Alpha School · Deep Dive · Academic Block Anatomy

Two students, one machine:
a day inside Alpha School

A stage-by-stage reconstruction of what an ordinary Alpha School day feels like for a 4th-grader and a 9th-grader — with the moments where the system is making a deliberate design decision called out [SYSTEM] like this. Built from Alpha's own documentation and sample schedule, parent and press reporting, and our synthesis; each claim is badged for evidence quality.

Scope — typical Alpha School offering (not the sports/Founding variants) Students — Maya, G4 (primary→middle) vs Ethan, G9 (middle→high) Evidence doc Alpha-stated · reported press/parents · inferred our synthesis

The two students

Same architecture, two very different textures

Alpha runs one machine across K–12: a two-hour AI-driven academic core in the morning, life-skills workshops in the afternoon, and a motivation economy running all day. What changes with age is autonomy, cycle length, tool stack, and what "motivation" means. Two composite students, built from documented band behaviour, show the contrast.

Student A · Primary → Middle

Maya

Grade 4 · age 9 · band "4–5"

Fluent reader, halfway through her grade-level math sequence — in the same room as a classmate still doing addition and another already into early algebra. [SYSTEM] level ≠ grade: every subject is placed and progressed individually.

Core block runs 2–2.5h in short cycles (~25 min)
Stack: AlphaMath Fluency + Synthesis (math), AlphaRead → TeachTales (reading + knowledge), AlphaWrite (writing), AlphaLearn Language, MobyMax (science)
Motivation = daily XP target + Alphas currency + Friday Emporium + her personalised dream reward
Afternoon: 7-week workshop block ending in a public performance

Student B · Middle → High

Ethan

Grade 9 · age 14 · band "9–12"

Longest autonomous stretches in the school. Runs his academic core against a SAT/college track, has elected two AP subjects with an advisor, and is pitching a real side-project in the afternoons.

Core block still ~2h productive — but surges push real academic time to 3–3.5h on many days
Stack: Math Academy, AlphaRead at SAT-track difficulty, Write of Passage (public writing), Membean, AP-track apps
Motivation shifts from candy-and-stickers to earned autonomy, real audiences, and college signaling
Proctored mastery tests, AP exams (College Board), SAT/PSAT timeline

The day, stage by stage

8:15 AM to 4:00 PM

Times follow Alpha's published sample schedule;³ the texture inside each block is drawn from parent accounts, guide job postings, and Alpha's own descriptions. doc = Alpha-stated, reported = press/parents, inferred = our synthesis from the model. The fourth column — the machine — is what the platform, economy, and guide console are doing while the students work.

TimeMaya · G4Ethan · G9The machine behind it

8:15–8:45
Arrival

Soft open

Drop-off, hang out, settle. Free social time before the ritual starts. reported

Soft open

Same, with more self-managed startup — older students often open Dash and check their day plan before Launch. inferred

Platform

[SYSTEM] Dash has already assembled each student's day: year goals decomposed into today's weekly/daily chunks, sequenced across apps — no "what do I do now" decision cost. doc⁴

8:45–9:00
~15 min

Morning Launch

Guide-led ritual: energy, connection games, goal-priming for the day. Explicitly not instruction. doc

Morning Launch

Same ritual, harder-edged content: weekly goals, AP/SAT milestones, shout-outs for shipped work. reported

Human OS

[SYSTEM] Motivation is a designed daily ritual, not an afterthought — guides are hired and trained as motivators, not lecturers (unbundling: apps teach, humans energise). doc²⁴

9:00–11:30
Core block

Academic core · cycles 1–5

Opens Dash, works through today's plan in ~25-min focus cycles with Q Breaks between. Math first (fluency sprint, then concepts), then reading, writing, science. Same-age kids at totally different levels around her. No homework ever. doc — minute-level zoom below.

Academic core · long blocks

Same skeleton, longer autonomous stretches: 30–40-min blocks of Math Academy, SAT-track reading, Write of Passage drafting. Self-books a proctored mastery test when ready. reported — zoom below.

