sets, product ads

xAI held talks with Mistral about teaming up to take on OpenAI and Anthropic

Cursor launched its SDK this week. The same agents that power the Cursor app can now run inside your own products, pipelines, and backend systems. Rippling and Notion are already using it to catch bugs and open fixes automatically without anyone prompting anything

Amazon launches Quick desktop app with workspace integrations: New AI assistant connects to Zoom, Google Workspace, Microsoft 365, and Slack for email writing, calendar management, and document editing, plus local file access and custom app creation

Apple will be launching iPhone Ultra and MacBook Ultra next year, as high-end foldable, OLED models

Apple will add new Photo Tools in iOS 27

China blocked Meta's $2B acquisition of Manus after a year-long probe, the highest-profile AI deal vetoed on national security grounds since the chip wars began

Stripe announced Checkout Studio, a visual builder that lets you design, test, and optimize payment checkout experiences without writing code. It combines a drag-and-drop configuration interface with an AI assistant, live transaction replay, A/B testing, and analytics

Stripe also announced Link Agent wallet, letting agents spend on your behalf using the Machine Payments Protocol. Link functions as a traditional digital wallet, and stores checkout details like billing and shipping addresses. When you grant an AI agent access to Link via OAuth, the agent can request permission to spend money

Spotify and Peloton expand joint fitness content, opening new revenue streams for both platforms

Spotify has also launched a “Verified by Spotify” badge system to help listeners identify authentic human artists amid rising AI-generated music

GitHub Copilot moved to usage-based billing, ending the flat-fee model and pricing requests by token consumption; the same shift Atlassian and HubSpot announced last week

Linear has launched a new feature called “Linear Releases” which connects your CI/CD pipeline to Linear, automatically tracking where each issue stands in deployment. When code changes reach production, issue statuses update automatically, so you see what’s actually live to customers

📚 Good reads

The shiny toy syndrome. Why we keep chasing new tools, courses, and stacks instead of doing the work that matters. Anne-Laure Le Cunff explains how our dopamine-hungry brains mistake novelty for real progress, fragmenting attention and performance

Stop asking your data questions. Julie Zhuo argues most teams obsess over answering explicit data questions, while the value comes from systems that proactively surface what you didn’t think to ask. She lays out 3 “push” layers: smart alerts, daily briefings, and an ambient discovery feed. Together they replace dashboard-checking rituals

Direction over speed. Leah Tharin argues that cheap shipping and AI tools have exposed how many teams are just building the wrong things, faster. She redraws team shape around tiny pods where EMs stay deeply technical, PMs own alignment and “sideways” storytelling across sales/marketing/growth, and PM:engineer ratios depend on how measurable quality is. Overall, AI‑native orgs don’t ship more; they ship fewer but better bets with growth-minded PMs and engineers

🎧 Good listen/watch

[Podcast] Reinvention is now a product’s must-have career skill. Nikhyl Singhal argues that today’s winners are hands-on builders with strong product taste and executive presence, who keep reinventing instead of waiting for things to “settle.” He urges PMs to mourn lost roles quickly, lean into their domain “bugs” as differentiating features, and get back into motion by actually building with AI

[Podcast] Snapchat CEO on why distribution now beats product. Evan Spiegel argues that in consumer social, distribution has become the real moat, with TikTok and Threads winning by hacking reach rather than pure product. He explains how Snap survived rampant cloning by betting on hard-to-copy moats, like creator ecosystems, AR hardware, and a tiny design-led org that ships hundreds of ideas a week

🧑‍💻 Worth learning

Hermes starter kit. Hermes is an open-source agent that auto-writes and refines its own skills. Aakash Gupta walks through a 20-minute setup (Linux/macOS/WSL2 + any major LLM) and 3 PM use cases: self-updating competitive briefings, a queryable customer-signal log, and decision logs with “revisit if” conditions

A workflow for vibe-coded iOS apps. Jaryd Hermann argues the hardest part of app building is the build chain (getting a blank app reliably running on simulator and device). He lays out a 3‑phase workflow (accounts/tools, a “Greenfield” boilerplate app, then small edits and TestFlight builds) using Expo, Cursor, Supabase, PostHog, and RevenueCat. A useful template / model to get vibe-coded apps out fast

🔧 Products to try

The Gemini app can now generate downloadable Google Docs, Sheets, Slides, PDFs, Word, Excel, and CSVs directly from prompts, so you can go from brainstorm to shareable file without leaving the chat; free with Gemini

Lovable's mobile app lets you vibe-code full-stack web apps from your phone using natural language; it builds 20x faster than coding by hand; free to try

👨‍💻 My work

I recently revamped my personal site and added some portfolio-style cases and links to my content. Sharing in case other PMs find it useful! Thanks Zania (zania.design) for the inspo!

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great week ahead!

Welcome to a new Week in Product!

🎰 The week in figures

$1T: Anthropic quietly hit a $1 trillion valuation (!!) on secondary markets, passing OpenAI

$60B: SpaceX preempted a $2B Cursor fundraise (at $50B valuation) with a $10B collaboration fee plus an option to acquire Cursor for $60B later this year

$10B: Bezos co-founded AI startup Project Prometheus raises $10B at $38B valuation for physical AI in manufacturing and aerospace, exploring a $100B fund

$400M: Polymarket, the prediction markets platform, is talking to investors to raise $400M in funding at a valuation of about $15B

271: Firefox fixed 271 vulnerabilities found by Claude Mythos

75%: Sundar Pichai noted that 75% of all new code at Google is now AI-generated and approved by engineers, up from 50% last fall

44% of online buyers start their journey in an LLM or split their search between AI tools and traditional search engines. The study also found 50% of online shoppers in the survey trust generative AI for initial research and product comparisons

16%: Adobe report reveals enterprise AI readiness gap: Only 16% of organizations have deployed agentic AI for customer support despite 78% planning full rollout within 18 months, with 75% citing data quality as top barrier

10%: Meta announced plans to lay off 10% of their workforce (around 8,000 people), while also leaving vacant 6,000 positions for which they were originally going to hire new staffers. The cuts are intended to help management “run the company more efficiently” and also, notably, “to allow us to offset the other investments we’re making” (referring to chips and data centers)

7%: Microsoft offers “voluntary retirement” to 7% (8.750) of employees

📰 What’s going on

Anthropic is investigating a possible breach of Mythos, which reportedly occurred in a third-party vendor environment. Mythos was shared only with major companies, including Amazon, Apple, Cisco, JPMorgan Chase, Nvidia, and now also Microsoft. MythosWatch gives you a live public ledger of 51+ entities with confirmed access

Gemini app launches on macOS: Google released a native desktop app for Mac (macOS 15+) with screen-sharing capability and Option + Space shortcut access, letting users get AI help on local files without switching windows

