Google Delays Gemini 3.5 Pro to July and Reorganizes AI Team to Catch Anthropic
Big Tech is panicking again. While everyone was expecting a shiny new AI model, a certain search giant is busy moving desks around and postponing its flagship launch because writing code is apparently harder than selling ads.
The search giant is reshuffling its newly formed AI coding unit, transforming a temporary 'strike team' into a permanent 'midtraining' division. This structural shift targets the most lucrative AI use case on the market: automated programming. Instead of just polishing a model's final outputs, the new team aims to inject coding capabilities much deeper into Gemini during the mid-stage of its training cycle, eventually scaling these skills to broader business tasks like automated presentation building.
This strategic pivot mirrors the playbook of Anthropic and OpenAI, who are rapidly expanding their coding tools beyond pure programming. But the reorganization comes at a heavy human cost, as top-tier talent is fleeing the campus. AI pioneer Noam Shazeer recently departed for OpenAI after his project's computing resources were reportedly redirected to other teams. Adding to the headache, John Jumper, the Nobel laureate who led the groundbreaking AlphaFold project, has jumped ship to Anthropic, shortly after being transferred to this exact coding unit.
The internal chaos has already trickled down to product releases. The highly anticipated Gemini 3.5 Pro, originally announced in May, has been delayed until July as engineers scramble to refine its performance. Currently, the company is collecting feedback on testing platforms like LMArena and its internal Antigravity system to optimize the model for complex agentic tasks and reduce high token consumption, leaving only the smaller Gemini 3.5 Flash available to the general public while Pro remains locked in a restricted enterprise preview.
Moving boxes around and renaming departments is the classic corporate response to a crisis, but it rarely replaces actual innovation. With massive compute power, proprietary TPU chips, and a monopoly on search, the search giant has all the chess pieces to win, yet continues to trip over its own organizational chart while nimble competitors run circles around it.
Source: The Information
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.