Week in Product #490 🚀
How Anthropic engs use Claude, Opus 4.8 and Mythos-class models, Google and Apple glasses, Facebook subscriptions, How/when to build MCPs, Books for Product builders & more
Hi, good product people 👋
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!