Learning OS + economy

[SYSTEM] Progression is mastery-gated (≥90%), not time-gated; missed questions resurface later (spaced repetition); "Test Out" lets anyone skip ahead; placement is per-subject⁸. [SYSTEM] XP = productive minutes, judged by an engagement model — fast-scrolling/skipping doesn't count; daily target 120 XP⁷¹⁴. doc reported

11:30–12:30
Lunch

Lunch & recess

Staggered by age. Real play — movement is part of the design, not a concession. doc

Lunch & recess

Same; older students use part of it for venture work, club calls, or advisor drop-ins. reported

—

Time-saved scoreboard keeps framing the bargain: [SYSTEM] "earn your afternoon" — the ~2h academic cap is sold as a deal students close every morning. doc⁶

12:30–3:15
~2 × 1h

Workshops

This block, her group is mid-way through a 7-week comedy/improv workshop: they write, direct, and perform a full sketch show for a 100+ live audience as the session's AlphaTest. Recent band-4–5 menu also includes go-karts, food-truck businesses, and "rejection training" (a 4th-grader cold-emailed 40 ornithology experts and got 6 replies). reported¹⁷

Workshops → AlphaX

Venture mode: his team runs a small real business/app/podcast with real customers and pitches to real audiences. Older bands also run expedition-style finals (5th–7th graders sailed Florida → Bahamas as a test). doc reported¹⁷

Life-skills product

[SYSTEM] Workshops run in ~7-week themed sessions (~4–5/year; one session shipped 18 workshops across 5 age groups), each ending in a public performance gate — real audiences as the motivation mechanism. [SYSTEM] Afternoons are the retention/love lever — parents' favourite product. doc

3:15–3:30
1:1s

Limitless Meeting

Books her own 1:1 with a guide from inside the platform when she needs help unblocking or checking a goal. [SYSTEM] self-advocacy is engineered: kids learn to request help, schedule, and prepare — a skill with its own expectations. doc²

Advisor + Guide 1:1s

Academic advisor time (AP load, SAT plan, college path) plus guide check-ins. He runs this cadence himself. reported

Coachbot console

[SYSTEM] guides work from alerts, not vibes: the console flags students below ~1.5× projected growth, logs interventions, tracks check-chart checklists. reported¹⁵

3:30–4:00
Closing

Town hall & cash-out

Peer shout-outs, student-led town hall. Friday closing is when the week's Alphas convert to Emporium shopping — and her personalised dream reward is checked off against completed work. doc⁷

Town hall & cash-out

Same, with venture demos. Top performers can enter the Rocket Ship autonomy tier — earned off-campus/independent learning privileges. reported⁶

Motivation economy

[SYSTEM] weekly settlement, not daily — delayed gratification is part of the curriculum; class-wide bonuses make peers cooperative, not just competitive. [SYSTEM] every child has one personally meaningful grand prize contingent on the work. doc

4:00
Dismissal

Home, empty-handed

No homework. Ever. [SYSTEM] total-time-on-learning is treated as the scarce resource to protect — evenings are for family, play, and sleep. doc

Home — or venture hours

No homework; some students voluntarily continue venture work or AP prep (which the system treats as bonus, not obligation). reported

Parent surface

Parents get the Activity Feed, weekly End-of-Week Report, and periodic Academic Snapshot (RIT scores, national percentile, US-median vs Alpha-median bars, grade-mastery colouring). doc⁸¹⁶

The human layer · a guide's day

The guide: the school's real teacher

Alpha unbundled the teacher: software delivers instruction, and the adults are redeployed as full-time motivation and coaching specialists — and yes, Alpha runs unusually high adult density. The job postings are explicit: "You are not a teacher."¹⁸ What follows is the guide's day, stage by stage, from Alpha's own hiring material and parent accounts.

