Steve Yegge, writing earlier this month about the latest incarnation of his multi-agent harness:

We’ll explore a new harness I’ve been building, called Wheelhouse. It is closed-source, made just for me. I have given up on building reusable harnesses. Indeed I believe harnesses will all soon be bespoke, and the people trying to sell you one will all soon be bebroke. Harnesses need to be part of your application, chemically bonded in. You won’t have any luck with someone else’s “reusable” harness framework. You don’t need it.

While I was on extended leave earlier this year, I started building my own harness, deeply intertwined with the iOS app I’ve built for writing. Had I not returned to work, I almost certainly would have ended up with something like what Yegge describes.

Stepping away from it has given me some perspective. Increasingly, I feel like agent harnesses are just another hobby. A contemporary model railway, if you will. Stringing together a bunch of agents to automate something complex is exciting and dopamine-inducing, but almost none of these harnesses will create value for anyone beyond their creator. And that’s ok!

What Yegge has built is interesting, no doubt. As I said, I would build the same thing for myself. But there is no way the future is a bunch of high-agency individuals running off on their own, building Rube Goldberg-like systems and letting them loose inside a company of any real size or importance.

Nor is it remotely economical to do so today. Agent harnesses at Yegge’s scale are expensive:

I’m only spending about $2800/month out of pocket for my $87k “worth” of tokens. Though that number keeps growing alarmingly.

In order to sustain my pace of development, I pay for currently twelve extra Max accounts, in addition to my personal one.

Only! This is your brain on agents.