Jorge Rondón

How I work

My day runs through several Claude Code sessions at once: one coordinates, the others work, each in its own lane. Most of these rules exist because something went wrong once and I didn't want it to happen twice. Now I'm measuring my own setup against a baseline before I change it.

1. 01

   One by one

   One task per lane, one lane per job. Fewer lanes beat a longer queue.

2. 02

   Brief before GO

   No session writes anything until I've read its brief and said GO. A GO for one task doesn't carry over to the next.

3. 03

   Evidence, not words

   Nothing closes on an agent's say-so. A status needs proof: the pull request, the live page, the read-back.

4. 04

   A rule that doesn't fire becomes a hook

   If a mistake comes back with a written rule in place, the rule isn't there when it matters. So it becomes a check that runs on its own.

5. 05

   The agent prepares, I run production

   Anything that touches production is a command I run myself, and each one checks the current state before it changes anything.

6. 06

   Every task teaches me something

   Each brief says what I'll learn, not only what will change.