A guide's day · K–3 cohort bandhalf coaching · half workshops · measured on three numbers

~7:45 – 8:15

Coachbot

Pre-day data review: yesterday's app performance per student, who is off pace, whose weekly goal is at risk. Then chooses the right motivational approach per child — a leaderboard challenge, a school-currency play, or a 1:1 about their potential.

[SYSTEM] motivation is prescribed per-child from live data — the guide's instinct is a managed input, not the whole method¹⁹

8:15 – 8:45

Soft open

Relationship time at arrival — the raw material for everything later. Guides are evaluated on students actually liking them (see "three numbers" below).

8:45 – 9:00

Launch

Runs the motivational ritual and sets each student's daily objectives from the morning data review. No instruction.

[SYSTEM] Launch = energy + data-driven goal-setting, explicitly not teaching¹⁹

9:00 – 11:30

1:1 + small groups

The core block is the guide's coaching floor: circulating while students work in apps, pushing each to 100% of weekly app goals, intervening the moment Coachbot shows someone off pace — before a weekly goal slips. When a child is stuck, the guide routes them to resources rather than explaining content ("did you use your resources?").⁷

[SYSTEM] "if students are working independently and you're standing still, something has gone wrong" — the KPI is intervention, not supervision¹⁸ [SYSTEM] content answers are forbidden: learning-how-to-learn is the product

10:15 / lunch

Q Break

Breaks and lunch double as informal coaching and relationship time (the "bird-watching with the guide" reward pattern from Price's own telling).

⁷

12:30 – 3:15

Workshops

Leads hour-long life-skills workshops — public speaking, focus, giving/receiving feedback — loud and on their feet, from a supplied playbook they personalise (not write). Then runs Test2Pass mastery checks: the child must demonstrate the skill to pass; participation isn't the bar; a 99% is coached to 100%.

[SYSTEM] life skills use the same mastery gate as academics — prove it, or keep coaching¹⁸¹⁹

3:15 – 3:30

Limitless Meetings

Student-booked 1:1s — unblocking, goal review, check-chart sign-offs.

3:30 – 4:00

Closing

Shout-outs and town hall; Friday includes the weekly data readout and Emporium.

Weekly / quarterly

Accountability

A guide's quarter comes down to three numbers: every student hits their weekly app goals, every student passes Test2Pass for each workshop, and ≥90% of students say they love you. "Missing one of the three is missing the job." Advancement: Guide → Lead Guide (mentors a guide team while keeping a cohort).

[SYSTEM] the performance contract encodes the school's values — outcomes AND relationship are measured, with affection as a hard metric¹⁸

ACCOUNTABLE TO

weekly cohort data · "comfortable being measured"

Who the adults are

Role · front line

Learning Guide

$100k base (some campuses $120k) · on-site, full-time

Leads one cohort band (K–1 or 2–3), ~15 students¹⁸
No teaching certificate required — any Bachelor's + ≥3 years with children (coaching, tutoring, camps, children's theatre all count)
Hiring bar: "someone kids listen to and want to impress… genuinely warm and completely immovable on expectations"
Final interview round: a full day on campus coaching real students¹⁸

Roles · around the cohort

Specialists & leads

the unbundled staff room

Lead Guide — mentors a team of guides while keeping their own cohort (career step, "impacts hundreds")¹⁹
Reading Specialist — owns literacy development alongside the guide¹⁹
Campus Lead / Head Guide — parent communications and campus logistics (guides are explicitly not on routine parent email); Head Guide pay ~$150k reported¹⁰
Reported ladder: Associate Guide ~$60k → Guide ~$100k → Head Guide ~$150k — 2–3× typical Austin teacher pay¹⁰

Do they really have more adults per student? Roughly yes — with the unbundling caveat. Parent-reported ratios: ~5:1 at private campuses, ~20:1 at charter-style operations¹⁰; job postings describe cohorts of ~15 per guide¹⁸. For comparison, US averages are 15.4 pupils per teacher in public schools and 12.1 in private (NCES, fall 2022)²⁰. But "guide" ≠ "teacher": no instruction, no grading, no curriculum writing — the adult density buys motivation, coaching, and emotional support at ~2–3× the normal staffing ratio, plus specialists (reading) and ops (campus lead) carved out of the traditional teacher's job description. About a third hold education degrees¹³. For a replica: the ratio and role design transfer; the no-credential mix does not survive Indian norms (B.Ed/D.El.Ed required) — see the replication report §3.