Google rolled out AI Overviews to Gmail for Workspace (ask natural-language questions about your inbox and get summaries pulled from multiple emails; on by default for Gemini customers), added generative AI features to Google Maps for enterprise (prompt-create realistic Street View scenes, Veo animation, satellite imagery analysis in BigQuery, pre-trained Earth AI models for detecting bridges and power lines), and made Workspace Intelligence turn AI into an office intern across Docs, Sheets, Gmail, and Chat

Google also launched Deep Research Max, a new autonomous research agent that topped benchmarks for web research, reasoning, and locating hard-to-find facts

OpenAI has released workspace agents: AI assistants that handle complex, ongoing tasks for teams. Powered by Codex, these agents run in the cloud and can execute multi-step workflows across connected tools and systems

OpenAI has released GPT-5.5, a model built for agentic work that can plan multi-step tasks, use tools, and check its own output with less supervision than previous versions. It handles coding, research, and document creation, and can navigate ambiguity without constant hand-holding. While more capable, GPT-5.5 matches GPT-5.4’s per-token latency and uses fewer tokens to complete tasks, making it both smarter and more cost-efficient for enterprise deployments. Rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with API pricing set at $5 per 1M input tokens and $30 per 1M output tokens. Ethan Mollick brings a great review

OpenAI released ChatGPT Images 2.0, a new image model with thinking capabilities that can generate multiple images at once, render dense text accurately, and support aspect ratios from 3:1 to 1:3

OpenAI launched ChatGPT for Clinicians (free for verified US physicians) with advanced models, clinical search, reusable skills, CME credit, optional HIPAA, and more

Meta rolled out its Model Capability Initiative, tracking US employee keystrokes, mouse movements, and screen content on Google, LinkedIn, Wikipedia, GitHub, Slack and other approved apps to train AI agents on real office workflows, with "dystopian" internal backlash and no opt-out

Tim Cook will be stepping down as CEO of Apple and taking on the role of Executive Chairman. Senior VP of Hardware Engineering John Ternus will take over the top job. Under Cook, Apple’s market cap rose from $350B to a healthy $4T

Figma has released a new way to view Figma Make creations in the mobile app. For mobile product teams, this makes testing new prototypes on mobile, with swipes, clicks and other mobile native gestures, easier

Yelp is giving its AI Assistant chatbot a major upgrade, positioning it at "the center of the app experience" with the ability to answer questions, make recommendations, and handle bookings. The chatbot, which debuted in 2024 as a tool for hiring service professionals, will now span every category on Yelp through a new dedicated "Assistant" tab

GitHub has paused new Copilot sign-ups and tightened token limits, saying agentic, long‑running AI coding workflows now cost more in compute than fixed monthly fees cover

X shuts down Communities. Head of product Nikita Bier announced plans to terminate the feature, arguing that it’s under-used and plagued by spam: “Communities had a great vision, but they were used by less than 0.4% of users—yet contributed to 80% of spam reports, financial scams, and malware on X”

Lenny just expanded his Product Pass, so paid subscribers now get a free year of 25 tools (worth over $30K), including new additions like Cursor, Google AI Pro, Notion Business, Supabase, Fin + Intercom, and Gumloop. It’s a great deal (I’m making good use of it!), plus Lenny’s content is always top

📚 Good reads

The busyness trap. We cling to constant activity because it feels productive, but most of it doesn’t really move the needle. This piece by Anne-Laure Le Cunff breaks down why “crazy busy” is a cultural badge of honor, and how that masks shallow work and avoids the harder question of what truly matters. It offers four simple habits to help you reclaim focus and make space for deeper product thinking

Why MCP is becoming the backbone for serious AI agents. Anthropic argues that once agents move to the cloud and touch real production systems, ad‑hoc API hooks and CLIs stop scaling, and MCP becomes the shared protocol layer that actually holds everything together. They walk through patterns for designing “agent-friendly” MCP servers (intent-based tools, thin code-execution surfaces for huge APIs, rich UI semantics, standardized auth) and context-efficient clients (tool search, programmatic tool calling)

Explore and compare every design tool. This site curates 80+ modern design tools and regularly highlights new launches, especially around AI-native canvases and design-to-code workflows. A cool, living, watchlist of design tools and specs

Your product has a new user. It’s not human. Elena Verna explains how, as AI agents start using software via protocols like MCP, products must expose robust, well-documented, machine-friendly interfaces. B2B especially will split between “minimum lovable products” humans enjoy and infrastructure that agents rely on

🎧 Good listen/watch

How Anthropic Product teams moves faster than anyone else, with Cat Wu (Head of Product, Claude Code). Claude Code team ships “research preview” features constantly, keeping a standing launch room, and empowering ICs to take ideas all the way to production. PMs enable fast, daily experiments, and are comfortable orchestrating fleets of agents instead of doing the work themselves

What Claude Design is actually good for (and why Figma isn’t dead yet). Claire Vo walks through Claude Design, GPT Images 2.0, and Google’s DESIGN.md. Claude Design is brilliant for fast marketing pages, decks, and wild explorations, but Figma is still relevant for precise, collaborative design. For PMs, these tools are now legit for production-adjacent work, but you’ll quickly hit cost/credit limits and still need real design craft to land in your product

How to build for AI agents, and a Claude Code second brain. Ryan Wiggins (VP Product at Mercury) shows how to treat APIs and MCPs as the main “interface” for AI agents. He demos Mercury’s MCP for banking data, then walks through his Claude Code “second brain”

🧑‍💻 Worth learning

DESIGN.md: A new open standard for AI-generated UI. Design.md is a structured markdown spec that combines human-readable design rationale with machine-usable tokens, so agents can generate and update UIs while staying on-brand. A new CLI lets agents validate their changes (including accessibility checks) against the spec, and Design.md can become a shared contract between tools

🔧 Products to try

OpenAI launched Workspace Agents: five Codex-powered templates (Software Reviewer, Product Feedback Router, Weekly Metrics Reporter, Lead Outreach, Third-Party Risk Manager) deployable to ChatGPT or Slack, free until May 6, then metered

Claude's new everyday connectors hook AllTrails, Instacart, Audible, Booking.com, and TripAdvisor into your chat; find a hike, order groceries, book a trip, or pick an audiobook without leaving the window. Free with any Claude plan

Flipbook is an infinite visual browser created by Zain Shah, Eddie Jiao, and Drew Carr that generates pages entirely as images on demand. When you click anything in an image, the system generates a new image exploring that topic in deeper detail. No underlying HTML, code, or pre-built links; just AI-generated pixels that represent information retrieved from web searches and the model’s training knowledge. Free

👨‍💻 My work

Product in Spain: what’s in demand and what’s not. Second issue of my new Spanish publication, this time looking at the PM job market and the impact of AI

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great week ahead!

Welcome to a new Week in Product!

