Evidence and definitions
Every figure on this site comes from a public source and was checked against that source on 28 October 2026.
"Checked" means the cited source says what the figure claims. It does not turn a company's own claim into an independently measured fact, so each figure also shows what kind of source it comes from. Figures are dated, and figures from different dates or with different units should not be compared directly.
Kinds of source
- Regulatory filing
- A document filed with a regulator, such as an SEC annual report or a nonprofit's tax return. Issuer-prepared; filing alone does not mean it was independently audited.
- Independent study
- An evaluation by a party other than the company, such as a university, a research firm or a government review.
- Journalism
- A figure reported by a news organisation.
- Company claim
- A figure the company publishes about itself: press releases, websites, blog posts.
- Investor claim
- A figure published by an investor or accelerator about a company it backs.
- Third-party estimate
- An estimate from a data provider such as a funding or web-traffic database. Indicative only.
- GitHub
- Public repository counts. Stars and forks measure developer attention, not learners or customers.
What the numbers mean
- Users, learners, students
- Usually everyone who has ever signed up or been served, including free use. Not monthly active users unless stated.
- Classrooms, schools, institutions reached
- Exposure or availability, which may include pilots and free use. Not the same as paying customers.
- ARR
- Annual recurring revenue: an annualised subscription run-rate at one point in time. Not the same as revenue recognised over a year.
- Funding
- Money raised from investors. It is not revenue and says nothing about profitability.
- Valuation cap
- A ceiling used when early investments convert to shares. It is not an actual valuation of the company.
- Website visits
- Modelled sessions from a traffic-estimation service. Not unique people.
- Effect size (g)
- How large a measured learning difference is, in standard-deviation units. In matched (non-randomised) studies it depends on who used the product and how much.
- Content-grounded
- The AI answers from specific course material. This does not by itself mean a particular technique such as retrieval-augmented generation.
- Adaptive
- The product changes what comes next based on the learner's answers. The word alone does not say which method is used.
Figures not shown
These figures appeared in earlier research about these companies but could not be confirmed, or were contradicted by their own source. They are not used anywhere else on the site.
contradictedKhanmigo / Khan Academy: Internal tutoring test result — +6 percentage points tutoring improvement
Source labels the changes +2.7% and +3.4%, 6.1% combined, not percentage points. Separate online product tests and an internal next-item measure, not a standardized learning-effect estimate.
contradictedAristotle (Hey Aristotle): Implied beta tutoring per student — ~1.5+ hours/student based on 1,500+ hours and 1,000+ students
Both inputs are lower bounds. Their ratio is neither an identified average nor a lower bound: the actual student denominator is unknown. Report the two totals separately; neither establishes retention or high-dosage exposure.
unverifiedSchoolAI: Estimated ARR — Third-party estimated $17M-$25M ARR
No identifiable accessible source or method for this range. The cited financing page does not substantiate ARR. Do not publish as a numerical current metric.
unverifiedAlpha School: Expected expansion — Expansion toward approximately 50 campuses
An approximate expansion narrative is not a verified operating-campus count.
unverifiedAlpha School: Learning-growth claim — 2.6× faster learning growth
Current homepage says 2x learning in 2 hours. The precise 2.6x historical calculation and underlying data were not recovered; a changed slogan is not evidence contradicting the historical result.
unverifiedAistotle: Team’s prior learner reach — 30M+ learners across applications built by the team, not Aistotle
Domain DNS failure; no current primary documentation recovered. Keep this separate from Hey Aristotle. No product traction or current price may be inferred.
unverifiedAistotle: Free plan ceiling — Free up to 50 learners
Domain DNS failure; no current primary documentation recovered. Keep this separate from Hey Aristotle. No product traction or current price may be inferred.
unverifiedAistotle: Pro price — Pro $24/month on annual plan
Domain DNS failure; no current primary documentation recovered. Keep this separate from Hey Aristotle. No product traction or current price may be inferred.
unverifiedAistotle: Team price — Team $60/seat annually
Domain DNS failure; no current primary documentation recovered. Keep this separate from Hey Aristotle. No product traction or current price may be inferred.