Zoom · 9:00–11:30

Inside the academic core

The core block is the product's heart: roughly two productive hours, assembled per-student each morning. Below, a reconstructed cycle-level view for both students — the tools they touch, and the design decisions embedded in each cycle.

Maya · G4 · core block~2h productive · 25-min cycles · target 120 XP

9:00 – 9:05

Dash

Opens the day plan: today's chunks of her year goals, ordered by the sequencer. Picks the first subject (or accepts the default).

[SYSTEM] executive-function scaffolding: the plan is pre-made; the only decision is "start"

9:05 – 9:25

AlphaMath FluencySynthesis

8-minute math-facts sprint (speed + accuracy drills), then concept work with interactive widgets — sometimes in a small team challenge.

[SYSTEM] fluency before concepts — Alpha's data claims fluent kids learn new math 2.5× vs 0.8× for non-fluent¹⁴

9:25 – 9:30

Q Break

Movement break. Outside, snacks, running around.

[SYSTEM] attention-cycle design: ~25-min focus + movement reset, protecting "good screen time"

9:30 – 9:55

AlphaRead

An article generated at her level ("~10% above current capability"), with guiding prompts and comprehension questions. History/science/civics facts ride inside these articles; a free-reading stretch may switch to a personalised TeachTales story.

[SYSTEM] the knowledge spine is delivered inside reading — ~5,000 Core Knowledge facts embedded in content, not a separate subject

9:55 – 10:15

AlphaWrite

Sentence → paragraph progression (The Writing Revolution pattern) with instant feedback on each attempt.

[SYSTEM] "10× the practice and feedback loop"; deliberately anti-ChatGPT — writing is treated as thinking practice

10:15 – 10:30

Snack · Q Break

Official sample-schedule snack window; outdoors if possible.

10:30 – 10:55

MobyMaxAlphaLearn Language

Science module (adaptive practice), or language work — dash routes her to whichever subject is furthest behind her plan this week.

[SYSTEM] the sequencer compensates weak subjects; ≥90% mastery gate, misses resurface later (spaced repetition)

10:55 – 11:20

Resurface queueTest Out

Spaced-review queue (yesterday's misses) — or, if she's hot today, a placement test to Test Out of a level.

[SYSTEM] anti-speed-run gate: lesson credit needs ≥80% on problem sets; 100% mastery tests pay bonus Alphas

11:20 – 11:30

Check chart

Checks off non-software goals (e.g., physical books read from her list) with a guide — these gate grade moves alongside software mastery.

[SYSTEM] deliberately hybrid: screens measure mastery; check charts keep real books and off-screen skills in the loop

XP TODAY

~100/120 · each XP = 1 productive minute

Ethan · G9 · core block~2h productive · 30–40-min blocks · surges to 3h+

9:00 – 9:05

Dash

Day plan + AP elective load check. He may move blocks around himself within the day's rules.

[SYSTEM] autonomy scales with demonstrated self-management

9:05 – 9:40

Math Academy

Rigorous adaptive math: worked examples, mastery sets, spaced cumulative review — pacing far beyond grade level for many students.

[SYSTEM] deliberately hard content: the engine sequences mastery + spaced review rather than "adaptive comfort"

9:40 – 10:10

AlphaRead

SAT-track comprehension work (goal: SAT-800 English) — or the classic book on his check chart, with reading-guide questions.

[SYSTEM] difficulty ramps toward a public, high-stakes target (SAT) so effort maps to visible future value

10:10 – 10:30

Break

Longer self-managed break; often social or venture time.

10:30 – 11:05

Write of Passage

Real writing for real audiences (essays, internet writing, pitch decks) with tight feedback loops.