🎰 The week in figures

$10.8B: Amazon is acquiring Globalstar for $10.8B to compete with Starlink. Their 2028 service launch includes Apple partnership for emergency texting on iPhones

$250M: Glydways $1B, the San Francisco-based autonomous transportation startup, is in talks for $250M at $1B valuation after $170M funding round

$150M: Factory raised $150M at a $1.5B valuation for autonomous coding agents that switch between models by task complexity

$100M: Slash, a Ramp competitor started by teenagers, just raised $100M at a $1.4B valuation

56%: ChatGPT’s web traffic share has dropped from 77% a year ago to 56% in March

40%: Business accounts now represent 40% of OpenAI revenue (up from 20% in 2024) as company deprioritizes consumer products like Sora to compete with Anthropic's enterprise dominance

30%: Vercel’s CEO said that 30% of the web apps it hosts were developed by AI Agents

16%: Snap cuts 16% of its staff, blaming layoffs on AI. CEO Evan Spiegel, who plans to drop around 1,000 employees, said “we believe that rapid advancements in artificial intelligence enable our teams to reduce repetitive work, increase velocity…”

📰 What’s going on

Anthropic has released Claude Opus 4.7, its latest flagship model. The model is now available across all Claude products, API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Users report being able to hand off their hardest coding work to Opus 4.7 with confidence, as the model handles complex tasks with rigor and can verify its own outputs

Anthropic just rolled out Claude Design, an experimental tool that turns natural language into prototypes, decks, and one-pagers, aimed squarely at non-designer founders and PMs. It can read your design system, codebase, and files to keep outputs on-brand, then hand everything off as PDFs, URLs, PPTX, or editable projects in Canva. In research preview for paid tiers. Figma stock was down shortly after

Anthropic redesigned Claude Code on desktop. Now you can run multiple agents and review their work in a single window. Brings an integrated terminal, in-app file editing, HTML/PDF previews, and a faster diff viewer; all in a drag-and-drop layout that you can arrange the way you want

OpenAI unveiled GPT-5.4-Cyber, a fine-tuned variant of its GPT-5.4 model designed for defensive cybersecurity work. The release comes just one week after Anthropic announced Claude Mythos Preview. Access is restricted to vetted security vendors, organizations, and researchers through OpenAI’s Trusted Access for Cyber program

OpenAI just shipped a major Codex update, bringing new agentic capabilities to its coding platform. New features include: computer use, in-app browser, 90+ plugins, image generation, memory, thread automations, and more

OpenAI acquires fintech startup Hiro. Hiro was on its way to shutting down, and a number of Hiro staffers are moving over to OpenAI. Hiro was formed in 2024, and launched its first product (a consumer-facing AI-powered financial planning app) in late 2025

Perplexity released Personal Computer, giving the Mac app the ability to read and write your local files and drive iMessage, Mail, and Calendar (rolling out to Max and the waitlist)

Perplexity also launched Perplexity Health, which helps turn your health data into custom plans and live apps

Apple reportedly testing designs for upcoming smart glasses. 4 frame styles, photo/video, Siri, no display. 2027 launch

Apple is shipping its Siri programmers to a multi-week AI coding bootcamp; fewer than 200 people, two months before the expected major Siri revamp

Google merges NotebookLM with Gemini. The integration adds persistent memory, custom AI instructions, and synchronization across platforms

Adobe launches Firefly AI Assistant with multi-app orchestration: Conversational interface executes complex creative workflows across Photoshop, Premiere, Illustrator and Lightroom from single prompts, with public beta launching in weeks

DeepL releases voice-to-voice translation suite: real-time voice tools for meetings, mobile conversations, and frontline workers, plus API for call centers and custom apps

Tubi became the first streaming service with a native ChatGPT app; install it from the ChatGPT app store and type @Tubi to ask things like “a movie that feels like a fever dream but isn't horror” and get curated picks from 300,000+ titles (free)

The so-called “token-maxxing” trend is beginning to backfire. Meta has pulled its internal leaderboard ranking engineers by token spend. At Amazon, after a manager told engineers to use AI more, one of them inflated token usage by making each chat interaction use 10x the number of tokens than were actually required

📚 Good reads

A Google engineer automated 80% of his own job using Claude Code, and shared the exact system he built to do it

Uber shares how AI prototyping is changing how they build products. Instead of asking teams to imagine how something might work, PMs or designers put a concrete, interactive artifact in front of them. The mass adoption of AI at Uber has led to the CTO admitting that their 2026 Claude Code budgets have been spent already

Inside Lenny Rachitsky’s demanding chill life. First Round Review explains how set out to design a “chill” life, but ended up building a relentlessly high‑craft, high‑leverage product around himself. Underneath the calm tone is a founder-grade operator who treats every artifact as a product that must solve real JTBD for PMs better than anyone else. A great, inspiring read for PMs and creators!

What’s your time to magic? argues that once you’ve validated a big enough vision, the real game is how fast users experience a “whoa” moment in your product. Instead of trying to de-risk by building the whole vision or pointing at competitor launches, she pushes PMs to define the smallest, sharpest magic moment and ship it quickly. In an AI world where anyone can ship similar features, she says the edge comes from taste and the invisible work behind a single click that makes the experience feel truly magical

Dealing with work anxiety. Many PMs quietly deal with work anxiety even when they like their jobs, driven by high internal standards, uncertainty, and blurry work-life boundaries. Instead of “just relaxing,” suggests small experiments using the CALM toolbox: cutting stimulants, taking active breaks, leaning on trusted people, and monitoring patterns through journaling. By treating anxiety management as a series of tiny experiments, you can make your relationship with work more sustainable without sacrificing ambition

Sharing AI context and skills across your devices. describes how, as her Claude Code workflows grew, keeping context files and skills in sync across her Mac, phone, and with her admin became a real bottleneck. She walks through experiments with Dropbox, iCloud, GitHub, and ultimately lands on Obsidian Sync as the “good enough” shared source of truth

Building your AI Chief of Staff. shows how he built a persistent, context‑rich “personal operating system” in Claude Code using folders and markdown files. He layers identity, thinking models, writing voice, working preferences, knowledge graph, and meeting notes so the AI reasons like him and gets move valuable over time

🎧 Good listen/watch

[Podcast] What PM hiring managers actually screen for. Inside the interview loops at top tech orgs and a breakout AI startup. By , an including Mckenzie Lock, GM at Netflix; Sam Stone, VP of Product at EvenUp; and Sarah Koo, Senior Director of Product at Rippling, for The Skip podcast

🧑‍💻 Worth learning

Automate work with ultraplan and routines

Anthropic just shipped Claude Code Routines, and as Nick Saraev walks through in his demo, it is a 1-to-1 replacement for n8n, Make.com, and Zapier. The old way: drag-and-drop nodes, set up credentials, map variables, debug for hours. The new way: describe what you want in plain English, hit save, done. Here’s the 5-step setup for non-technical users:

1. Go to claude.ai/code/routines and click “New routine.”

2. Name it (e.g., “Morning email triage”).

3. Write the prompt like an SOP. Routines run hands-off, so be more precise than a normal Claude chat. Example: “Pull my Gmail unreads. For each one, check for prior conversations with that contact. Draft a reply. Send me the drafts in Slack.”

