Anthropic's CEO called for an AI slowdown. Then his company shipped two models anyway.
The second one is Sonnet 5.5. Cheaper than the last version. Not a frontier model, they promise. Just better at coding and knowledge work. You know, the kind of tasks that don't advance anything important. They're very clear about this. It doesn't push boundaries. It just does the work better. Which is different somehow.
Here's the play: call for a slowdown when you're ahead. Sounds responsible. Gets you on podcasts. Then keep shipping while everyone else pauses to think about what you said. It's the corporate equivalent of yelling "wait up" while sprinting past someone.
The model costs less because it's not frontier. Frontier means scary and expensive. Non-frontier means friendly and affordable. Same way a knife isn't dangerous if you call it a kitchen tool. Anthropic swears this one won't end the world. It'll just write your Python scripts faster and summarize your emails. Totally harmless. Unless you're the guy whose job is writing Python scripts and summarizing emails. Then it's f*cking catastrophic.
Retail traders are already pricing in the AI revolution for the third time this year. They bought Nvidia at the top. They bought the dip. They bought the dip after the dip. Now they're loading up on hyperscaler cloud stocks because surely this time the math works out. It won't. But they'll keep buying because someone on Twitter told them AGI arrives in 2027 and they want to retire by 2028.
Anthropic says it's being responsible by not advancing the frontier. They're just making the current frontier cheaper and better at the exact tasks humans get paid to do. Totally different thing.
Photo by Brecht Corbeel on Unsplash
Leave a Comment