Google exec: new Gemini releases now tested by thousands of SWEs for weeks before launch
inductionheads · x · 2026-10-01
Logan Kilpatrick of Google says the company has gotten much better at testing models at scale: most new Gemini revisions now go through thousands of software engineers for weeks of internal use before release, which he hopes has meaningfully narrowed the gap between benchmark scores and real-world performance.
More from Companies & People
- Google Called the Bell Labs of AI: Transformer, MusicLM, MoE, TPU All Born There — AccBalanced · 2026-10-01
- AI Scout: free tool tracks every researcher leaving OpenAI and joining Anthropic — Xianbao_QIAN · 2026-10-01
- Singapore punches above its weight in AI as the world's No.4, but Claude's web integration lags OpenAI — teortaxesTex · 2026-10-01
- AMD Buys World Labs for $8.2B, Anthropic S-1 Leaks $518B Compute Commitments — 20VC · 2026-10-01
- xAI reportedly uses Grok Bot agents to build Grok Bot, from design to merged PRs — elonmusk · 2026-10-01
- Sakana AI hiring first product sales lead to build enterprise GTM from scratch — SakanaAILabs · 2026-10-01