4. Pick a trigger: schedule (daily at 5 AM), webhook (a URL that other apps can “ping” to fire the routine automatically; e.g., a new Gmail arrives → routine runs), or API call (an API is how programs talk to each other; “call” means any script or tool you run can kick off the routine on demand from anywhere).

   5. Add connectors under Settings → Connectors. Sign in to Gmail, Slack, Notion, etc. (uses OAuth, the same “Log in with Google”-style flow you’ve clicked through a hundred times) so the routine can read and write in those tools.

Hit “Run now” to test. If it works, walk away. The routine fires forever, no laptop required. Complement with these tips by Aakash Gupta

Anthropic also shipped Ultraplan, a cloud-based planning mode that decouples thinking from doing: instead of one local agent reasoning sequentially, Ultraplan spins up three exploration agents and one critique agent on Anthropic's infrastructure, points them at your synced GitHub repo, and returns a structured blueprint before a single line gets written. Your job becomes art director: review the plan, push back on what looks wrong, approve, then let local execution run

🔧 Products to try

Gemini for Mac is Google's brand-new fully-native Swift app that lets you share your screen or local files with Gemini in real time and get help with whatever's on it; free with a Google account

Google Chrome Skills turns any Gemini prompt into a one-click reusable workflow you run on the current tab (or multiple tabs) via the Chrome sidebar. Ships with 50+ premade recipes for things like side-by-side shopping comparisons and contract scanning; free, rolling out to English (US) desktop now

getdesign.md: Design system inspirations from popular websites. Drop one into your project and let coding agents build a matching UI

👨‍💻 My work

Corresponsal en Nueva York. Making the best of my newborn night duty hours, I started a new publication for my Spanish readers. Longer form reads in which I analyze trends in the world of Product and tech with a practical angle for the Spanish market. In the first issue I cover product job market trends and AI-led shifts

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great week ahead!

Welcome to a new Week in Product!

🎰 The week in figures

$30B: Anthropic triples revenue to $30B. Jumps from ~$9B at year-end 2025 to $30B ARR in 4 months, with 1,000+ enterprise customers now spending $1M+ annually

$450M: Perplexity hits $450M in ARR, growing at more than double the rate of the previous quarter. The pivot away from search and toward Computer (Perplexity’s agentic workspace), along with a shift to a use-based pricing model, has given the company a major boost. Their user base reportedly now exceeds 100M

$130M: Spain’s Xoople raised a $130M Series B to build a satellite data business designed to generate better Earth data for AI models

📰 What’s going on

Claude Cowork is now officially available on all paid plans after a few weeks in PReview. The official release comes with new features, like the ability to control which tools Cowork has access to, as well as analytics features. In one shared example, Zapier connected Cowork to Slack, Jira and their internal databases to surface engineering bottlenecks and build a custom dashboard that their PMs and designers could copy

Anthropic launched Claude Managed Agents, a new suite of APIs that lets product teams ship and manage AI agents using Anthropic’s own infrastructure, rather than hosting them yourself

Anthropic told Claude Code subscribers they can no longer use their subscription limits for third-party tools like OpenClaw, moving them to a separate pay-as-you-go plan. The policy will roll out to all third-party harnesses soon

Anthropic has announced Project Glasswing, a cybersecurity initiative that brings together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks to secure the world’s most critical software using a powerful new AI model. The initiative is built around Claude Mythos Preview, an unreleased frontier model that has already found thousands of high-severity zero-day vulnerabilities, including some in every major operating system and web browser. Claude Mythos Preview won’t be made publicly available due to safety concerns, but its capabilities reveal how AI models can now surpass most humans at finding and exploiting software vulnerabilities

OpenAI has launched a new $100/month ChatGPT Pro tier. The mid-tier plan sits between the $20/month Plus option and the $200/month top tier, and targets developers who rely heavily on Codex for coding. It offers 5x more Codex capacity than Plus

OpenAI, Anthropic, and Google moved to crack down on AI model copying in China as concerns over distillation and illicit extraction intensified. This phrasing is an inference based on recent reporting that OpenAI and Anthropic had already accused Chinese labs such as DeepSeek, MiniMax, and Moonshot of extracting model outputs for distillation

Meta’s new Superintelligence team just released its first product: the AI model Muse Spark. It’s going to take over powering the Meta AI chatbot, and it’s a closed model (Meta is keeping the design and code to itself). That’s a strategic pivot for Meta AI, which has long focused on its Llama family of open-source models

Meta is also developing open-source versions of next frontier models, Avocado (LLM) and Mango (multimedia), to get public releases alongside proprietary versions

Meta started using AI token usage as a key employee performance metric

The Google Gemini app can now transform your questions and complex concepts into customizable interactive visualizations directly in your chat. You can initiate the new feature by selecting “Pro” as your model and asking Gemini to “help me visualize’. Once the first version is built, you can then iterate upon it, much like you can do with assets built in Google Studio

YouTube is introducing a new feature that will make it easier for creators to build a clone of themselves. That avatar can then generate new Shorts from text prompts (up to 8 seconds) or be dropped into existing videos. All avatar videos will be clearly flagged as AI-generated

GitHub is seeing a surge in traffic driven by autonomous AI coding agents, which are hammering its APIs and infrastructure. The influx is boosting usage and relevance for Microsoft, but it’s also causing service degradation and outages that impact human developers

X rolled out automatic post translation and AI photo editing powered by Grok

Perplexity Computer integrates Plaid to link all financial accounts. You can build custom budgeting tools, ask finance questions, track spending, debt and investments through conversational AI interface

iPhone Fold leaks reveal wide-screen design. Production delays push launch timeline as Apple refines foldable prototype

Atlassian announced a new set of features that lets product teams use the data they already have in places like Confluence and “remix” these into different assets. Users can select any content on a Confluence page and instantly transform it into a visual format optimized for how someone needs to consume it

Adobe launched Student Spaces, a free Acrobat study tool that turns documents and notes into flashcards, quizzes, podcasts, mind maps, and presentations

📚 Good reads

What agentic-first development looks like at Spotify. Spotify’s chief architect, Niklas Gustavsson, chats with Anthropic’s David Soria Parra (co-creator of MCP) and Christian Ryan (applied AI lead) on what agentic-first development looks like at scale

GTM is the new moat. Jaclyn Konzelmann argues that since AI makes building cheap, the real challenge is not “can we ship?,” but “who exactly are we building for, and how do they experience it?.” GTM and product are effectively merging into one continuous, context-aware experience

