Right. Come here. Look at this. Google — the company with the biggest kitchen in the world, its own farm, its own chips, its own model on the menu — just told every engineer in the building they're allowed to cook with the competitor's knives.
Reported Monday, first by Business Insider and confirmed by a Google spokesperson: all Google engineers can now use Anthropic's Claude Opus 5 for internal work, inside Antigravity, Google's in-house development platform, on per-user quotas. Until now, Google barred most employees from outside coding tools — Claude Code, OpenAI's Codex — and pointed them at Gemini. The only exceptions were a few Google DeepMind teams and "high-priority" projects, according to three people who spoke to Business Insider. Everyone else ate what the house served.
Why the chef changed the menu
Because the brigade complained, and they were right. Engineers had "grown frustrated with Gemini's limits on coding tasks," the reporting says, even after recent improvements, with Anthropic and OpenAI continuing to perform better on coding. Productivity pressure on one side, a padlocked pantry on the other. That's not a strategy, that's a hostage situation.
Google's statement, and I'll read it exactly: "Engineers have access to select third-party models in Antigravity, which is aligned with our external Antigravity enterprise offering. Gemini remains our primary and foundational model for internal development, with third-party models available on a quota to support specialized use cases."
"Specialized use cases." The specialized use case is writing code. At Google.
Let's be fair, because I am always fair. Gemini is a serious model — its Pro tier is the best reasoning-per-dollar on the market by most counts, and Google just shipped a voice model that undercuts OpenAI's. This isn't a bad kitchen. It's a kitchen that finally admitted no single station makes every dish best. And Google is hardly a stranger to Claude: it's an Anthropic investor that has said it plans to put up to $40 billion into the company. Amazon, Anthropic's other giant backer, made the same move earlier after its own staff complaints, opening Claude and Codex to employees. The two companies that fund Claude both let their engineers pick it over their own AI. Read that again.
Tasting notes from the pass
- The timing is delicious. The same day, Anthropic quietly cut Claude Code's weekly usage by about 17% — ending a temporary 50% boost and replacing it with a "permanent" 25% bump. Portions shrink at the exact moment demand from Mountain View arrives. That's not a coincidence, that's a menu price.
- "On a quota" is doing a lot of work. Google isn't switching. It's rationing the good stuff and telling you the house wine is still the house wine.
- The scoreboard changes weekly. Claude Opus 5 topped the independent intelligence index this summer at half the price of the tier above it. GPT-6 Astra leads graduate-level science reasoning. Gemini leads on price. DeepSeek leads on cheap. Nobody leads on everything, and anyone who tells you otherwise is serving you a frozen lasagna and calling it fresh.
The verdict
Here's the lesson, and it isn't about Google. If the company that built Gemini won't let its own engineers rely on Gemini alone — if it needs Claude on the line for the hard tickets — then you, with one chatbot tab open, are running a kitchen with one burner and calling it a restaurant. Every model is brilliant at some plates and raw in the middle on others, and none of them will tell you which is which.
So do what Google just did, only without the memo and the quota. Put the same question to ChatGPT, Claude, Gemini, Grok and 30 more at once and let them check each other's work — one cross-examined, cited answer instead of one model's confident guess. That's the entire point of Gangsta AI: every top model on the pass, and the one that's undercooked gets sent back before it reaches your table.
Now go and see which models are actually earning their station tonight on our best AI models rankings. And for heaven's sake, taste before you serve.
Sources
- Dataconomy — Google opens Claude access to engineers across the company (spokesperson statement, $40B investment plan)
- TechBriefly — Google grants engineers access to Anthropic Claude for coding (prior policy, DeepMind exceptions, Amazon precedent)
- Seoul Economic Daily — Google lets engineers use rival Anthropic's Claude
- dev.ua — Google has opened access to Anthropic's Claude to its engineers (Business Insider report summary)
- Analytics India Magazine — Anthropic cuts Claude Code's usage limits by 17% after slashing promotional boost
- Hero photo: Googleplex, Google headquarters in Mountain View, by Austin McKinley — Wikimedia Commons (CC BY 3.0). Byline avatar: Gordon Ramsay photo by Dave Pullig, Wikimedia Commons (CC BY 2.0)
Originally published on Gangsta AI News. Gangsta AI lets you compare 30+ AI models side by side on a single prompt — free, no login. Try the AI model comparison tool.
Top comments (0)