- Product & Design Pulse
- Posts
- Product & Design Pulse v95
Product & Design Pulse v95
What's Under the Hood 🔍
Malcom Jones
June 08, 2026
Welcome to this week’s edition of Product & Design Pulse, where we explore the latest in tech, product, design, and innovation! Last week forced the industry to confront what's actually happening under the hood. The Anthropic Institute published the most revealing internal data any frontier lab has shared, showing that over 80% of its production code is now written by Claude, engineers are shipping 8x as much work as two years ago, and Claude Mythos Preview solves open-ended engineering problems 76% of the time. The report explicitly raises the possibility of recursive self-improvement and calls for a verifiable global mechanism to pause development if needed. Google, meanwhile, made two moves that signal how seriously it's preparing for an AI-dominated future: it quietly began paying Play Store developers for proprietary app code to train its AI coding tools (without mentioning AI in the pitch email), and it issued $10 billion in equity to Berkshire Hathaway, a move Ben Thompson argues signals that even Google's $126 billion cash pile won't be enough to fund the compute buildout ahead. And in Brussels, Meta won a partial DMA victory when the EU court struck down the Marketplace designation on procedural grounds, while upholding Messenger's gatekeeper status. The week's message was unusually direct: AI is accelerating AI development faster than most people realize, the companies building it know it, and the race for capital, data, and regulatory positioning is intensifying accordingly.
🎧 Audio Overview [BETA]
For those who don’t have time to read 😁 |
Last week…
Meta Wins DMA Challenge on Marketplace, Loses on Messenger
The EU's General Court ruled that the European Commission made legal errors when it designated Facebook Marketplace as a "core platform service" under the Digital Markets Act, finding the decision "lacks sufficient reasoning" and relied on "hypothetical and incomplete" analysis. However, the court upheld the designation of Messenger, agreeing it is distinct from the Facebook social network and rightly subject to gatekeeper obligations. For the regulatory landscape, this is a mixed signal: the DMA can be successfully challenged on procedural grounds, but its core framework for regulating Big Tech gatekeepers remains intact.
The Anthropic Institute published a detailed report showing that over 80% of merged production code at Anthropic is now authored by Claude, engineers ship 8x as much code per quarter as they did from 2021-2024, and Claude Mythos Preview's success rate on open-ended engineering problems hit 76% in May 2026, up 50 percentage points in six months. The report outlines three possible futures: capabilities plateau but current AI diffuses widely, compounding efficiency gains continue with humans still setting direction, or full recursive self-improvement arrives where AI builds its own successors. For the industry, this is the most transparent look any frontier lab has published at how close we are to AI systems that meaningfully accelerate their own development, and Anthropic is explicitly calling for a verifiable global coordination mechanism to slow or pause if needed.
Google Quietly Paying Play Store Developers for App Code to Train AI
404 Media reveals that Google has been emailing select Play Store developers with millions of downloads, offering to buy access to their codebases through a "confidential content offer pilot" that never mentions AI in the email itself, though a link leads to a page about "partnerships to improve our AI products." Developers retain full IP through a non-exclusive license, and the program targets both active apps and archived prototypes. For the AI coding race, this is Google acknowledging that publicly scraped code isn't enough to compete with Claude Code and GitHub Copilot, and that high-quality proprietary training data now has to be purchased directly.
Thompson analyzes Google's unusual decision to issue $10 billion in equity to Berkshire Hathaway rather than taking on more debt, despite having $126 billion in cash and ample borrowing capacity, arguing that the deal signals Google believes AI compute demand will be so large that even its massive free cash flow won't be enough to fund it. He frames the deal as a signal to the market that Google expects AI Cloud revenue to eventually rival or surpass its advertising business, making the dilution worthwhile if Cloud margins continue expanding. The essay positions Google as transitioning from a company defined by search advertising to one defined by capital deployment at infrastructure scale, where the ability to raise and spend money becomes as important as the products it builds.
