Google reportedly starts its biggest Gemini 4 pretraining run yet
VraserX · x · 2026-07-29
The post says Google has reportedly started its largest pretraining run yet for Gemini 4 and uses that as a springboard for a bigger question: is frontier AI still mostly a software race, or has it become an energy-and-capital race?
The substantive point is not just the model name, but the implication that pushing the frontier now depends increasingly on massive compute, power, and financing alongside model engineering.
More from Infra
- YouTube-style semantic IDs tackle recommender memory walls with dual-purpose tokens — _reachsumit · 2026-07-29
- Meta’s Memory Layer pushes Instagram Reels item coverage to 100% and freshness to 20 seconds — _reachsumit · 2026-07-29
- VaLiDRec uses variable-length LLM-aligned IDs and runs 87.49× faster than LC-Rec — _reachsumit · 2026-07-29
- India’s AI inference market will reward companies that co-optimize models and hardware — santoshpanda · 2026-07-29
- Hermes Agent Desktop impresses users with parallel tools and remote local-model setup — Teknium · 2026-07-29
- Chinese threat actor pivots infrastructure and leaks 775 API-key IDs from an AI reseller — cyb3rops · 2026-07-29