AceBuddy

The problem

AceBuddy turns a syllabus into AI-generated micro-sessions and Ace Cards. The MVP worked for one exam at a time. To grow, it needed to handle several exams per learner without data leaking between them, sell study stacks to people who weren’t signed up yet, and let contributors publish and get paid.

Key decisions

• Each exam is its own space

  Preferences, progress, Ace Cards, analytics, notes and the AI copilot all follow the exam you’re in. Adding a second exam never leaks another exam’s sessions, dates or subjects, and the cache is keyed the same way, so switching exams never flashes stale data.

• Onboarding that ends in a real plan

  Standardized-exam learners pick subjects and a date; university students upload a syllabus or build one from their courses. Generation progress streams live, and learners land on a usable plan instead of an empty dashboard.

• No blank screens while AI works

  Every session moves through queued, enriching and enriched, with retries and detection of poor results. The dashboard waits until content and video exist, so nobody stares at an empty lesson after signing up.

• Re-enriching sessions when the AI gets better

  On the backend, I built session re-enrichment. Whenever we improved the agent’s schema for breaking a syllabus into sessions, the sessions learners already had stayed on the old structure. Re-enrichment finds outdated and skeleton sessions and runs them through the updated agent, so existing learners get the improvements too, not just new sign-ups.

• A marketplace that works logged out

  Guests can browse, see prices in their own currency, check out and pay with Stripe. Coupons go from 1 to 100%, and a fully discounted order completes without a payment redirect while still granting the stack.

• Contributors earn per exam

  Contributors apply for a specific exam, backed by their study history and scorecard, can reapply after a rejection with the reason shown, publish stacks, set coupons and get paid through Stripe Connect. Approval for one exam never unlocks another.

• Shipping in slices

  Feature flags for the marketplace, collections and contributor tools let staging ship product slices without waiting on one big release, and analytics run safely in local development.

Outcome

• Took AceBuddy from MVP to its first and subsequent live versions, across the frontend and backend.
• Existing learners’ sessions now pick up every improvement to the AI agent through re-enrichment.
• Turned a single-exam study app into a multi-exam platform with learner, marketplace and contributor sides.
• Guests can buy study stacks, and contributors can publish and get paid, without the team touching anything by hand.