Implemented
The site has flagship ML pages, Atlas paths, source-linked evidence, diagnostics, signed-in learner state, saved topics, review entry points, privacy pages, and an iOS companion contract.
Launch packet
Private beta launch packet for the ML theory learner loop.
This page turns the current product truth into a launch plan: what a serious learner can try, what the demo may claim, what is still experimental, and what must pass before paid growth.
Last reviewed: July 5, 2026
Product truth
What is real now, and what is not.
Experimental
The learner loop uses FSRS-style review state, Q-matrix skill links, and guarded PFA-style mastery updates. Those signals guide study actions; they are not proof of learning lift.
Data-insufficient
Live production rows are still too sparse for calibrated IRT, validated PFA personalization, or paid-acquisition claims about retention improvement.
Functional funnel
The beta should prove that a learner can find a useful next action without a guided sales call.
1First proof of valueOpen a flagship ML topic with prerequisites, examples, failure modes, references, and follow-on topics.2Map the next stepUse Atlas to move from one concept to neighboring prerequisites and downstream topics.3Create learner stateSign in, save topics, start review, and keep the account-deletion path visible.4Continue on iOSUse the iOS companion for Today, Atlas, Saved, and Profile once TestFlight upload and signed-in smoke pass.
Beta cohort fit
The first beta group should test the learner loop, not just browse the library.
The ask is narrow: one week, one real study goal, and blunt feedback about the next action. That keeps the beta useful before paid acquisition or broad marketing.
Good fit
Graduate ML applicants, research engineers, self-studying ML practitioners, and math/statistics learners who can spend one focused week testing the loop.
Not the right fit yet
Learners who need a fully guided beginner course, live tutoring, app-store polish, or verified learning-effectiveness claims before trying the product.
One-week feedback contract
Use one topic path, save at least one item, attempt review, and report whether the next action is clear, where the curriculum loses you, and what would make you return.
Marketability
Demo media comes after the product path is repeatable.
The strongest video asset is not a static explainer. It is a real run through the app: graph, topic, evidence, signed-in state, and mobile continuity. Generated presenter footage is useful only as a finishing layer after real captures pass.
Readiness source: data/content/launch-media-readiness.json. Policies: real-product-footage-first and paid-acquisition-waits-for-retention.
Public web still set
readyUse these stills for static launch materials only; signed-in web, iOS continuity, and video footage remain separate gates.
Publishable
Five-minute founder walkthrough
draftRecord one current walkthrough from the production or production-equivalent app, then review for private data and stale UI.
Not publishable yet
60-90 second product demo
draftCut the longer walkthrough to a short route order: launch, Atlas, flagship ML topic, Evidence, signed-in learner state, and iOS handoff.
Not publishable yet
30 second ad creative
blockedWait for activation and week-one return evidence before running paid acquisition.
Not publishable yet
15 second vertical hook
blockedMake this only after the longer cut is approved, so the hook does not outrun the product truth.
Not publishable yet
Presenter-led pitch video
blockedUse a presenter video layer only after the script and real app captures pass review.
Not publishable yet
Recording sequence
The short demo has to follow real routes.
Source: data/content/launch-demo-script.json. Every step points at a receipt and keeps raw emails, bearer tokens, and private notes out of frame.
- 1Open launch packet/launchTheoremPath is a graph-backed ML theory learning system for serious learners.
- 2Show one path/atlas?trace=scaling-laws&fresh=launch-demoAtlas shows where a topic comes from and what it leads to.
- 3Show flagship content/topics/scaling-lawsA page should teach the idea, test the learner, and connect the next step.
- 4State the evidence boundary/evidenceThe learner model is wired, but learning lift and calibrated ability are not calibrated yet.
- 5Show iOS Todayios://todayThe iOS app continues the same study loop.
- 6Show iOS Savedios://savedSaved state and review are the habit surface, not a separate demo app.
Public web stills
The screenshot set now comes from real captured routes.
These are optimized production-style WebP captures recorded in the launch receipt. They are safe for static launch materials, but they do not prove signed-in state, iOS continuity, or learning outcomes.
Static captures are the public-web layer only. Signed-in web, account deletion, iOS continuity, and video footage remain separate release gates.
Monetization boundary
The site can invite beta users now. It should not sell outcomes before the learner metrics support them.
Free public library
Flagship ML pages, Atlas paths, and evidence surfaces remain useful without payment.
Future paid layer
Candidate paid value is advanced review, saved trail history, research interview prep, and private progress reports.
Paid growth gate
Ads wait for measured activation, day-7 retention, and a repeatable private-beta onboarding path.
Release gates
Checkpoints before a broader push.
- Production web build and critical anonymous flows are green on main.
- Signed-in web walkthrough covers Today, Atlas, Saved, Profile, and deletion.
- iOS archive, App Store Connect validation, privacy labels, and screenshots pass.
- Fresh demo screenshots show real routes without secrets, raw emails, or tokens.
- Evidence and methodology pages state FSRS, Q-matrix, PFA, and data-insufficient boundaries.
- Paid acquisition waits for measured activation and day-7 retention baselines.