Optimising for programmer happiness

A few weeks ago I spent a weekend camping with friends, and what struck me most was the relief of not talking to an agent.

Leading a team through Slack

At work I'm in conversation with them all day. It's like leading a team of developers where the only tool you have is Slack, so you spend the whole day chatting. It's exhausting, even when the results are thrilling.

I wrote last week about how thrilling and draining this way of working can be. The camping trip made it clear to me where a lot of the drain comes from. It isn't the amount of work. It's the constant conversation.

Optimising for something else

So much AI dev tooling is focused on producing more code, and better code. But what if we optimised for programmer happiness instead? Productivity is going to be through the roof either way.

The idea isn't new. Ruby was designed to make programmers happy, and Rails lists optimising for programmer happiness as the first item in its doctrine. I think agent tooling needs the same kind of priority.

So that's what I'm optimising for in the dev tool I've been building these past weeks. I want it to be a joy to use.

Tasks, not chat

That means moving chat out of the centre. Instead, tasks become the unit the work revolves around.

When an agent needs your input, it asks a clear question with options rather than pulling you into a conversation. Something like: should the export include archived projects? Yes, no, or only when the user asks for them. You pick one, and the agent carries on.

You act as the team lead and make the calls. The agents do the rest.

I'm calling it Skelcrew. Sign up for the mailing list and I'll email you when it's ready to try.

All writing