Hi friends 👋
Welcome to a new Week in Product!
🎰 The week in figures
$50B: OpenAI told investors it reached roughly $50B in annualized revenue at the end of September, below a circulated $68B figure that included partner gross revenue
$40B: Alphabet’s Isomorphic Labs is in funding talks at a valuation of at least $40B. The company was formed out of DeepMind in 2021, and researches potential medical uses for AI technologies
$15B: Reports indicate that DeepSeek’s latest funding round could reach $15B
$2.3B: Uber will acquire workplace-meal marketplace ezCater for $2.3B in cash, as Uber Eats expands into catering
$475M: Oratomic, the quantum startup raised a $475M Series B at a $5.4B valuation, up from $1.5B after its $300M Series A in July
$200M: Arena, the AI model benchmarking platform is now valued at $3.1B after raising a $200M round
$200M: Ledgebrook, an AI-powered specialty insurance platform, raised $200M at an undisclosed valuation
$150M: Valon, an AI-powered mortgage-servicing software provider, raised $150 million in a Series D at a $2.3B valuation
30K: Meta and Microsoft cut back internal Claude use, steering staff to their own coding tools; Meta's Claude Code users fell from about 60,000 to 30,000 (!!)
372: OpenAI published 372 AI-generated advances across mathematics and theoretical computer science. Many of these were formally verified, but OpenAI withheld its exact model and prompts used
7%: HubSpot is cutting 7% of its workforce (roughly 660 jobs). CEO Yamini Rangan denied that automation prompted the layoffs
3%: top Chinese AI models trail U.S. rivals by just 3% on benchmarks, down from 15% earlier this year
📰 What’s going on
Anthropic released Claude Haiku 5.5, their low-cost, low-latency option. This release most significantly impacts developers running high-volume agentic workloads
Anthropic launched Claude Dashboards. It connects to company data, including systems like Salesforce or Snowflake, and builds charts that update as the underlying numbers change. Ask why sales dropped, for example, and Claude can build a live view of the answer
Anthropic also launched Claude Motion, which turns a report, chart, or idea into a short animation. It generates code instead of AI video footage, so all the content stays editable
As of October 6, new Cowork tasks for Claude Pro and Max run on Anthropic’s servers, and the “Only on your computer” option is being removed. Existing local tasks stay local, while new sessions can keep running after your laptop closes
OpenAI is rolling GPT-6 and its “Intelligent UI” in general availability. The model can assemble interactive controls and visual tools around the task in front of you. All plans, including Free, have access now
OpenAI says it will begin testing clearly labeled visual ads during image generation in the US later this month. With ChatGPT reaching a claimed 1.2B users each week, this is a huge storefront
Google announced Gemini Agent for work: persistent cloud execution, memory, and connected work tools support tasks that run for hours or days. The agent can have its own work identity, with email, calendar, and Drive storage
Google froze its open-source bug bounty program after a flood of AI-generated reports, full of hallucinated functions and fake code paths, overwhelmed reviewers
Apple will tighten macOS "Full Disk Access," requiring very explicit consent, after warning that AI agents sharply raise the risk of apps reading files, mail, and messages
Apple and LG co-developed a smart home line including doorbell, lock, thermostat, and cameras, launching after Apple's hub debut
Meta, Walmart, Stripe, and Sierra unveiled an open protocol for authenticating AI agents and monitoring their actions, giving businesses visibility into whether they are engaging with a human or a bot
Zuckerberg reportedly rushed out Meta’s Muse AI agent after rival startup Instinct’s comparable product began gaining traction
Mistral unveiled Le Chonk, a trillion-parameter model that it claims is the strongest open-weight system built outside China. Le Chonk is optimized for coding and cyberdefense, but intended to compete with general-purpose models
Asana launches Command, an orchestration layer where coding agents get the full context of specs, decisions, and meetings instead of starting from a single pasted prompt
🤖 Futurism
Honda and Taisei will pilot wireless highway charging for EVs at up to 120 kph starting fiscal 2027, allowing batteries to recharge while driving
📚 Good reads
Will AI agents have network effects? Andrew Chen argues that today’s consumer AI agents are mostly solo tools with weak inherent network effects, but they can design in network effects via agent-driven viral loops and shared identity/reputation layers. Agentic viral loops could spread by auto-creating shareable artifacts (microsites, docs, mini‑apps) that nudge others to sign up
Booking.com explains it’s AgentHub platform making building travel AI agents less painful. The platform lets teams spin up configurable, template-based LLM agents that plug into internal tools and data across the travel funnel. Their first use case, an AI Trip Planner Q&A agent, already shows big gains in answer relevance and reduced LLM failures. They’re now pushing toward multi‑agent, multimodal, and more personalized experiences
Uber’s MCP Gateway. Uber built a centralized MCP Gateway for AI agents to call thousands of internal and third‑party services. AutoCrawler auto-discovers APIs and native MCP servers, generating tools that serve teams. On top, features like Omni MCP, response projection, and “code mode” make tools discoverable and cheaper to use in-context
Building cool apps is not enough. Jaryd Hermann reflects on shipping 3 polished apps that went nowhere and argues that today great product no longer guarantees users or revenue. Distribution moats come from doing the uncomfortable growth work, growth hacking, and content. I learned a similar lesson with henrytheapp.com, but after reading this piece, I also feel like I made it with my 70 users 🎉 (thank you!)
🎧 Good listen/watch
[Podcast] The roadmap to becoming a product builder. Aakash and Ankit Shukla (CPO of Freshworks) chat about why Product Builder is the next evolution of PM. The piece walks through Ankit’s practical ladder for shipping AI products, and a concrete playbook for PMs to build projects and portfolios
🧑💻 Worth learning
[Podcast + article] Use cases for Jev: a decision engine for routing, classification, and product logic. The podcast also covers some examples of ChatGPT Sites powering internal dashboards and personal tools
🔧 Products to try
Claude Docs, Slides, and Design are out of beta and available on all Claude plans now, including Free
Moxie lets you use OpenAI, Gemini, and DeepSeek models inside Claude Code (Anthropic's coding assistant) from your Mac menu bar, and it switches accounts for you when you hit a usage limit; Free for 2 providers, $39 for 2 years, or $79 forever
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!
