Both cite uploaded material. What is actually different?

What happens after a miss. Zelvi names the misconception, re-teaches it, has the student retry, records whether they got there on their own, and hands a student who is still stuck to a person. Those records add up to the measure a district or college sees.

Can a school use both?

Yes. Flint for teacher-built activities across many subjects; Zelvi for a course where one syllabus defines what must be learned and every recovery is recorded.

Comparisons

An activity builder, or a recovery loop for one course.

Flint and Zelvi both cite the material a teacher uploads. Flint gives teachers more tools inside each activity. Zelvi follows each miss through to a recorded result and a person.

Choose Zelvi if

You want every wrong answer followed through: misconception named, re-taught, retried and recorded.
You want students stuck on the same misconception brought into one short live session with their instructor.
You want a district or college to see each school, teacher and class on one recovery measure.
You teach gateway college, allied-health or workforce courses and need interactive STEM and anatomy diagrams.

Choose Flint if

You want web search, a whiteboard and generated images, slides and videos inside the tutor.
You want AI grading against your rubric feeding a gradebook with weighted categories.
You want per-question time limits, open and close dates, and a shared library of ready-made activities.
You need published evidence and a large teacher community today.

Feature by feature

Zelvi compared with Flint, feature by feature
Feature	Zelvi	Flint
Answers cited to the material	Every grounded answer cites the course page; anything beyond the material is labelled	Numbered citations to uploads and to web results
After a wrong answer	Misconception named, re-taught, retried and recorded; still stuck goes to a person	Session analysis of strengths and gaps, and a follow-up activity built from them
Students stuck on the same thing	Grouped by misconception and invited to one short live session with the instructor	Analysis names the students who share a gap
District and college reporting	Each school, teacher and class on one measure: how often a stuck student got it right on their own	Usage analytics and session results
Tools inside the tutor	54 STEM and 27 interactive diagrams, 6 anatomy labelling tasks, realtime voice, Python that runs in the browser, and every sum in a reply re-checked by code	Python calculator, code runner, whiteboard, web search, and generated images, slides and videos
Grading open answers	Answer-key marking; for written work, a suggested mark against your rubric that the teacher confirms and releases	AI grades against your rubric and you can override any grade; gradebook with weighted categories
Time limits and deadlines	Due dates, and a time limit on quizzes with a countdown and auto hand-in	Time limits per session or per question, open and close dates, auto-submit
Published evidence	Measured in every pilot on your own students, with sample sizes; first results publish at the end of the pilot term	An ESSA Tier 4 rating
Primary audience	Colleges, allied-health and workforce programmes, with K-12 sign-in supported	K-12 schools, with an offer for universities

Where Flint fits

Flint gives each teacher a very capable activity builder: a Python calculator, code, web search, generated media, rubric grading with override, time limits and a large activity library. Zelvi does fewer things and follows every wrong answer through to a recorded result and, when needed, a person.

What our side of this rests on

Answers cited back to your own material, with the instructor one tap away when the material does not cover itOn our go/no-go benchmark every answer retrieved the right passage, carried a valid citation and stayed inside the material, and every out-of-scope question routed the learner to the instructor rather than a guess — against 54.2% for handing the model the whole document.
Students stuck on the same misconception are grouped, and the instructor brings them into one short live sessionGroups are built from the escalations Zelvi already opened, keyed on concept and misconception, and only students enrolled on the course can be invited.
A district sees every school, then each teacher, class and student, week by week, on the one measure: how often a stuck student got it right on their own after Zelvi helpedDistrict figures are added up from each school’s own counts, never an average of percentages, and a week with too little practice shows a dash rather than 0%. A district administrator can open any of its schools; a school administrator cannot see the district or a neighbouring school.
Interactive STEM visuals — graphs, geometry, circuits, reactions — that the learner can manipulate54 STEM diagrams and 27 interactive ones ship today, and every one is re-drawn and checked before a release.
Talk to the tutor out loud, interrupt it, and be understoodThe realtime voice loop ships today and is exercised before every release.
When a student says something that may mean they are in danger, Zelvi stops tutoring, says it is an AI, gives crisis resources for their country, and alerts the institutionEighty-nine automated checks, covering crisis language in English, Spanish and Hindi, and coursework such as "why does Ophelia commit suicide" that must not raise an alert. The alert email never contains the student’s words.

Questions people actually ask

Both cite uploaded material. What is actually different?: What happens after a miss. Zelvi names the misconception, re-teaches it, has the student retry, records whether they got there on their own, and hands a student who is still stuck to a person. Those records add up to the measure a district or college sees.
Can a school use both?: Yes. Flint for teacher-built activities across many subjects; Zelvi for a course where one syllabus defines what must be learned and every recovery is recorded.