Hi friends đ
Welcome to a new Week in Product!
đ° The week in figures
$30B: OpenAI is reportedly working on a $30B raise at a $1.25T valuation
$15B: DeepSeek reportedly raised about $15B
$1B: Flapping Airplanes, a 1-year-old Silicon Valley AI lab building models that train on far less data, raised a $1B Series A at a reported $6B valuation, 4x its seed
$550M: Legora, the Swedish AI platform for law firms and corporate legal departments, closed a $550M Series E at an $8.8B valuation, one of the largest legal tech raises
$320M: Otter.ai secured $320M for enterprise growth, go-to-market strategy, international expansion, and product innovation
$250M: Pigment, an AI business planning platform, raised $250M Series E at a $2.3B valuation
$200M: Whop, a platform for launching businesses and selling digital products, raised $200M Series D at a $5.1B valuation
50M: Spotify purged 50M AI slop tracks from recommendations. The streaming platform removed AI-generated spam and updated its algorithm to prioritize human artists, while launching an "AI Persona" badge for synthetic creators
10M: Metaâs Muse has surpassed 10M downloads in the US and Canada since its launch in September. +1/4 of users make purchases through it. The app also hit the milestone in just 40 days, something only 11 other apps have done. Itâs nearly three times ChatGPTâs 3.7M US downloads over its first 40 days
83%: Microsoft AI says its new system, MAI-DxO, solved 83% of a set of hard cases from the New England Journal of Medicine. The doctors in the same test got 36%. The doctors were primary care physicians and general internists working alone, not the specialists and teams who usually handle cases this hard
5%: Anthropic pledged 5% of its shares to a new Public Benefit Fund to help with job displacement and other societal issues caused by AI
đ° Q3 earnings
Meta: reported negative free cash flow of $2.12B in Q3 and increased its full-year capex range to $135â145B, raising the floor from $130B. Fell 7%+ post-earnings, erasing $125B, despite 26% sales growth. Creating non-voting shares, set to trade in 2027, to fund AI infrastructure
đ° Whatâs going on
Apple will present its first touch-screen MacBook Pro at the Oct. 28 LA event; plus an M6 MacBook Pro, new iMac, and iPad mini
You can now take an app built in Gemini and import it into Google Sites. For example, you can ask Gemini to build an interactive calculator that shows how a proposed pricing change impacts different customer tiers, export it to Sites, and share it with your sales team to review without any engineering needed
OpenAI said its models and Codex are coming to Google Cloud, with GPT-Astra, GPT-6.1 Sol, and GPT-6 Luna in private preview for select enterprise customers
OpenAI expanded ChatGPT ads to 26 more countries, bringing the total to 89, with 100,000+ advertisers; only Free and Go plans show them
Anthropic has launched Claude Duet, a new standalone app where Claude operates as a proactive agent that monitors your connected apps, flags what needs attention, and completes tasks in the background. The research preview is now available to Pro and Max subscribers in the US. Duet keeps working when your phone and laptop are off. It also carries over your Claude memory, skills, and connected apps, and Anthropic lists 6,000+ connectors, including QuickBooks, Stripe, Gusto, Docusign, HubSpot, and Shopify
Perplexity released Decider for Mac, an on-device app running a compact 4B pplx-decider-flash-v1.2 model that can classify text or images privately without an API key
Amazon renames Kindle Unlimited to Kindle Plus, and expands access to audiobooks, + adds AI-narrated audiobooks for $11.99/mo
1Password launched a new platform that lets AI agents log into user accounts without ever seeing their credentials. The service, 1Password for Agents, creates a verification system where agents request access, users approve the attempt, and 1Password fills in credentials on the agent's behalf, keeping passwords hidden from the AI
Zoom announced deepfake risk detection that analyzes live video and audio for signs of alteration and shows a confidence indicator, due in December
Canva released Canva Craft, its first proprietary image model. Itâs trained on design templates rather than photographs and generates assets up to 2K across 30 different aspect ratios
Replit has launched a Monitoring Agent that acts as an on-call engineer. When something breaks in production, it reads the logs and metrics, finds the cause of the breakage, validates a fix, and emails you a summary
Granola Snapshots can pin meeting feedback to specific Figma frames. Once Granola is connected to Figma, each piece of feedback from a design review is attached to its frame, so you can find which comment was for which screen
