The AI Token Apocalypse: Why Costs Are Exploding

The U.S. government restricted Anthropic's Mythos class of AI models on June 12, 2026, adding uncertainty around closed AI systems and shifting the AI conversation from hype to cost and productivity. Companies are increasingly questioning AI's measurable business value, accelerating interest in open-weight models amid growing divisions between closed and open AI approaches.

Member-only story The AI Token Apocalypse: Why Costs Are Exploding Governments, business leaders, and the race for productivity are reshaping how AI is adopted worldwide. Hello AI world. I took a short break over the last few days. But while many people were enjoying the summer, the AI world kept moving — and some important signals started appearing. One thing is becoming harder to ignore: AI is still struggling to prove its productivity value at scale. In my view, June 2026 marked an important shift. On June 12, the U.S. government sharply restricted Anthropic’s Mythos class https://www.anthropic.com/news/fable-mythos-access of models. Beyond the policy itself, the decision added more uncertainty around closed AI systems and companies such as Anthropic and OpenAI. This feels very different from 2025. Last year, the biggest AI story was vibe coding and the promise of AI agents. In 2026, the conversation has moved toward cost, token efficiency, and measurable business value. Many companies are asking a simpler question: Is AI actually making us more productive? That pressure is accelerating interest in open-weight models, especially outside China. At the same time, AI is becoming connected to class struggle https://en.wikipedia.org/wiki/Class struggle , geopolitics, and public trust. I have rarely seen the AI community this divided between closed and open models.