← All notes
Jul 2026

What a desktop robot taught me about scope

A small hardware project turned into a lesson about restraint. The best version of Stackchan does one thing, not ten.

[ AGENT PHOTO ]

I wanted Stackchan to do everything — talk to me, remind me of things, control other stuff on my desk, maybe read my calendar out loud in the mornings. Within about two weeks I’d built a small creature that did five things badly instead of one thing well.

This is an old lesson wearing a new costume. I’ve watched product teams do exactly this with enterprise agents — try to make the first version handle every edge case instead of nailing the core interaction — and I still fell for it on a project with no stakeholders except me and no deadline except my own patience.

What I cut

The reminders went. The desk-control stuff went. What was left was the one thing that actually made the project interesting in the first place: a consistent, opinionated personality you could have a short conversation with. Everything else was scope creep wearing the disguise of “while I’m at it.”

"“While I’m at it” is where most good ideas quietly turn into bad ones."

Cutting scope on a side project with no one watching is a strange kind of practice. There’s no roadmap pressure forcing the discipline — you have to supply it yourself, which turns out to be the actual skill.

[ OPTIONAL DIAGRAM ]
The nugget

If you can’t describe what your thing does in one sentence without an “and,” it isn’t ready to be judged yet. Cut until it is.

Next noteAll notes →
Product management is applied philosophyJun 2026
Greg O'Brien2026