A K-12 “vibe coding” platform whose headline 700,000-student figure implies exactly one class roster per registered teacher, whose binding privacy policy never mentions FERPA and states the service is “not intended to be used by children,” and whose free tier runs on OpenAI credits scoped to a single awareness week. Co-led by Brighteye Ventures and Day One Capital.
The company’s own independently-reviewed HundrED profile said 150,000 learners in 31 countries as of August 2023. The figure is cumulative since 2018, spans a discontinued consumer hardware app, and rests on deliberately anonymous accounts — which makes it unauditable, including by imagi.
Marketing claims full FERPA, COPPA and GDPR compliance with “zero data retention.” The operative privacy policy never mentions FERPA, states the service is “not intended to be used by children,” names Shopify for personalized advertising, and describes retention rather than disclaiming it.
OpenAI does — via $1M in credits scoped to CS Education Week and the Hour of AI. The application layer is Lovable, a company valued at $6.6B. imagi owns the curriculum wrapper, which is the cheapest layer in the stack to reproduce.
Key Finding: The team is real and unusually persistent — eight years, one full pivot, a killed consumer product, and a serious senior GTM hire. But no independent source anywhere reproduces the 700,000-student figure except by quoting imagi, the compliance marketing is contradicted by the company’s own binding privacy policy, and the growth engine runs on a third party’s expiring credits layered on a partner 1,200× its size.
The product that generated the coverage is a classroom-safety wrapper around two other companies’ infrastructure. imagi owns exactly one layer — and it is the cheapest one to reproduce.
OpenAI credits, $1M, scoped to CS Education Week and the Hour of AI. Promotional and time-boxed, not perpetual.
Lovable — $653M raised, $6.6B valuation, ~$500M ARR. No contractual obligation to preserve a K-12 channel.
Teacher-controlled access, prompt screening, anonymous student accounts. Genuinely useful. Genuinely thin.
CSTA, ISTE and state-standards alignment plus ~30,000 teacher relationships. The one durable asset in the file.
Undisclosed. No cloud provider, no AI vendor and no subprocessor is named in any imagi legal document.
The capital asymmetry is 1,200 to 1. Lovable’s $6.6B valuation against imagi’s $5.5M lifetime capital (DERIVED). Lovable can build its own education tier, reprice, restrict under-13 access to manage its own COPPA exposure, or be acquired — with no obligation to imagi in any scenario. The honest counterweight: Lovable today has no education product at all, and its only student offering is a higher-education discount with no K-12 tier. imagi is effectively Lovable’s K-12 channel rather than a tolerated parasite on it, and neither Lovable nor Replit is currently pointing capital at K-12. That is a real position — it is simply one held at someone else’s discretion.
Launched free to teachers in April 2022. Adopted into school curricula worldwide. Deprecation announced November 2023 citing financial unsustainability. Shut down 1 August 2024 — roughly eighteen months after going free. imagi’s entire top-of-funnel currently runs on that same arrangement, one layer further removed, with the subsidy coming from a third party whose relationship is with an awareness campaign rather than with imagi.
Co-lead. Genuine edtech specialist — >€150M AUM, exits including Genially, TeachFX and Ironhack. Credible and on-thesis.
Co-lead. A Budapest B2B-software fund whose stated focus is cybersecurity, identity and data infrastructure. Edtech is off-thesis — likely a Hungarian-founder relationship deal.
The consumer Android app was formally discontinued 4 December 2024. The company pivoted to US K-12 software the same year.
The original programmable LED wearable. Kickstarted in 2019 — 624 backers, 106% funded, delivered. Still sold today as a low-volume accessory.
The only third-party evaluation found. Impact rated “Exceptional.” But it is a curation panel, not an efficacy study — and it cites no peer-reviewed outcome research.
The imagi Edu iOS app carries a 1.0-star rating from two ratings. The product is primarily web-delivered, so this is a minor surface — but it corroborates nothing.
Every account of imagi’s scale traces to a single company-sourced number. Three pieces of arithmetic complicate it.
imagi’s independently-reviewed HundrED profile, updated August 2023, recorded 150,000 learners across 31 countries. The current claim is 700,000 across 140. That is 4.67× the learners and 4.52× the countries in roughly two years (DERIVED). A country count quadrupling is not what district sales looks like — it is what a free global awareness campaign looks like.
700,000 students divided by 30,000 registered teachers is 23.3 students per teacher (DERIVED). That is not a measured number. That is one US class roster, exactly — consistent with teacher signups multiplied by an assumed class size, which is a modeled reach estimate rather than a count of distinct learners.
