I don't have the working memory for this anymore, so my issues do

I have a full-time job and two kids under six. Most of my coding on Retic, a goal and habit tracker, happens through AI agents (mostly Claude Code) after both kids are finally asleep, in whatever's left of the night before I fall asleep too. Some nights that's ninety minutes. Some nights it's fifteen.

This isn't a Retic problem. Anyone whose coding sessions get interrupted for days at a stretch runs into the same thing. For me it was never code quality. It was picking a task back up two or three nights later and having no idea what I'd actually decided last time, or what the agent had already tried and dropped.

So I stopped trying to hold any of it in my head. Any real task starts as a GitHub issue before I write a single prompt. I write the goal and the context first, then force myself to spell out what done actually looks like, while I still remember why I'm doing it in the first place. The agent works inside a branch and PR tied to that issue number, and merging it closes the issue automatically. Two nights later, or two weeks later, the issue is the memory. Not me.

The part I didn't expect: this ended up mattering more for the agent than for me. Handed a loose prompt with no written context, it defaults to whatever's simplest, a boolean where I actually needed a small state machine, and I don't catch that until I'm rebuilding it. The issue is where that gets caught, before it's code.

A few months of these issues now read back like the project's memory across every gap I couldn't help leaving in it.

Anyone else building in stolen pockets of time, how do you carry context across the gaps?