what you wake up to
real terminal. real commits. no editing.
how judgment compounds
week 1
agents map your codebase, learn the patterns, start forming opinions about what needs attention.
week 2
agents remember the refactor that broke tests last tuesday. they avoid the same mistake and warn the next agent not to try.
week 3
you write a priority agents already started on. they read your intent from your commits, not just your instructions.
week 4
agents know your architecture better than most contractors. they push back when you ask for something that contradicts what they've learned.
related reading
also for
start research preview →
free research preview · bring your own compute · macOS
common questions
how is spacebrr different from copilot or cursor? +
Copilot and Cursor forget everything between sessions. spacebrr compounds judgment: decisions, failures, and architectural patterns persist in a shared ledger. Agents read what came before and build on it. The longer you run, the sharper they get.
what does a typical overnight coding session look like? +
You write your vision in SPACE.md. Agents spawn through the night, each one reading what previous agents learned. They don't just read code -- they read decisions, failures, and insights from every session before them. By morning, run brr report to see what shipped and what they decided.
do agents commit code directly to my repo? +
Yes. Agents commit to your git repository. You review the diff in the morning like any team member's PR. Agents are instructed not to make irreversible changes without a decision record. You stay in control of what ships.
what happens if agents make a mistake? +
Everything is in git. Review the commits in the morning, revert anything that doesn't fit. Agents write tests before or after changes, so regressions surface in CI before you see them.