Hello!
Today I'm poking fun at one of tech's favorite talking points: that "taste is what will matter in the future." I half-joked you could turn it into a drinking game. But underneath the joke is something I've been seriously working on, which is what actually matters when you're managing lots of AI agents at once.
Here's the real tension. AI products have moved fast from models behind the scenes, to chat, to agents, to agents orchestrating other agents. And chat threads just break at that scale. You manage agents, they manage agents, those manage more agents. Trying to run all that through a single conversation thread is like trying to command an army with one-on-one chats.
So I keep coming back to the question of what UI meets this moment. Satya Nadella framed it as "macro delegation and micro steering," and I think he's right that it's a bigger shift than chat was. Anyone who's tried to manage a team through a single Slack thread knows a conversation thread isn't enough.
That's what we're building in Intent, with isolated workspaces, a living spec that keeps parallel agents aligned, and a new stacks UI that makes seeing, knowing, and steering large numbers of agents easier. We shipped a batch of updates just yesterday too: stacks header controls, moving stacks around with undo, cross-workspace note search, and token usage breakdowns by agent and model.
So sure, drink every time someone says taste will matter. But the practical work right now is figuring out the interfaces that let you stay on top of many agents without babysitting every one of them.