The August version of this idea was “AI 101 for the team, eight to ten modules, async with office hours.” I’d had it in my notes since August 7th. Not a project yet. A thing I kept meaning to turn into a project.
September 12 I sat down and redesigned it properly. Self-paced skill tree, task-gated progression. You learn something, you prove you can do it, you unlock the next node. The question I couldn’t skip was who’s learning first.
Not developers. Developers have resources, they figure it out. The first learner is someone who has already tried ChatGPT. Got one good result once, got one embarrassing result once, and now either uses it carefully or has mostly stopped. Confident in their domain. Hesitant about the tool. The main fear isn’t job loss, it’s looking foolish. Paste the wrong thing in a client doc. Have a manager catch it.
That description changes the whole trunk structure. Every task under 30 minutes, because this learner abandons anything longer. No new installs, no API keys. Start with a small win before the hallucination demo, not after, because opening with “here’s where it fails” doesn’t get you to lesson two. Get the cost question out early, because “will this charge me unexpectedly” is a real anxiety for this person and a footnote won’t address it.
The platform went live at branchandbadge.com branchandbadge.com on September 13. One day from redesign to production. The pilot is five colleagues from work, which is down from my original estimate of five to ten. If I’m committing to grading submissions within 48 hours, five is what I can honestly promise.
Don’t get me wrong, this is still early. The trunk has four modules. Branches haven’t been built. Badge issuance, email notifications, and paid tiers don’t exist yet. The assessment and security pass this morning was mostly closing RLS gaps and sandboxing the AI grader worker. Foundation work.
But the learner persona is load-bearing. Every design decision since September 12 traces back to that “already tried, got burned once” description. You can build a general curriculum or you can build one for a specific fear. I know which one gets finished.