Built for Pursuit · AI-Native Program Manager

Builders don't need answers.
They need better questions — and proof they're getting sharper.

This is a working prototype, not a deck. It fuses two of my projects: Animus, a Socratic AI coach that builds metacognition instead of dependence, and SkillTree, a BCG:X hackathon winner that replaced the resume with evidenced skills. Together they answer the question your program lives on: how does a Builder without a degree prove what they can actually do?

Enter the prototypeSee the operator view

01

The coach refuses to answer

Gnothi asks one question at a time, live through Lovable AI. It reflects your thinking back and pushes. Handing over the solution would be the failure mode.

02

Your words become the evidence

Every time you demonstrate prompting, debugging, systems thinking or persistence, the model records it as a tool call, quotes you, and the skill tree grows. Nothing is self-reported.

03

The operator sees it first

The same evidence, plus attendance and reflection cadence, ranks who's slipping — and drafts a specific, guilt-free re-engagement message in a click.

What I'd do in week one

  • Instrument persistence. Attendance alone tells you someone left, not that they were about to. Reflection cadence and evidence velocity tell you a week earlier.
  • Make outreach cheap. The reason at-risk Builders get a generic check-in is that a specific one takes 20 minutes. This makes it take 30 seconds and stay specific.
  • Model the mindset live. This whole prototype is the workshop: here's a problem, here's the system I built for it, here's where the AI got it wrong.
  • Give Builders artifacts. An evidence map with quoted proof is something a Builder can walk into an interview with. A completion certificate isn't.

Cohort data is fictional. The coaching conversation and the drafted interventions are real AI calls, made server-side.