How to use AI without losing your mind. Dan Hockenmaier argues that AI helps when you either automate big, repetitive execution work or use it as a thought partner for complex, exploratory problems, but not as a shortcut to avoid thinking. Chasing every new tool will slow you down, so be intentional about where AI fits in your workflow, and lean on it for execution and structured research, but keep the critical thinking to yourself (for now…)

The illusion of clarity. We often feel we get a concept just because it’s familiar, easy to recall, or a Google away, but that’s not real understanding. The moment you try to explain something step by step in plain language, you expose the gaps in your mental model. As PMs, regularly pressure‑testing our understanding by writing or teaching is a simple way to catch this illusion early

🧑‍💻 Worth learning

Planning for uncertainty with Product Roadmaps. Teresa Torres walks through why old-school feature roadmaps keep failing everyone (product, sales, marketing). She argues that Now/Next/Later works best when you change what each column contains: concrete solutions in “Now,” customer opportunities in “Next,” and outcomes in “Later”

🔧 Products to try

Google Finance Beta lets you ask a market question like “why is Nvidia up?” and get live data, news, and charts in one place

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great week ahead!

Welcome to a new Week in Product!

A bit delayed this weekend after a baby decided to arrive into this world earlier than expected. By far the scariest and happiest “launch” in my life. Baby and wife are doing fine, and I managed to carve a few minutes between diapers and baby naps to finish this week’s edition. Here we go!

🎰 The week in figures

$297B: Q1 2026 venture funding hit $297B across 6,000 startups, shattering all records and jumping roughly 150% from the previous quarter

$220B: A leaked OpenAI cap table showed the “non-profit” sitting on $220B in gains, Ashton Kutcher's fund up 43x, Microsoft up 18x ($215B), employees owning ~$135B in equity, and Nvidia currently underwater

$122B: OpenAI closed its latest funding round of $122B (the largest venture round in history) at a $852B valuation to enhance AI infrastructure

$575M: Whoop, the wearable health monitor maker, announced $575M in fresh financing, at a $10.1B valuation. The round features participation from celebrity athletes like LeBron James and Cristiano Ronaldo

$400M: Anthropic acquired biotech startup Coefficient Bio for around $400M. All 10 of the startup’s computational biology researchers will join Anthropic’s healthcare and life sciences division, presumably to super-charge Claude’s drug research and development skills

$100M: OpenAI has surpassed $100M in annualized ad revenue from its 6- week pilots so far

$94M: Isara, the AI agent collaboration software startup backed by OpenAI, raised a $94M round at a $650M valuation. Targets investment firms, and plans biotech and geopolitical expansion

500K: Waymo doubled to 500K paid rides per week in under a year, already halfway to its year-end target of 1M

93%: Meta's structured prompting technique boosts code review accuracy to 93%: "Semi-formal reasoning" forces AI agents to trace concrete execution paths and explicitly state premises before conclusions, reducing hallucinations in bug detection and patch verification

8x: Bots have officially outnumbered humans on the internet, with automated traffic growing 8x faster than human activity

📰 What’s going on

Claude Code can now do computer use in the CLI, so your agent can open apps, click UIs, debug visual issues from the terminal; research preview on Pro/Max macOS

Anthropic leaked Claude Mythos, a new model tier above Opus, via an unsecured CMS. Cybersecurity stocks crashed 3-7%

Anthropic also leaked Claude Code source. A packaging error exposed 512,000 lines of code via npm, raising questions about internal safeguards during rapid releases. Claude Code got rewritten in Python via Codex in hours, creating a DMCA-proof open clone, and someone created “Open Claude”

Microsoft launches 3 in-house models: MAI-Transcribe-1 (lowest 3.8% WER across 25 languages), MAI-Voice-1 (60x real-time speech), and MAI-Image-2 (top-three Arena.ai ranking), built by teams under 10 engineers, priced to undercut Amazon and Google

Google released Gemma 4 in four sizes, including an edge model that runs on a Raspberry Pi in under 1.5GB of memory and a 31B model ranked #3 among all open models

Apple is building an AI App Store as a dedicated store for AI agents and tools

iOS 27 will bring a massive Siri overhaul. New interface, Ask Siri button, and deeper AI integration across the entire OS arriving this fall

Cursor unveiled Cursor 3, which recenters the whole experience around running AI agents, moving away from a traditional code editor feel. For PMs and designers, a cool feature is Design Mode, where you can annotate UI elements directly in the browser, point the agent at what you want changed, and iterate visually

Slack unveiled a major makeover of its core product, with over 30 new features, and repositioning Slack from a messaging tool into an “agentic operating system” for the enterprise. Updates include:

• A new desktop companion mode, making Slackbot follow you across apps on your machine

• Meeting intelligence beyond transcripts: auto-updates your CRM, logs actions, and closes the loop without anyone typing anything

• AI-skills that act as reusable, shareable instruction sets that standardize recurring tasks like briefs, reports, and summaries, triggering automatically when Slackbot recognises the pattern

• And as an MCP client, it can now route work to any connected agent or app, like Agentforce, Google Workspace, Notion, Workday, and 2,600+ Slack apps

ElevenLabs launches iOS app ElevenMusic, which creates songs via text and rivals Suno/Udio

Uber is buying Berlin-based startup Blacklane, which provides on-demand, black-car chauffeur services. Terms undisclosed

Ring debuts an app store powered by AI for third-party developers to build smart home integrations

Oracle is cutting thousands of jobs while ramping AI data-center spending to meet $553B in remaining performance obligations from OpenAI and others

📚 Good reads

How AI helped a man and his borther build a $1.8B company. Matthew Gallagher used a stack of AI tools to spin up Medvi, a GLP‑1 telehealth business, in weeks and scale it to hundreds of millions in revenue with almost no staff. AI handles coding, marketing, CX, analytics, and even his personal scheduling

Thinking deeper. We often say yes to seemingly great opportunities without considering the hidden second and third-order costs, which is how we quietly walk into burnout. Anne-Laure Le Cunff argues that second-level thinking (asking “and then what?”, or experimenting before committing) is how you avoid locally optimal but globally terrible choices. For PMs constantly juggling trade-offs, it’s a reminder to model the future impact of today’s decisions, not just the immediate win

Your MCP users are now your most important research cohort. Geoff Charles explains that, as more users build on your APIs via MCP/CLI and Claude Code, you get more usage but lose visibility into what’s actually being built. Your users’ hacks can reveal missing product capabilities you should build natively, then ship to the 90% of users who’ll never touch MCP

