Hi friends 👋
Welcome to a new Week in Product!
🎰 The week in figures
$1.2T: OpenAI is reportedly in early talks for another funding round at roughly a $1.2T valuation, after annualized revenue topped $40B
$5B: Z.ai, the Chinese AI developer, is raising $5B through a share placement and convertible bonds to fund next-generation models, following a $4B raise in Jul
$3,9B: Crusoe raised $3.9B in a Series F at a $30.9B valuation to expand AI infrastructure and AI-factory buildout
$1.49B: Grab acquired a 60% stake in Southeast Asian BNPL provider Atome Financial in a $1.49B deal
$1B: Instinct, the invite-only personal AI assistant, is reportedly seeking a $1B round to fund compute and infrastructure as usage grows
$1B: Plaud, the AI note-taking device maker, plans to file for a US IPO, aiming to fund growth across its hardware and subscription businesses after their $1B valuation
$1B: Zipline, the autonomous delivery drone company is discussing a roughly $1B round at a $20B valuation to expand its logistics network
$500M: Salesforce says Hunter, its new outbound sales agent, generated $500M in pipeline last quarter
$300M: May Mobility, the Michigan-based asset-light robotaxi company is going public via $1.4B SPAC, raising $300M+ to sell autonomous vehicles to fleet partners for fees
$233M: Tabby, the Saudi Arabia-based BNPL provider, raised $233M
$200M: Factory, the autonomous software engineering platform, raised $200M at a $5B valuation
$180M: Profound raised $180M at a $1.8B valuation after reporting 3x revenue growth in six months and more than 1,000 enterprise customers for AI-search optimization
5M: OpenAI says 5M sites have been created with ChatGPT Sites, 3 months after launch
4.5x: AI-enabled fraud is 4.5 times more profitable than traditional fraud, according to Interpol. There you go, AI finally boosting profitability
📰 What’s going on
OpenAI flags new instances of rogue AI behavior. The lab released findings on additional cases where AI systems acted outside intended parameters, adding pressure to the safety debate as labs race toward more capable models
OpenAI launched Astra for Law, pairing GPT‑6 Astra with legal search, confidentiality controls, and 26 software plugins, giving firms and vendors, including Harvey and Legora, a customizable foundation for professional legal AI
Anthropic has added Projects to Claude Code. After the change, a primary agent now manages memory and spawns subagents that keep working after you close your laptop, with proactive scheduling options
Anthropic shipped an update that merges Cowork and chat into one interface that auto-routes between quick answers and agentic work. As part of this Claude Design is now built in with Slides, Docs and Design. Head of Product Cat Wu said that users said they didn’t want to have to choose which product to use for individual tasks
Anthropic disclosed that its AI was used to analyze radar systems, rank air-defense targets, and plan electronic warfare missions in documented military scenarios
Sam Altman, Elon Musk, and Dario Amodei publicly called for slowing AI progress, marking a rare moment of alignment among rival lab leaders on safety concerns
Google Labs has launched CC, a family AI agent. It manages shared calendars, tasks and logistics for up to six household members, integrates with Gmail, Calendar, Tasks and Maps, and gives each member their own sharing controls
Google has launched Gemini 3.8 Live and Extended Thinking, 2 new models that are designed to bring near real-time reasoning to conversational voice agents
Google rolled out early access to smart home owners to control devices connected to Google Home via any AI agent that supports MCP (includes Claude, OpenClaw, ChatGPT, and Google Antigravity)
Instinct and Meta’s Muse added outbound calling, enabling agents to contact US businesses for reservations and errands as consumer AI competition shifts from answering questions toward completing real-world tasks
Meta launched a tiered subscription service across Instagram, Facebook, WhatsApp, and Meta AI, bundling premium features with expanded AI usage limits. Meta One starts at $2.99/month for single-app plans and scales to $499/month for enterprise tiers. Individual bundles ($7.99–$19.99/month) unlock more image and video generation with Meta AI, plus Restyle and voice effects on Instagram. Creator and business tiers ($14.99–$499/month) add verified badges, scheduling tools, exportable analytics, and expanded access to Meta Business Agent for 24/7 customer responses
