Hi friends 👋
Welcome to a new Week in Product!
🎰 The week in figures
$852B: OpenAI completes $7B employee tender offer at $852B valuation. The buyback provides liquidity without an IPO, signaling the company may delay its public debut while focusing on enterprise growth
$5B: Databricks, the data and AI platform raised $5B at a $190B valuation
$1B: Cognition AI, the coding-agent startup, is in talks to raise $1B at a $40B valuation
$1B: River AI raised $1B to build home and small-business servers that run and customize AI models locally
1B: Sundar Pichai announced Gemini reached 1B MAUs faster than any other Google product; powering Gmail, Drive, Search
$500M: Harvey, the legal AI startup, is in talks to raise at least $500M at a $15.5B valuation to develop proprietary models and compete with rivals including Legora
$400M: Lovable raises $400M at a $13B valuation as enterprise demand for AI-assisted app-building accelerates
$150M: Kevin Weil, the former OpenAI product chief, is seeking $150M at a $750M valuation for his unnamed AI model training startup
2K: Uber and Pony.ai will deploy 2,000 robotaxis across Europe, expanding beyond Zagreb into four additional cities
📰 What’s going on
Google launched Gemini 3.7 Flash for coding and AI agents. They claim sizable gains in software engineering and business workflows that come at half Gemini 3.6 Flash’s original per-token price through 2026
Google Play added Venmo as a payment option, giving developers another way to capture growing consumer spending on apps and games while reducing checkout friction for Venmo users
Apple is training a China-specific AI model with Alibaba, while Apple Intelligence will also use Qwen
Apple's glass-centric 20th-anniversary iPhone stays on track for 2027, with Pro models using glass front and back curving into the sides
Anthropic makes auto mode Claude Code’s default for Pro, Max, and Team plans, replacing reflexive permission prompts with a classifier that apparently catches more dangerous commands while enabling longer autonomous runs
Claude is adding invisible statistical watermarks to output from new models, encoding the signal into token choices so it can survive copy-paste and light edits
OpenAI began previewing Ultrafast, a new setting that runs its GPT-5.6 Sol model up to 14 times faster, generating up to 750 tokens per second
OpenAI launched Computer History, an opt-in feature that lets ChatGPT recall what you did on your Mac across apps and sites without taking constant screenshots
OpenAI's first device leaks as a screenless, doughnut-shaped ChatGPT speaker with moving parts, priced $300-$400 for a 2027 release
ChatGPT now books restaurants through Yelp Guest Manager across the US and Canada, though changes still need the Yelp app
DeepSeek shipped V4-Pro, adding adjustable "thinking" levels so developers can dial reasoning up or down depending on the task, plus cheaper off-peak pricing
SpaceXAI unveiled a bot working like a team of AI agents to compete with OpenAI, Anthropic, launched alongside Grok 4.6, which the company says reaches frontier-level performance over Grok 4.5 at the same price
Meta's new Glimmer model offers a first hint at Zuckerberg's personal intelligence vision (his vision: a future where everyone has a personal AI agent providing PhD-level tutoring, business creation tools, and lifestyle optimization). The 30B parameter model is essentially an open version of Muse Spark. Zuckerberg pledged to release more open-weight AI models
Facebook launches AI-powered Creator Studio app for iOS. The standalone app gives creators an AI assistant for personalized growth recommendations, auto-drafted comment replies in their own tone, and daily priority feeds as Meta competes with TikTok and YouTube for creator attention
Microsoft merges consumer and business Copilot apps. It kills AI podcasts, Group Chats, Deep Research, and its Clippy-style mascot Mico as it consolidates into a single experience by August 18
Spotify will tag AI Persona profiles with a badge and cut their music out of recommendations as part of wider AI policy changes
Klarna is supercharging its European membership tiers, from €4.99 to €44.99 a month, packed with cashback, lounge access, and up to €6,000 in annual perks. This follows the same playbook Klarna ran in the US about eight months ago
🤖 Futurism
Lab-grown brain organoids could outdo silicon AI on energy use and adaptive learning, though technical and ethical hurdles block progress
Robot dogs are being deployed to patrol instead of human security guards
📚 Good reads
Anthropic found that when it set several Claude agents loose on the same project without telling them about each other, they sabotaged each other with self-replicating malware before eventually negotiating a truce
AI agents are stepping outside the box. Following the story above, Andrea Filiberto explains how recent sandbox escapes by models from OpenAI, Anthropic, and Moonshot aren’t signs of rebellion, but of capable agents exploiting overlooked paths to hit their goals. This behavior is both a safety concern and a powerful marketing narrative in the AGI race, especially around cybersecurity. See AGI podcast below
How to move the needle. John Cutler argues most teams overcomplicate ways of working with frameworks and cycles, when strategy lives in front-line teams shipping small, learning fast, and iterating on durable lanes. He pushes for simple decision-rights, light routines, outcome-centric lanes, and accepting that different work has different shapes
How to build a better feedback system for professional development. Most feedback is vague and socially awkward, so Adam Fishman lays out 5 tactics to “engineer” clearer input, including self‑assessments, peer feedback, and an AI “manager mirror.” The core idea: stop relying on one off-the-cuff, vibes-based conversation and design a repeatable system that turns feedback into actionable growth items
How we use computers is about to change forever. Peter Yang argues we’re shifting from manually using laptops to orchestrating cloud agents with our voice. Personal computers will increasingly live in the cloud, with products designed as tools for agents vs traditional SaaS apps
Most people want IC roles. Elena Verna digs into a 300+ ppl study and finds that, if pay and prestige were equal, most managers and ICs would pick hands-on work over people management. Money-trapped managers are often burnt out and stretched thin. We need more high‑impact IC paths with real influence and upside
🎧 Good listen/watch
[Short video + article] What changes when AI keeps learning after release? Dwarkesh Patel is one of AI’s biggest podcasters, and he’s long argued that continual learning (an AI updating its own internal settings from experience, so work today makes it better tomorrow) is a missing ingredient in today’s AI. Continual learning will transform AI models from software you buy into something closer to an employee you develop. That changes safety, competition, privacy, pricing, and lock-in
[Podcast] Entering the AGI era. Joshua Achiam, Chief Futurist at OpenAI, says we may have already crossed into an AGI-level world. This is most visible in cybersecurity, where models are finding software vulnerabilities faster than humans can patch them
🧑💻 Worth learning
Google Analytics’ new AI-powered tools. GA has quietly shipped a bunch of features that help you find what you do next. Avinash Kaushik walks through the new AI features
🔧 Products to try
Claude in Chrome now turns the browser side panel into a Cowork session that keeps history, Skills, connectors, and multi-step tasks synced across Chrome and Claude’s apps; available on Max and Team plans, rolling out to Pro users over the coming weeks
Adobe Podcast cleans noisy recordings, removes echo, and lets you edit speech from a transcript so rough audio sounds studio-ready; free plan available
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!