[SYSTEM] authentic audience as the incentive — motivation migrates from points to reputation

11:05 – 11:25

MembeanAP-track app

Vocabulary to ~4,000 words and/or elected AP subject work (science/history via adaptive coursework).

[SYSTEM] choice architecture: AP elections each fall with an advisor — student owns the trade-off

11:25 – 11:30

Proctored test slot

If a mastery test is due, he books a video-proctored session and tests when ready — pass at ≥90% or the content resurfaces.

[SYSTEM] integrity layer without teachers: proctoring + mastery thresholds replace seat-time trust¹⁰

XP TODAY

on track · surge days stretch the core past 120

Honest footnote on "2 hours". Two productive hours is the marketing average; the real academic block runs 9:00–12:00/12:30¹⁰ and reaches 3–3.5h on many days and during "surge" periods where students bank platform hours. reported Critics also note the growth claims need care: the famous "2.6×" figure¹¹ was methodologically broken as published (an averaged standard-error column)¹²; recomputations of Alpha's own summaries land nearer 1.3–1.7× — still substantial, but treat all growth numbers as claims, not audits.

Across ages

How the machine bends with age

One architecture, five textures. This is the answer to "what do we know about different age groups" — the knobs that move are cycle length, self-direction, tool stack, workshop style, and what the economy pays out for.

Band	Core block texture	Stack signature	Afternoon workshops	Motivation economy	Autonomy
K–1	Shortest cycles, most movement breaks, heavy fluency drills (math facts, phonics), most guide hands-on reported	AlphaMath Fluency · Mentava/Lalilo/ClearFluency · Synthesis	Play, swimming, bike-riding; Polite Society — introductions, eye contact, ordering at a restaurant alone as the AlphaTest reported	Immediate, concrete, frequent rewards; personalised prizes front-and-centre	Low — adults scaffold most transitions
2–3	Full Dash self-direction switches on; reading pipeline (AlphaRead) starts at ~3; AlphaWrite begins doc	+ AlphaRead early · AlphaWrite entry	Sports, making, speaking basics, teamwork challenges	Daily XP target + streaks begin to matter	Medium — self-start with adult proximity
4–5 MAYA	2–2.5h core; writing to paragraphs/short essays; knowledge spine visible in reading; self-advocacy skills (book own 1:1s, proctored tests) doc	+ Math Academy at 4 · MobyMax science · AlphaLearn Language	Alpha Afternoon Live (7 weeks improv → full sketch show, 100+ audience); go-karts; food-truck businesses; rejection training reported	Full currency stack: XP, Alphas, class bonuses, Friday Emporium, dream-reward check-offs	Medium-high — plans the day with light scaffolding
6–8	Vocab/grammar layer added; science via Khan; surges to bank platform hours; peer comparison pressure builds reported	+ Membean · eGUMPP · Khan (science)	Sailing expeditions (5th–7th graders sailed Florida → Bahamas as a final), Navy-SEAL-style challenges, programming toy cars, "declined card" disappointment drill reported	Streaks + leaderboard pressure rise; workshop access becomes a privilege lever	High — runs own schedule, books own meetings
9–12 ETHAN	AP elections with advisor each fall; SAT goals; Write of Passage; longest autonomous stretches; some run real businesses during "school" doc	Math Academy · AlphaRead (SAT-track) · Write of Passage · AP-track apps	AlphaX ventures — businesses, apps, podcasts, real pitches; coding, content creation, leadership	Points recede; reputation, real audiences, earned autonomy (Rocket Ship tier), and college signaling carry the pull	Highest — functionally self-managed scholars

What's actually load-bearing across every band: (1) mastery gates replace seat time; (2) motivation is engineered daily — rituals, currency, personalisation — not delegated to content quality; (3) guides are a motivation layer, not an instruction layer; (4) afternoons are a real product with public performance gates. The 2-hour core is the headline, but the economy and the workshops are why kids want to come back.