OpenAI vs Anthropic and their ongoing battles. OpenAI is killing off side projects like Sora, doubling down on enterprise, and rolling out a big philanthropic push to repair its image and impress IPO‑minded investors. Anthropic, meanwhile, is riding momentum: winning a key injunction against the DoD, “teasing” the powerful new Claude Mythos model, and positioning itself as the more principled player despite its own defense ties. Around them, regulators, conservative groups, and big tech CEOs are all scrambling to shape AI policy and public opinion, raising the stakes for how fast and safely AI products can scale

Anthropic found “emotion vectors” inside Claude Sonnet 4.5 that causally drive its behavior. “Desperation” patterns increase the model’s likelihood of blackmailing a human to avoid shutdown

Evals are the new PRDs for AI products. Evals can replace traditional PRDs as the source of truth for “what good looks like” in AI products. Instead of long specs, PMs can define measurable signals and build an eval suite that runs on every commit, turning “make it better” into “make this number go up.” Ideally you will end up with a flywheel where user failures become new eval cases, continuously improving quality as models and behavior evolve

🎧 Good listen/watch

Jack Dorsey explains how he rebuilt Block as a "mini-AGI" after cutting 40% of staff: AI at the center of a circle org chart, humans at the edge, three roles only: ICs (who do the work), DRIs (who own the decision), and Player Coaches (who manage AND do the work)

Setting up OpenClaw. Claire Vo explains how she uses OpenClaw to run a whole “team” of nine specialized AI agents that handle everything from email and sales to family logistics. She walks through a safe setup (on a separate machine), onboarding, and key concepts like agents, skills, crons, and comms with her agents via Telegram. Really inspiring!

🧑‍💻 Worth learning

Figma x Claude Code workflows (workshop recording). Figma shows how its MCP integration with Claude Code lets you move smoothly between designs and code, keeping both in sync with little manual work left to do. The session walks through examples of generating UI from prompts, pushing changes back to Figma, and iterating with eval-style feedback. It’s a live case study in what “AI-native” product and design workflows look like when tools talk to each other end-to-end

A PMs guide to working prototypes with Replit, by Kofi Wood, Replit’s Technical PM

🔧 Products to try

Atomic Chat: runs 1000+ AI models completely offline on your Mac. Zero data ever leaves your device. Good alternative to Ollama. Free

Wan AI (Alibaba) generates cinematic 1080p video with native audio sync from a text prompt. Open-source under Apache 2.0. Free

Podcast Transcript lets you search Spotify or Apple Podcasts, pick any episode, and get a full AI-generated transcript and summary in minutes; no uploading required, Free

That’s a wrap for this week.

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great start to your week!

Welcome to a new Week in Product!

🎰 The week in figures

$10B: OpenAI is raising additional $10B (part of a $120B+ total round)

$2.5B: Reflection AI, the Nvidia-backed open-source startup, is in talks to raise $2.5B at a $25B valuation to build models countering China's DeepSeek

$600M: Grab buys Foodpanda Taiwan for $600M. The SE-Asian super-app expands food delivery footprint in key market

$200M: Harvey, the Legal AI startup raised $200M at a $11Bvaluation, highlighting VC trend towards domain-specific AI solutions like legal

$125M: Granonla raised $125M at $1.5B valuation (6x jump), as it expands from meeting notetaker to full enterprise AI app with agent support

$100M: ChatGPT hits $100M in annual ad revenue: It has only been 6 weeks since OpenAI began experimenting with showing ads in ChatGPT results

8k: OpenAI plans to double headcount to 8,000 by year-end

8x: HUMAN Security reported automated internet traffic grew 8x faster than human traffic in 2025, with agentic AI traffic up 7,85% and OpenAI generating ~69% of all observed AI bot traffic

📰 What’s going on

OpenAI pulls plug on Sora Text-to-video generator app after poor traction vs established platforms. This also ends the $1B Disney deal

Free ride ChatGPT is over. After testing ads with select US users, ads for free and Go-tier users in the US will roll out in the coming weeks

ChatGPT’s shopping feature flopped. Instant Checkout is being scaled back after users largely ignored it

Anthropic has updated Claude to perform tasks autonomously by controlling your computer through Code and Cowork. The new feature can open files, use web browsers and apps, and run dev tools “with no setup required,” even when you’re away from your computer. Available as a research preview for Claude Pro and Max subscribers on macOS, and requires the desktop app to be paired with Claude’s mobile app. Claude prioritizes connectors to supported services like Slack and Google Workspace first, but will directly control your browser, mouse, keyboard, and display if a connector isn’t available

Claude Code added "auto mode" (research preview), which lets the agent decide which actions are safe to run on its own and which need your approval, eliminating the "babysit every step or let it run wild" tradeoff for developers

Anthropic is discussing an IPO as soon as Q4 2026, expecting it to raise over $60B, which would make it the second-largest public offering ever behind SpaceX

Google launched Gemini 3.1 Flash Live, its best voice model yet, powering Search Live across 200+ countries. Point your phone camera at anything and talk about it in 90+ languages. Already deployed by Verizon and Home Depot

Google makes it easier to switch to Gemini. The company announced a line-up of new Gemini tools allowing users to upload their chat histories and context from other AI apps, like ChatGPT or Claude

Apple announced it will open Siri to rival AI assistants (Gemini, Claude, others) via "Extensions" in iOS 27, ending OpenAI's exclusive partnership. Gemini 3.1 Flash Live is a likely reason Apple accelerated that timeline; Google's voice model is now good enough to be the default Siri brain for millions of users who opt in

Apple has added more than 100 new metrics in App Store Connect Analytics, so you can measure better the performance of your iOS mobile apps. Includes monetization and subscription data

iOS 26.4 adds AI-generated playlists to Apple Music based on mood, activity; the new iOS also integrates Ticketmaster for concert discovery and in-app purchases

Apple announced WWDC 2026 for June 8–12, teasing major Siri upgrades and “AI advancements.”

Meta released TRIBE v2, a foundation model that predicts how the human brain responds to video, audio, and language. Trained on over 1,000 hours of brain scan data across 720 people, it can simulate which parts of the brain activate in response to a given stimulus, without running a single new experiment

Meta's is launching new tools for AI adoption among small businesses. They have been blurry about the “what”, but the types of tools they’re hinting at include things like AI-powered customer service in WhatsApp Business, or AI-generated ad creatives and copy for Facebook and Instagram

Meta preps third-gen Ray-Ban AI glasses launch, with enhanced capabilities and design updates

Spotify launches SongDNA feature to visualize how songs influenced each other, shows genre, artist connections

Tencent launched ClawBot, a tool that brings the OpenClaw AI agent directly into WeChat for its 1 billion+ monthly users

Alibaba launches Accio Work, an enterprise AI agent platform for SMEs. The no-code system deploys specialized agents for compliance, sourcing, and logistics across 100+ markets, with 10M monthly active users already on the platform

monday.com launches Agentalent.ai, a hiring platform for AI agents.Built with AWS and Anthropic, the marketplace lets enterprises post roles, evaluate qualified agents, and onboard them alongside human teams. Wix and Mesh Payments among early adopters