Walmart launches Agent Gateway for AI shopping while Amazon blocks agents, creating first major retail split on agentic commerce; Walmart's MCP server integrates ChatGPT, Gemini, and Muse with spending controls and dedicated agent profiles
Block announced Dessa, a speech-to-speech model trained from scratch using nothing but human conversations. Dessa will laugh mid-sentence if something is funny, finish sentences for the user if they forget the name of a movie or song, and also get louder when asked to speak up, which Block says nobody taught it
AI uncovered the most significant PHP vulnerabilities ever in a language powering 77% of sites. Amazon was attacked before the flaw went public, and Governments and companies scramble to patch
đ€ Futurism
OpenAI will build a humanoid, with a 2027 demo planned
đ Good reads
Foursquareâs founder reflects on a heartbreaking $0 outcome for shareholders. Dennis Crowley shares how Foursquareâs 17-year journey ended in an acquisition that wiped out both preferred and common equity, despite having built a $100M+ revenue business. Heâs candid about leadership missteps (including his own), and itâs a reminder that great products, strong culture, or market impact donât guarantee a good exit
On not doing 996. Elena Verna argues that grinding 70-hour weeks in AI isnât a competitive advantage, and itâs just borrowing against employeesâ future health, families, and performance. She shares how strict boundaries, seasonality in effort, and a 5-year regret test help her stay ambitious at Lovable without skipping precious family time
Fear-based decisions are running a lot of your meetings. Anne-Laure Le Cunff explains how we often stay quiet or play it safe at work because our brains overreact to social threats (embarrassment or criticism). Fear-based choices feel safe in the short term but limit learning, impact, and career growth. She suggests simple habits, like metacognition, experiments, and learning in public, to be more curiosity-driven
Anthropicâs AI-native SDLC could redefine how software gets built. Anthropicâs playbook is an AI-native SDLC where each phase hands off via shared markdown artifacts in git (like intent.md, spec.md, plan.md) that both humans and agents read and write. Is this a sketch of the next paradigm in software development?
đ§ Good listen/watch
[Video Podcast] Bill Gates warns of 20 âtumultuousâ years of AI disruption. Gates tells Nilay Patel that AI is both the most dangerous and most beneficial technology humans have ever built. He expects humanoid robots driving blueâcollar automation within ~20 years, and a deep split between people who live at the tech frontier and those who opt out
[Video Podcast] An 86-minute guide to finding PMF with Rob Snyder. Rob Snyder shares his framework to help founders systematically find product-market fit. He gives practical examples on sourcing better ideas, running early sales calls, and avoiding certain traps
đ§âđ» Worth learning
[Podcast] Inside comp negotiation for PM roles at top AI companies. This episode breaks down how PMs can use structured due diligence and role-alignment calls before ever countering. Colin Lernell explains how OpenAI and Anthropic handle no-negotiation policies in practice, and when taking a lower level can be a wise move. He also shares a role-alignment doc thatâs unlocked huge equity bumps at places like Apple
4 practical rituals to drive AI adoption on Product teams. Figma, Sephora, Mercari, and Atlassian show how they make AI stick by using it as a team. Simple norms, like starting every crit with an AI-generated prototype, help make AI a regular part of how teams ship, and are making teams more efficient
đ§ Products to try
Peter Yangâs free AI skill to turn raw videos into polished shorts. Peter Yang, a prolific PM content creator, introduces /video-short, a free AI skill that transforms a simple talking-head recording into a polished short with captions, B-roll, motion graphics, and optional AI-generated video and music. You can install it in your favorite AI agent, point it at a raw video, and it will do its thing
Idealab.ai. Bill Gross has started 175+ companies through Idealab. He wrote down 31 of the methods he uses to turn real-world problems into businesses, refined them with AI, and found the output beat his own brainstorming. He canât build them all, so heâs giving them away: 1.2 million+ business ideas, free
OpenAI's Decisions API sorts, ranks, and routes batches of up to 400 questions with short answers, and OpenAI says it runs up to 10x faster than using GPT-6 Luna
WhatsApp MCP lets your AI agent search and read your own WhatsApp chats (read-only, but it's unofficial and against WhatsApp's terms)
Thatâs all for this week.
Feel free to drop your comments/questions/feedback. Would love to hear what youâd like to see more about in WiP, so I can make it better for you.
Have a great start to your week!