Appendix A

The tools and the role each plays

Alpha's stack is deliberately assembled from bought specialist apps orchestrated by a built platform. Filter by role to see who does what in the day.

Tool	Role	What it does in the day	Built / bought	Evidence
TimeBack / Dash	Orchestration	The Learning OS: day-plan assembly ("Dash"), subject routing, mastery/test-out flows, XP + time-saved displays, proctoring bookings, parent + guide dashboards, cross-app sequencing (reported to use 1EdTech standards: OneRoster, LTI, QTI, Caliper)	Built ($100M+ claim)	doc
NWEA MAP Growth/Screener	Assessment	Adaptive benchmark 3×/year (fall/winter/spring) + admission placement; anchors per-subject starting levels; feeds parent Academic Snapshots	Bought	doc
AlphaMath Fluency	Instruction	K–3 math-facts automatisation — the daily "sprint" that makes later math faster	Built	doc
Synthesis	Instruction	K–3 concept math via interactive widgets + small-team challenges	Bought	reported
Math Academy	Instruction	G4–12 rigorous adaptive math: worked examples, mastery sets, spaced review (replaced IXL after mid-2025)	Bought	reported
Mentava / Lalilo / ClearFluency	Instruction	K–3 foundational decoding and fluency	Bought	reported
AlphaRead	Instruction	The flagship reading engine: AI-generated articles calibrated ~10% above level, embedding the Core Knowledge spine⁴ (history/science/civics inside reading); continues in HS toward SAT-800 difficulty	Built	doc
TeachTales	Instruction	Generative "choose-your-own-adventure" long stories personalised with the child's friends/interests — engagement-reading volume machine	Built	reported
AlphaWrite	Instruction	Writing mastery progression sentence → paragraph → essay (Writing Revolution model) with instant feedback (autonomous tutor reportedly built by Eigen)	Built	reported
Write of Passage (adapted)	Instruction	HS writing for real audiences — essays, internet writing, venture documents	Adapted	reported
AlphaLearn Language	Instruction	Foreign-language/vocabulary work, G3–12	Built	reported
Membean · eGUMPP	Instruction	Vocabulary to ~4,000 words (G6–12) and grammar (G6–8)	Bought	doc
MobyMax · Khan Academy	Instruction	Science: MobyMax G3–5, Khan G6–8; pre-HS science is deliberately thin ("a grade level in ≤20 hours")	Bought	doc
AP-track apps	Instruction	HS students elect AP subjects; adaptive coursework assigned per elective; school registers/pays College Board exams	Bought	doc
In-app AI tutor layer	Instruction	Explanations, hints, scaffolding, question generation inside each app — deliberately not an open chatbot (anti-cheat). 2025 static human-checked content → 2026 dynamic generation	Built on LLM APIs	reported
XP / Alphas / Emporium	Motivation	Productive-minute XP (120/day target), ~10¢ currencies for milestones, class bonuses, Friday store with real goods (up to iPads/Switches), per-child dream reward, time-saved scoreboard	Built	doc
Coachbot (guide console) + check charts	Human layer	Real-time class view, <1.5× projected-growth alerts, intervention logs, 1:1 booking ("Limitless Meetings"), check-chart checklist tool for non-software goals	Built + ops playbook	reported
Workshops + AlphaX	Human layer	7-week themed sessions with public AlphaTest gates; HS venture platform (real businesses/apps/podcasts); expert-taught intensives	Built + outsourced experts	doc
Proctored mastery tests	Assessment	Video-proctored testing at ≥90% mastery thresholds; replaces exams/seat-time trust; placement "Test Out" available any time	Built (booking) + bought proctoring	doc

Appendix B

Replicating the stack: what's easy, what to build on, what to rent

Verdicts for a lean build (our context: solo + AI-assisted engineering, India-first, CBSE/NEP localisation). "Weeks" = focused engineering time, content effort excluded. The system's real IP — the motivation architecture and orchestration — is the cheap part to rebuild; the expensive parts are content, item banks, and proof.