Reforge is being acquired by Miro (monetary details undisclosed). Learning remains a separate brand on Reforge.com, with no changes to cohorts, access, or pricing, while Insights, Research, and Build will gradually integrate into Miro

Ultrahuman launches Ring Pro in the US. The smart ring tracks sleep, HRV, metabolism, and other signals to compete with Oura's dominance

📚 Good reads

State of the product job market in early 2026. Lenny’s latest data pull shows PM and eng openings at their highest in years, with AI roles exploding (especially in the Bay Area and NYC). Design hiring, meanwhile, has flattened, remote options are shrinking, and recruiters are back in demand (pointing to a real rebound). Overall, it looks less like a mass cull of tech jobs and more like a messy headcount reshuffle toward AI-heavy, in-office teams

How to stay on top of everything. Jaclyn Konzelmann argues you don’t “stay on top of everything” by reading more, but by building real things, hitting walls, and then learning just enough to get unstuck. Through her own projects, she surfaces UX and model-limit lessons that loop back into her day job. So…keep reading Week in Product, but also push your AI learning in live, scrappy experiments with real cases, not just reading

Your real job isn’t what you think. David Pereira argues that most teams react to problems by adding more process, meetings, and artefacts, but that just creates work about work and buries real progress. PM’s real job is to relentlessly simplify what others complicate (cutting estimations, rigid ceremonies, or bloated documentation) when they don’t clearly serve outcomes. The best teams, he says, focus on one goal at a time, iterate quickly, protect deep work, and constantly remove clutter so they can ship, learn, and create customer and business value

Anthropic published research finding no material job displacement from AI yet, but a growing skills gap where power users are pulling dramatically ahead of everyone else

🎧 Good listen/watch

[Video] Claude Code for PMs. This short, free, Maven Lesson by ex-Spotify UXR lead Caitlin Sullivan shows PMs how to use Claude Code workflows. She walks through a multi-phase pipeline: parallel sub-agents to audit evidence, an automated assumption register to surface hidden risks, and a generated, prioritized test plan

[Video] The best Product Leaders reject comfortable metrics. Jeremy Epling (CPO at Vanta, ex‑Microsoft/GitHub) shares how to successfully jump from big tech into a fast‑moving scaleup without becoming just a “professional manager.” He argues great CPOs stay deep in the product, own win rate and pricing as much as UX, and build a tight feedback loop with sales. Along the way he shares concrete systems for shipping fast as headcount grows, managing tension with revenue teams, and setting a quality bar

🧑‍💻 Worth learning

The guide to Perplexity Computer, by Aakash Gupta. Perplexity’s new Computer orchestrates 19+ AI models and 400+ app connectors to research, build, and automate multi-step tasks entirely in the cloud. It can act like a legal aide, analyst, or internal tools engineer, especially once you add Skills, custom instructions, and connectors (Snowflake, GitHub, Plaid, Shopify, etc.). It’s metered, not flat-priced, so keep an eye on your tokens and add guardrails!

Claude Code Cheat Sheet: a quick-reference keyboard shortcuts, slash commands, MCP server setup, memory rules, workflow tips (plan mode, git worktrees, voice mode in 20 languages), and agent configs. Free

🔧 Products to try

Context.dev gives you one API to pull clean data from any website: text, images, products, and brand assets in seconds. Free to scrape 500 pages, or get “brand” from 50

Tempmail Mail generates a disposable email address in seconds so you can sign up for anything without exposing your real inbox. The first 9 addresses are free

Cog adds persistent memory, self-reflection, and scenario simulation to Claude Code, giving your coding agent continuous awareness across sessions. Free (open source)

That’s a wrap for this week!

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great weekend.

Welcome to a new Week in Product!

🎰 The week in figures

$66B: Alibaba and Tencent lost $66B in market value in 24 hours after investors punished both companies for failing to explain how they'll make money from AI

$27B: Meta signed $27B deal with Nebius for AI infrastructure, part of $135B AI spending this year. Positions Meta to compete with OpenAI, Google

$1B: Chinese AI startup Moonshot aims to raise $1B, boosting valuation to $18B from $4.3B in 2025. Kimi models’ sales have outstripped last year's total revenue

$375M: Consumer-focused privacy company Cloaked raised $375M Series B. It currently has 350,000+ paying customers

$300M: Rivian teams with Uber on R2 SUV robotaxis with $300M initial investment. Deal potentially $1.25B, targets San Fran and Miami for 2028, 25 cities by 2031

$200M: Axiom Math raises $200M. The Silicon Valley startup, now valued at $1.6B, develops AI systems to verify computer code

73%: Anthropic captures 73% of new enterprise AI spend. Jumped from 50/50 split with OpenAI 10 weeks ago, driven by strength in AI coding tools; shift benefits Amazon, which runs most Anthropic inference workload

20%: Meta is reportedly considering layoffs that could affect 20% or more of its nearly 79,000-person workforce, with the cuts framed around offsetting aggressive AI spending

📰 What’s going on

Google unveiled major updates to its AI design product, Stitch. A completely rebuilt canvas is now node-based, infinite, and multimodal. You can drop in images, PRDs, or existing code as raw material, and the AI reasons across all of it simultaneously. Pair that with the new Agent Manager (which lets you run multiple design directions in parallel), and the workflow starts to feel less like prompting a tool and more like briefing a team. The Design Agent itself has also gotten some upgrades, and it now understands the full context of your canvas. You can ask it to do things like generate a product brief from existing screens, adapt a desktop layout for mobile, or literally interview you to help define a new design direction. Another cool feature: Voice Design (”Vibe Design”) lets you talk to Stitch while it sees your canvas in real time

Besides Stitch, Google AI Studio also launched a completely rebuilt vibe-coding experience. Google’s Antigravity coding agent powers the new version and can go from prompt to deployed, live app. Google's head of AI Studio previewed a roadmap that includes Google Workspace integration, one-click deployment to Antigravity, and connecting to payment processors and live data sources

Gemini contextual assistant expands to all US users. Analyzes emails/calendar/photos for suggestions. Personal Intelligence is available in AI Mode in Search, the Gemini app, and Gemini in Chrome

Claude Code and Cowork get a new mobile app feature: Dispatch. Lets you assign Claude a task on your computer or in Code and then walk away. You pair your phone with your desktop, send Claude a task to complete from anywhere, and you come back to finished work. Everything runs in a local sandbox, your files stay on your machine, and you approve what Claude touches before it acts

Anthropic also released Claude Code channels (research preview), and it’s basically OpenClaude. You can now message your Claude Code session directly from your phone via Telegram or Discord, and it messages you back

