AI × Design Bi-Weekly #001: Opus 4.8, Figma Make, and the death of the handoff

I almost wrote the obvious version of this post: new Claude model, here are the benchmarks, here’s what it means. Then I looked at my own week and realized the model barely mattered. What actually changed is quieter and bigger. The agents stopped living in a separate tab you go and talk to.

This week they moved in, into Figma, into Adobe, into Notion, into my editor, into the Google search bar. They’re in the building now. And once the thing can act inside the software you already use all day, the interesting question stops being “what can it make” and becomes “who’s responsible when it makes the wrong thing, and how do I see it coming.” That’s not a model problem. That’s a design problem, and it’s the one sitting on our desks.

Claude Opus 4.8 and the one button nobody’s talking about

So yes, Claude Opus 4.8 landed on Wednesday, with the usual wink that something bigger (”Mythos”) is a few weeks out. I’ll be honest: I didn’t feel the model. The jump from 4.7 is real but it’s the kind of real you read in a chart, not the kind you feel in your hands. I’ve stopped getting excited about those deltas, and I think you should too. They’re table stakes now.

What did make me sit up was a feature most people scrolled right past. You can now tell the model how hard to think. There’s an effort dial — spend more on this, spend less on that — and Claude Code picked up “dynamic workflows” for the big, sprawling, multi-step jobs.