Piece	Verdict	How to build it	Effort
Dash · day-plan + goal decomposition	Build now	Plain web app: year goals → weekly/daily chunks, subject routing, simple sequencer. No ML needed for v1 — rules + schedules.	2–4 weeks
Motivation economy	Build now	XP ledger + currency + streaks + class bonuses + store + per-child dream-reward tracking. Pure CRUD + rules engine — and it's the piece that makes everything else work.	3–6 weeks
Guide console + alerts	Build now	Real-time class view, growth-threshold flags (<1.5× projected), intervention logging, 1:1 booking. Dashboards + a config table.	3–6 weeks
Parent reports	Build now	Weekly report + periodic snapshot templates fed by platform data; scheduled jobs.	1–3 weeks
Fluency trainer (math facts / typing)	Build now	Micro-drill web app + spaced scheduling (FSRS/SM-2 open algorithms). Alpha's own strongest pedagogical claim sits here.	2–4 weeks
Evidence layer (honest growth math)	Build now	Pre-registered growth methodology + published intake/attrition — a deliberate differentiator vs Alpha's broken 2.6× claim.	2–4 weeks
Adaptive practice engine	Build on OSS	OpenMAIC (MIT, THU-MAIC) for courseware/experience generation; FSRS open-scheduler for spaced repetition; open item banks + IRT libs (catsim/girth) when we outgrow rented assessment. Or wrap an Indian engine (Mindspark/iPrep PAL, est. ₹2–8k/student/yr) for v1.	6–12 weeks
Reading pipeline (levelled gen + comprehension)	Build on OSS	LLM generation + readability banding (textstat-class tools) + our Jev/Passage cognitive-load lint (already adapted + egress-locked) + comprehension-question generation. Content from open corpora: NCERT/OER/Pratham/CK-12 — not Core Knowledge text (copyrighted).	8–14 weeks
Writing tutor (sentence → essay)	Build on OSS	Writing Revolution progression + LLM rubric graders + instant-feedback loop; anti-LLM-submission stance is a product/policy design job, not a model job.	6–10 weeks
AI tutor layer (hints, no open chat)	Build on OSS	Constrained tutor on LLM APIs: hints/explanations/question-gen with guardrails; OpenMAIC's Socratic skill (dsh-openmaic, MIT) as a reference implementation.	4–8 weeks
Vocab / grammar	Build on OSS	Open word lists (AWL/COCA/NCERT glossaries) + FSRS scheduling + usage-in-context generation. Hindi/regional vocabulary = the localisation build Alpha never needed.	4–8 weeks
Placement + benchmark testing	Rent / buy	India: Ei ASSET / ACER PAT / MAP (international) rather than building norms. Keep our own mastery tests on top; rent the normed benchmark.	1–2 weeks integrate
Proctoring	Rent / buy	Thin wrapper over video-proctoring services + integrity policy. Keep data minimal (DPDP-sensitive).	2–3 weeks
Full K–12 knowledge spine content	Hard / later	~5,000-fact knowledge spine mapped to NCERT/NCF — generation pipeline (with Jev QA) helps, but this is a content-org effort measured in months. Start G3–8 only.	3–6+ months (content)
MAP-scale item bank + norms	Hard / later	Multi-year instrument building. Rent forever if possible; only build if volume justifies.	Years
Device telemetry / engagement vision model	Hard / later	Alpha scores "productive minutes" via an engagement model. We deliberately keep this app-level progress only (privacy, DPDP, and the earlier editorial call: no excessive student tracking).	deliberate non-goal

Build order that respects the evidence: the motivation economy + Dash + guide console come first (weeks, high leverage, no content dependency) — they're what made the same IXL content "a game" instead of "a fight" — then the fluency trainer and writing loop, then the reading pipeline with the Jev load-lint as the QA gate, with benchmarks rented and the knowledge spine built behind it. Workshop kits are the retention product and can start as curation. Full detail and costs live in the companion report's §8 (two-phase build plan).