Work with us —
by application, not purchase.
Each cohort takes a small number of students. Not a subscription — a hands-on pilot where we build a real, verified teaching loop around one learner and prove it works.
01 · What the pilot includes
Personal AI learning agent
Diagnoses errors, chooses the next teaching move, verifies mastery, and updates the student model — every day.
A personalized learning space
A dedicated instance of the product tuned to one learner's syllabus, goals and pace — not a hand-built one-off site.
Human quality calibration
Real people handle cold-start, weekly calibration and exceptions — not daily tutoring. The AI makes the day-to-day teaching calls.
02 · Who does what
| Stage | Human | AI |
|---|---|---|
| Cold start | Confirm syllabus, build the initial student profile, seed 1–3 strategies, set quality bounds | Generate the first candidates, set initial weights |
| Daily | Handle only low-confidence calls, errors and exceptions | Generate, rank, deliver, verify, update |
| Weekly | Review quality & method effect; add, cut or revise strategies | Summarize transfer, error recurrence, strategy performance |
| Cycle end | Interview learner & family; judge renewal and product value | Produce the learning-evidence & strategy report |
03 · A four-week cycle
Week 1
Cold start
Profile, syllabus, strategy seeds, quality bounds.
Weeks 2–3
Teach & verify
Daily loop runs; weekly calibration adjusts strategies.
Week 4
Evidence
Delayed retrieval & transfer checks; results compiled.
Review
Decide together
Interview, learning report, and what's next.
04 · The three starting strategies
The minimal set that covers find the problem → help understand → verify & consolidate.
Diagnosis first
Surface the error type before deciding whether to explain or backfill a prerequisite.
Scaffolded explanation
Tune language, length, analogy, examples and hint strength — but always return to the formal goal to verify.
Targeted practice & retrieval
Generate variants for the gap, fade hints over time, schedule delayed review.
05 · What we won't claim
We measure learning by unprompted transfer that's still remembered later — not watch time or completion.
Knowledge is strong but not infallible — we keep reliable sources and verification in the loop.
This is an early pilot. Every figure shown is illustrative until your own data replaces it.
06 · Apply
Tell us about the learner.
We review each application and reply within a few days. If it's a fit, we schedule an interview and a needs assessment before any commitment.
Thank you.
Your application is in. We'll be in touch within a few days about next steps.