Jeff Dean — Google's 30th employee, a 27-year veteran, and one of the most influential engineers in company history — is stepping down to co-found and run Discovery Loop, a public-benefit corporation aiming to use AI to "automate complete experimental loops" and dramatically accelerate scientific research. He's joined as co-founders by Sanjay Ghemawat (Google senior fellow), Quoc Le (Google Brain co-founder), and Oriol Vinyals (DeepMind senior research scientist). Funding is co-led by Radical Ventures and Khosla Ventures, with Kleiner Perkins, Lightspeed, Doerr Capital, and Alphabet itself participating as a founding investor and cloud partner.
The exits land alongside a broader Google DeepMind leadership reshuffle: Demis Hassabis is moving from day-to-day CEO to Chair of DeepMind and Chief Scientist of Alphabet, framing it as a chance to focus on AGI strategy ("we have arrived at a pivotal moment in human history"), while 13-year DeepMind veteran Koray Kavukcuoglu is promoted to SVP overseeing Gemini development, frontier research, and the Gemini app (now 950M+ monthly users; Gemma downloads past 900M). Discovery Loop is also explicit about pursuing recursive self-improvement — using AI to help build more powerful AI with less human iteration in the loop.
NewsAPI Step 1b main query returned 0 results again this run — the URL specified in this run's own instructions still joins every term with + (AND, not OR). Same unresolved news-agent/CLAUDE.md line 62 issue flagged for 30 straight prior reports (since 2026-07-07) — now 31. This run's OR-built workaround query surfaced 42 raw hits (10 PyPI-adjacent, remainder mostly celebrity/lifestyle/science noise), 0 additional genuine signal beyond what RSS/WebSearch already caught. Recommend the fix land directly in CLAUDE.md line 62 (swap + for OR between phrase-quoted terms) since report/memory writes alone can't patch it.
Resolved: Goodie AI's blog has been unverifiable for 11 consecutive days because the sweep was checking the wrong domain (goodie.ai). This run confirmed the real domain is higoodie.com, surfacing two previously-missed fresh posts. Recommend adding higoodie.com/blog to the Step 1a RSS list in a follow-up commit.
Shopify's product changelog RSS (changelog.shopify.com/feed.xml) returned HTTP 500 this run — a new failure mode for this feed (it had been healthy since early August). Direct-fetch fallback to the changelog page itself worked fine and confirmed no new post beyond Aug 4's already-logged Microsoft Advertising item, so no signal was lost, but the RSS endpoint itself is currently down.
shopify.dev's changelog RSS (shopify.dev/changelog/feed.xml) returned 404 this run — different failure mode than the 500 error that resolved on Aug 5. Direct-fetch fallback to the changelog page worked and surfaced a genuinely fresh Aug 5 item (WebMCP support for Liquid/Hydrogen storefronts).
Otterly.ai blog WebFetch returned empty content again this run — consistent with the recurring extraction failure seen most days over the last two weeks. No new post confirmed beyond Jul 30's already-logged Ocean Cleanup piece.
Firecrawl: 2/3 scrapes clean this run (TechCrunch's Jeff Dean/Discovery Loop piece and TechCrunch's Anthropic chip-team piece scraped cleanly); the TechCrunch Shopify AI-search page was again fully blocked by a Cloudflare Turnstile challenge with zero article content recovered — same recurring failure mode as Aug 4 and Aug 5's blocked pages — backfilled via WebFetch/WebSearch instead.