The AI race has become suffocating.
OpenAI and Anthropic used to trade blows on a roughly ten-week cycle. That interval has now been crushed to about eleven days. No one inside the arena gets to sit this out. Google DeepMind has been pulled in as well, and just overnight it rolled out Gemini 4 Argon, the model carrying the hopes of the whole village.
It looks like a launch. It also doesn’t.
Argon is not going fully open. It is being released first through the “Fairwind” program to a small group of trusted cybersecurity defenders, and it is going through the U.S. government’s voluntary pre-release model access process. Broader access for developers, enterprises, and consumers comes later. The first wave is paid API customers and Google AI Ultra subscribers.
The marketing after the Gemini 4 announcement has been triumphant, as if OpenAI and Anthropic were about to be swept off the board. At the same time, Bloomberg reported that Google’s own employees are privately dumping on the new model, saying it is painfully weak at real-world coding: high scores, low capability.