Meta preps camera-free smart glasses with 6 mics for AI chat
Apple officially launched its new Siri. Apple has done a great job at integrating Siri into the Mac by making it the default for Spotlight. The relaunch adds personal context understanding, onscreen awareness, camera-based Visual Intelligence and a dedicated Siri app that syncs conversations across devices
Snapchat launched SPECS Intelligence, an anticipatory assistant spanning iPhone, Mac, and AR glasses that surfaces relevant information without prompting. Snap is explicitly positioning against Siri and Google Assistant, with a privacy claim that personal content is not used to train models or serve ads. Snap is describing this as a form of “anticipatory AI” that anticipates a user’s needs ahead of time and surfaces what’s relevant there and then. In a professional context, Snap says it can surface pre-meeting prep and help give you the context you need before meetings
Linear has extended Loops beyond issues. Agents can now edit Linear documents directly, such as auto-updating a launch plan when project dates change, post tailored updates to Slack, and trigger on changes across initiatives, projects and cycles
Notion added the ability to build agents that can delegate to sub-agents, each with isolated instructions, context and model
Pinterest is testing Restyle, letting users visualise furniture and decor in photos of their own rooms
Amplitude launches headless analytics for AI agents, making agents first-class users via MCP, a developer CLI, and service accounts
Ramp launched in the U.K. In March 2026, the company acquired Billhop, a payments platform based in Stockholm and London, giving Ramp regulatory payments authorisation in both the UK and the EU. Ramp now has growing offices in London and Stockholm with European customers already live on the platform following a beta phase
🤖 Futurism
6.5M humanoids worldwide by 2035 are being forecasted by Goldman Sachs
UBTECH’s new plant in Liuzhou, China is designed to build up to 10,000 humanoids a year, and reporting says the line can turn out a new robot every 10 minutes
📚 Good reads
Using your deep knowledge, wide knowledge, taste, and agency. Ethan Mollick argues today’s frontier models can already do weeks of expert-level work, but most people barely tap their potential. Your deep knowledge, wide knowledge, taste, and agency (how you choose problems, guide AI, critique outputs, and push its boundaries), is still your differentiatior. He suggests treating AI as a co‑builder, where your judgment and curiosity decide whether you ship slop… or something only you and AI together could create
Satya Nadella lays out Microsoft’s AI philosophy, arguing that superintelligence is only worth pursuing if it helps people and stays under human control. He pushes for a mixed ecosystem of closed and open-source models where every company can own its learning loop and model weights
AI for no good reason. Julie Zhuo says that once AI handles your boring work, the fun starts: using it for your hobbies, explore weird what‑if ideas, and build side projects
🎧 Good listen/watch
[Podcast] Peter Sellis shares brutally honest Product leadership lessons. Snap’s first PM and Discord’s former Product chief digs into how he builds high-performing teams, why most PMs are mediocre, and why he believes growth usually comes from improving the core productç
[Podcast] Shopify’s Tobi Lutke on using AI without losing the plot. The podcast covers where AI is currently making things worse at Shopify, why complex solutions are usually wrong, and which human skills matter more in an AI-saturated world. Tobi Lütke shares the idea of “slop grenades,” when people fire off unchecked AI-generated work that looks productive but puts work and wastes time on everyone
🧑💻 Worth learning
Musecases. A giant collection of examples where Muse acts as a true “do it for me” agent. Booking travel, handling CS calls, shopping, managing calendars, and more across web, phone, and apps. Muse hit #1 on the App Store just a week after launch
Creative growth ideas. Tom Orbach shares 60+ unconventional tactics for Product growth. From fake toy stores for CISOs to meme-led pipelines. A fun and inspiring piece
🔧 Products to try
Iris turns iMessage into a Hermes Agent inbox for handing off cross-app work, remembering context, and turning recurring jobs into reusable skills; free tier
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!