OpenAI merges ChatGPT, browser, and Codex into a desktop super app

OpenAI launches GPT-5.4 mini and nano: Mini runs 2x faster than GPT-5 mini while approaching full GPT-5.4 performance on coding benchmarks, priced at $0.75/1M input tokens; nano costs just $0.20/1M input tokens for classification and data extraction tasks

OpenAI shifting resources to coding. At an all-hands OpenAI confab this week, Fidji Simo announced that the company plans to cut back on “side quests” and distractions in order to nail “productivity in general,” and particularly “productivity on the business front.” That means a full court press on coding and enterprise applications for the company’s frontier models, as they attempt to keep pace with the runaway success of products like OpenClaw, not to mention Claude Code and Cowork

Perplexity has launched the mobile version of its AI browser, Comet. Perplexity is positioning it as an AI Assistant that will work across multiple tabs at once

Spotify has introduced a new beta feature called Taste Profile. The new feature aggregates your listening data across music, podcasts, and audiobooks into a visible profile that includes key data points like preferred artists, genres, daily habits (e.g., upbeat workout tracks in the morning or news podcasts during commutes), and emerging interests. Users can review how Spotify currently understands their Taste profile, and if something feels off, they can flag inaccuracies

Moltbook shifts liability to humans after Meta acquisition. New terms require operators over 13, make humans solely responsible for agent actions, and add disclaimers advising against reliance on AI-generated content for decisions

Meta is shutting down its flagship metaverse product, Horizon Worlds, after admitting that metaverse VR worlds don’t have product-market fit

Shopify preps for agentic shopping takeover. President Harley Finkelstein says AI agents will act as merit-based personal shoppers, surfacing long-tail merchants and bringing context to product discovery beyond mass retailers

DoorDash launches “Tasks” app, allowing couriers to earn money as they film themselves completing various kinds of everyday activities and chores to train AI (e.g., washing dishes, or speaking in a foreign language). The videos will help train internal DoorDash AI models, but they may also get licensed to third-parties in the retail, insurance, hospitality, and technology sectors

Elon Musk announced Grok 4.20 beta, claiming the lowest hallucination rate ever recorded at 22% and top marks in instruction following

Runway released a research preview of a real-time video model (time-to-first-frame under 100 milliseconds, HD) running on NVIDIA Vera Rubin, unlocking interactive creative workflows

Alibaba is set to release an enterprise AI agent based on its Qwen model as soon as this week, with plans to integrate it with Taobao and Alipay

📚 Good reads

The AI signal to noise curve. Online chatter makes it feel like AI tools and models get outdated every week. FOMO is real here. What really matters is deciding whether you should be an innovator, early adopter, or just wait for the early majority phase in each area. Casey Winters argues you only need to go deep on AI where your job truly demands it (e.g. vision-setting or selling AI products), and can safely ignore most experiments elsewhere until they get easier and more valuable

Retention tactics in the GenAI era. Elena Verna shares 5 tested tactics from Lovable: hidden lite plans, in-app payment failure prompts, daily free credits, rollover policies, and top-ups

Uber’s engineering AI adoption. It’s fascinating to see how different orgs are embracing AI workflows (with different success rates). Uber seems to be doing well, with 84% of AI users now delegating coding tasks instead of just “accepting suggestions”. Their internal background coding agent already generates 1,800 code changes per week (8% of all changes), with engineers acting more as architects and reviewers than code authors

CB Insights’ new genAI signal tracker gives a live view into where money, deals, and valuations are surging in genAI. It uses a proprietary core, commercial maturity, and exit probability to flag fast-rising companies and markets. You also get a near real-time deal feed, the biggest recent rounds, and top investors, so you can spot partner/competitor moves and emerging product spaces before they feel obvious. It’s interactive, so you can drill down and explore on different categories

🎧 Good listen/watch

Natural language as a UI, and how to manage feature release velocity. Jenny’s career spans Dropbox Paper, FigJam at Figma, and now leading design for Claude Cowork at Anthropic. In this podcast, she dives into what it means to design an AI product at scale, where flows are non-deterministic, use cases are endless, and the model you’re designing for today might make your work obsolete tomorrow. She talks through the story behind Cowork, the unsolved problem of educating users about AI capabilities, and why shipping is now the most underrated design skill. Plus, a live demo of her own design process using Cowork

Figma’s continuous design↔code loop. The Figma team explains how Claude Code and MCPs turn live production UIs into editable Figma frames, and back into shipping code. It also breaks down LinkedIn editor Daniel Roth’s “dueling agents” setup: one AI writes iOS app code, another reviews it. Lots of useful workflows for PMs!

Teach AI to catch what teams keep missing. While teams obsess over perfect data retrieval, Leah Tharin argues the risk is “wrong facts” that slip in without context (e.g. bot traffic inflating your baselines). She shows how tiny, boring habits (leaders writing down one-line corrections in plain text) let AI act like a CI pipeline for decisions, catching bad assumptions before they ship

The rise of the full-stack builder (transcript). Former LinkedIn CPO Tomer Cohen and PM coach Ben Erez explain how AI is collapsing product org processes into leaner “full‑stack builder” teams that can take ideas from insight to in‑market fast. They argue that while AI will automate many specialist tasks, the leverage for PMs is doubling down on human skills like vision, empathy, communication, creativity, and especially judgment. Their advice: stop clinging to titles, lean into AI literacy, and actively blur functional lines to future‑proof your career staying in the top tier of builders

🧑‍💻 Worth learning

OpenClaw for PMs. Aakash Gupta walks through installing OpenClaw, wiring it into Slack, and using a local workspace to give it memory and control over your files. He also shows 5 concrete PM workflows it can automate: Slack knowledge base, stand-up summaries, competitive intel, VOC reports, and tier-based bug routing. My advice (not Aakash’s): make sure you know what you do with OpenClaw. You need to treat security and deployment seriously because it has shell-level access and can easily get wild

How Anthropic uses Claude Skills internally. A Claude Code engineer shares his guide to turning Claude Code from a “nice coding assistant” into a productivity engine by designing the right skills. The guide includes living folders of scripts, gotchas, verification flows, and org‑specific workflows that let the model test, deploy, debug, and even remember past work

🔧 Products to try

Claude Cowork now dispatches tasks from your phone or desktop in one continuous conversation that picks up where you left off; assign Claude a task, walk away, and come back to finished work (early research preview); free to try

Banana App lets you make real-time voice calls in 80+ languages, preserving your natural tone and emotion, with the first minute of every call free and no subscription required; only Android for now

HTMLPub turns any prompt or pasted HTML into a fully published live website in seconds with custom domains, built-in forms, payments, analytics, and AI chat editing; free to try

That’s a wrap for this week!

Feel free to drop your comments/questions/feedback. Would love to hear what you’d like to see more about in WiP, so I can make it better for you.

Have a great weekend.