The Tool That Works Is the One You'll Actually Update

We've tried GitHub Issues and Projects multiple times. It never sticks.

We always start with good intentions — organized labels, milestones, proper status columns. It looks great for about a week. Then it quietly dies. What actually works? A wiki page. Messy, flat, full of half-baked ideas and checkboxes that get rearranged constantly. It shouldn't work. But it does.

The problem was never organization. It was friction. Updating a GitHub issue feels like a commitment. You're filing something. Giving it a title. Assigning it. There's a ceremony to it. Updating a wiki feels like thinking out loud — you just move things around and keep going.

As AI accelerates everything, the ability to change direction quickly matters more than the ability to track things formally. A structured system that's 3 days stale is worse than a chaotic list that's current. We'd rather have an accurate mess than an organized fiction.

Here's the interesting part though: AI agents are great at GitHub Issues. Give an agent a well-scoped issue and it executes. Maybe that's the tell — it was never designed for the organized chaos of human thinking. It was designed for the structured, repeatable, automatable.

For us, the wiki is where ideas live. GitHub is where agents work.

"The best productivity tool is the one with the least resistance between thought and action."