Before it had a name

A while back I wrote about how around a year ago, a few of us put together a rough, small on-device agent at work in just a week as a proof of concept. Not long after that PoC landed, there was a lot of excitement around what it could do, so we kept working on it. At first, feedback was that it was a neat trick, but not the revolution that agents had promised, so it became a lower priority. That didn’t last long.

The on-device agent kept coming up and drawing interest, and over the next few months we made real progress. Not a polished product, mind you, but it was enough to get permission to keep working on it. None of us had ever built anything like it. We were backend engineers; most of what we had written had no UI. Now we needed a UI that would run on multiple platforms, so we went looking for the best cross-platform options with the fewest moving parts. We ultimately decided on Flutter for the front-end. The backend we wrote in Go. We even had a terminal client written in Bubble Tea. On paper, it should have taken us forever to ship, after all, we didn’t know Flutter or Bubble Tea, but we shipped it in a little over six months.

Agentic tools made it possible. “Learn a new framework and build something with it” used to be a quarter of ramp-up before you wrote anything significant. With an agent riding along, learning Flutter was building in Flutter. We’d ask how something worked, get a working example against our actual code, and keep moving. If the code didn’t need to be production quality, you could move really quickly.

In those six months, we delivered a real client, backed by a real service, that could actually do things on your behalf. This wasn’t a “look, it can talk,” but “it can go do the thing you asked.” The reactions told the whole story. Some people lit up; “wow,” “I had no idea we could do this.” Others squinted: “…why do I need this again?” Both were correct. It was impressive but not yet obviously necessary.

But I’d stopped wondering whether this was real. Watching a tiny team, fitting this in around everything else, ship a cross-platform agent that did real work in a stack we hadn’t known six months earlier convinced me that agents that can actually act on your behalf weren’t a someday thing.

Postscript, from later: about five months after this, an open-source project called OpenClaw showed up with more or less the same idea, out in the open. We all know the story after that. Within months the whole internet had found it. We’d been quietly building toward the same idea for a year, without a name for it. Watching it land felt like confirmation: we had the right idea all along.