The count runs “since launch” in 2018, so it folds in the consumer imagiCharm app the company discontinued in December 2024, 624 Kickstarter backers, 2,000 hardware units and every free-tier trial. The seed is being marketed partly on the accumulated exhaust of a business that no longer exists.
The architecture makes the number unauditable. imagi’s student accounts are deliberately anonymous and teacher-generated, with no PII. That is genuinely good privacy design — and it means anonymous accounts cannot be deduplicated, by anyone, including imagi. No ARR figure, paid-seat count, contract value or net revenue retention number appears in any source. The disclosed growth figures — “30-fold user growth,” “three-fold ARR” — are multiples off a base that has never been published.
What $18 per student is actually competing against. K-12 computer science is not sold per student. CodeHS — the category’s strongest operator, which raised $4.23M in total, has taken no capital since 2017, and paid cash in 2025 for Tynker — states plainly in public RFP responses that it “does not offer per student pricing options.” The category prices per teacher at secondary level and per building at elementary. The only genuine per-seat anchor in K-12 is Minecraft Education at $5.04 per user per year, which makes imagi’s $18 roughly 3.6× the reference price (DERIVED). For comparison, Khan Academy sells district Khanmigo access at $10 per student per year — and that product, with roughly 2 million seats across 795 districts, is the sector’s first public AI-tutor disappointment: only about 15% of users engage regularly, and Sal Khan has publicly called the rollout “a non-event.” imagi is asking nearly twice Khanmigo’s district price, from a standing start, in a year when global edtech venture funding has fallen roughly 82% from its 2021 peak.
What the operative legal document actually says. The binding privacy policy published by imagi Education, Inc. contains no mention of FERPA, states the service is “not intended to be used by children” and that the company does “not knowingly collect any personal information about children under the age of majority,” names Shopify as its only third party — for hosting, payments and personalized advertising — and describes retention that “depends on different factors” rather than disclaiming it. The separate imagi Edu policy page, where the FERPA and zero-retention language reportedly lives, did not render content on repeated fetches. No SOC 2 reference, no Data Protection Addendum, no subprocessor list, and no entry in the Student Data Privacy Consortium registry that most US districts require before purchase.
Seven structural risks that the $4.5M seed does not resolve.
OpenAI’s $1M credit is scoped to CS Education Week and the Hour of AI, not granted in perpetuity. Growth, freemium access and probably the “zero data retention” property all sit on it. When credits lapse, imagi absorbs inference cost against an $18/student list price.
The operative privacy policy omits FERPA, disclaims child users, names Shopify for personalized advertising and contains no zero-retention clause. No SDPC agreement found. A single district counsel review converts this from a paperwork gap into a sales-stopping event.
Cumulative, lifetime, cross-product, free-tier-inclusive and structurally unauditable by design. The company’s own independently-reviewed 2023 figure was 150,000. No paid-seat count is disclosed anywhere.
The differentiated AI product is a permission wrapper around Lovable’s IDE. Lovable can build an education tier, reprice, restrict under-13 access for its own COPPA exposure, or be acquired — with no obligation to imagi.
EST. ~$3.4M annual burn on 31 heads implies roughly 16 months on $4.5M (DERIVED) — before the stated US hiring, district sales build-out, new platform launch and compliance remediation. District procurement runs 9–18 months and is gated on spring budget adoption.
The ESSER liquidation deadline passed in March 2026 and districts are cutting programs while specifically reassessing which ed-tech spend produces measurable value. imagi publishes satisfaction percentages with no methodology and no efficacy research.
Peninsula School District expects to save $200,000–$220,000 a year by vibe-coding its own internal tools rather than buying vendor software, with one app that would have cost $30,000–$40,000 built in hours. imagi’s product teaches that exact capability to that exact buyer. Separately, 2025–26 research on student-AI interaction finds vibe coding produces confidence gains without corresponding comprehension gains — a problem for a curriculum branded to a term with a roughly twelve-month half-life.
imagi is a two-year-old company with an eight-year-old cap table. The founders did not quit through a failed consumer hardware business, and that persistence is the strongest thing in the file. But the round is being marketed on a reach number that spans a discontinued product and cannot be audited by construction, sold into K-12 on compliance claims the company’s own binding privacy policy does not support, and grown on inference credits that a third party granted for one awareness week. The diligence question is not whether 700,000 students exist — it is how many seats have ever been paid for, and that number appears in no source, anywhere.
Based entirely on publicly available information, including the TechCrunch announcement of July 23, 2026. Every figure is labeled CONFIRMED, DERIVED or EST. in the body text.
