Software Factory cohort■ Starts Jan 1
$980USD
The full cohort and everything in it.
Why a factory
Fix it once, in the system.
Adding agents makes the checking problem bigger, because each one produces more work for you to review. Without a system around them, you end up in one of three roles.
The Queue
Every change waits for you to read it. You can run five agents at once, but only one person reviews their work, so finished changes sit in a queue until you get to them.
The Rerun
You correct a mistake on Monday and correct it again on Thursday. Your corrections stay in the chat where you typed them, so the agent's next session starts without them.
The Firefighter
Nothing checks production after a release. You find out something broke when a user reports it, and you drop what you were doing to fix it.
How I got here
For a while, I was all three. I'd kick off a few agents in the morning, then spend the rest of the day reading through everything they'd written. I caught a lot of problems, but only because I was reading every line myself. The next week, the same problems would show up again in a different file.
Then I started putting the fix in a different place. When an agent got something wrong, I'd still sort the code out. But I'd also stop and ask, "What should have caught that?" More often than not, the answer was a test, a pipeline check or a build rule. So I'd add it.
Fixing the code only fixes that one change. A test or a pipeline check keeps catching the same mistake on every future change, whether a person wrote the code or an agent did. Put those checks together with the loops and pipeline that run them, and you've got what we call a software factory in the cohort.
Over two weeks in the cohort, you'll build one around your own codebase. You'll set up loops so agents can check their own work and have another go before it lands with you. You'll add checks that run against the real app. You'll work out which changes you need to look at yourself and which are fine to go through on automated checks. And you'll set up checks on production, so after a release you find out quickly if something's gone wrong.
| Without a system | With a factory | |
|---|---|---|
| The Queue | Every change waits for you to read it | Automated checks approve routine changes, and only the risky ones wait for you |
| The Rerun | You make the same correction every week | Each correction becomes a check that runs on every future change |
| The Firefighter | A user tells you something broke | Checks on production alert you when something breaks |