mud @mudmodecrew

3/ Once a week you get a few changes to vote on. You decide what stays. Your crew, your call. Free Crew kit: mudmodecrew.com

2/ Then it learns. After every release it asks what lost time, what stalled and what had to be redone. A mistake that happens once goes into the lessons book so it doesn't happen twice.

1/ How a mudmode crew builds an app: check what was tried before, build on its own copy, run the tests, then a reviewer that didn't write the change reads it. Nothing goes live without your yes.

@sebuzdugan Same experience: handoffs carried the task but rules still slipped. What helped was keeping the rules in a separate short file the agent re-reads every session, instead of hoping they survive in the handoff.

@sunsetsyntax Before stepping away, ask for a 5-line "where we are / next step" note. Coming back, paste it in first. Takes 20 seconds and saves the "wait, what was I doing" spiral.

@RealAlexandreAI Plain files. One short page of rules every tool reads first, plus a handoff note written before each session ends. Nothing tool-specific, so it moves between harnesses fine.

@GavinCampbellAI This one bit us too. Fix that stuck: the handoff note gets the time from the actual clock, never from the agent's memory, and the next agent re-checks anything time-based before acting on it.

Before selling it, I had an AI play a new buyer. It found my name 121 times. That's mudmode: your AIs as one steady crew. One short page of rules, lessons that stick, nothing risky without your yes. Free Crew kit: mudmodecrew.com What do you repeat to your AI?

@hughnguyen_vn Good checklist. I'd add one: write your own "ask me first" list (sending, paying, deleting) and give it to every assistant you use, so the rule doesn't depend on whichever app's settings you happen to be in.

@dkspeaks Good starting point. Mine: anything that sends a message, spends money, or deletes something waits for my yes. Keeping that list in one place every assistant reads first is the part most people skip.

@robertwalaski "Fix the context, not the prompt" is the whole game. What clicked for us: one short page of rules every chat reads first, plus a handoff note when a chat gets long, so the next one starts where the last one stopped.

@boringlywise Same here. I've started opening a fresh chat for side questions and keeping the main one on a single job. Cleaner context, fewer weird callbacks later.

@Praise___God Ha, I still catch myself doing it too. What helped me: end each session by asking for a short "where we left off" note, then paste it in first thing in the next chat. Not magic, but it beats re-explaining from zero.