TBPN: Google 'wasn't AI-pilled enough' — selling TPUs vs. scaling tension
zacharynado · x · 2026-11-20
A TBPN clip argues Google as a company "wasn't sufficiently AI-pilled": monetizing compute via TPU sales frustrated researchers and may have driven historical exits, while if you believe you're in the foothills of the singularity you'd want to hoard compute — a structural tension. DeepMind is credited for foundational research but described as lagging organizationally. The attached (unverified) chart shows a "Gemini 4 Neon" benchmark table beating Claude Opus 5.5 and GPT-6.2 Sol on most tasks, plus a CursorBench 4.1 cost-vs-score scatter where it leads on price-performance.
More from Companies & People
- The AI Bill Is the Cheap Part: 30-Engineer Teams Waste $1M a Year on Bad Processes — blaizedsouza · 2026-11-20
- Why Anthropic Is Structurally Positioned to Win: Talent Rails Plus Pure Intelligence Monetization — matt_slotnick · 2026-11-20
- Granola Cofounder on Designing Calm AI for Users Who Don't Want It to Be the Main Character — soleio · 2026-11-20
- Why Anthropic's "rails" strategy lets it attract talent hyperscalers can't — matt_slotnick · 2026-11-20
- Hyundai now owns 100% of Boston Dynamics — wavefnx · 2026-11-20
- Ex-AWS and poolside biz lead joins Hugging Face to push open-source compute — huggingface · 2026-11-20