Updated
Updated · 9to5Google · Jul 16
Google Tests Gemini 3.5 Pro With Partners After Missing June Launch Over Coding Setbacks
Updated
Updated · 9to5Google · Jul 16

Google Tests Gemini 3.5 Pro With Partners After Missing June Launch Over Coding Setbacks

3 articles · Updated · 9to5Google · Jul 16

Summary

  • Google said it is testing Gemini 3.5 Pro, an upgraded Flash model and other models with partners after the Pro version missed the June launch window promised at I/O 2026.
  • Late-June training-data updates aimed at boosting Gemini 3.5 Pro's coding skills produced disappointing results, Bloomberg reported, suggesting development was reset after Google touted major gains in mid-May.
  • 75% of new code at Google was AI-generated and engineer-approved as of April, up from 50% last fall, even as some engineers argue important code should still be written by humans.
  • Internal AI coding efforts are also constrained by tool capacity and fragmentation, with Google trying to unify separate systems across DeepMind, Cloud and Android while rivals press ahead in coding-focused AI.

Insights

Is Google’s AI delay a sign of responsible caution or a deep crisis against its fast-moving rivals?
Could decentralized training end the AI dominance of tech giants by slashing development costs and failures?