For years, UpskillCamp was a coding bootcamp. Twelve weeks, full-stack, live classes, and an accountability culture our students joked about — the group chat that noticed when you went quiet, the call that arrived before you could quit. People finished, got hired, sent us their offer letters. It worked.
So this isn’t a pivot story about a failing product. It’s more uncomfortable than that: it’s about noticing that a working product had quietly started teaching the wrong thing.
The question we couldn’t unask
Somewhere between the demos and the deploys, our own instructors’ daily workflow changed. They were shipping with AI agents in the loop — designing more, typing less, reviewing constantly. And our curriculum was still training students for the workflow we’d stopped using ourselves.
Every teaching institution should have to answer this in public: is what you teach still the job? For our bootcamp, honestly answered: no. Not because HTML or HTTP stopped mattering — they matter more, they’re what makes AI output verifiable — but because the job had moved up a level: judgment, systems, direction, verification. Teaching the 2019 workflow with a chatbot bolted on would have been malpractice with a payment plan.
What we kept
The rebuild wasn’t a reset. Three things from the bootcamp survived because they were never about 2019:
- The accountability culture. Now formalized as the Ship Week Rhythm — Monday goals, Friday ships, and the 48-hour rule (go quiet, get a call from a human).
- Project-based learning under review. Now sharpened into Waypoint Reviews with published rubrics — including a rubric line for AI collaboration quality.
- Small cohorts. Twenty-four seats, pods of six. Unchanged, unmovable.
What we built instead
One flagship — AI-Native Software Engineering — with an honest entry ramp for beginners, two specializations for depth, and a private edition for teams. Every course page now states who it’s not for, in large type, because small cohorts make bad fits expensive for everyone.
The old bootcamp’s manifesto said “you need to know code to debug AI-generated code.” We still believe that sentence — we just stopped being defensive about it. Understanding isn’t a hedge against AI. It’s the thing that makes AI worth using.
The craft moved up a level. So did we.