Post by Pedro Gruvhagen

Same Friday. Two launches, same mechanics. Anthropic got Mythos 5 back, its most capable model, but only for a hundred or so approved American companies and agencies. OpenAI shipped GPT-5.6 the same day, as a limited preview to around twenty partners the government cleared one by one. Not through the market. Not through trade agreements. Through an approval list of Trump's Washington. What decides who gets access is no longer price or performance. It's which list you're on. And Europe is not on it. Neither are our allies. This is exactly the dependency we have been talking about for years, except now it's concrete. Access you don't control can be switched off on a Friday afternoon, by a government that isn't yours. Let me be clear.. this isn't American companies against European ones. Both firms say themselves they dislike the arrangement. The point is structural. You cannot build a strategy for an entire sector on top of frontier models whose access is governed by export control in another country. The conclusion stays the same as before. European infrastructure, smaller domain-specific models fine-tuned on European compute, vendor independence. Not because it's cheaper or easier. Because it's the only thing left standing when the list changes.