Inspirit AI sits in a category that barely existed a few years ago and is now crowded: paid AI enrichment for teenagers, sold partly on learning and partly on what it might do for a college application. The programme itself is one of the more credible ones. Students in grades 9 to 12 join a live online cohort of about ten sessions, roughly 25 hours in total, taught in small groups by instructors drawn from Stanford, MIT and comparable places. There is no programming prerequisite.
You start with what machine learning actually is, work through enough Python and enough of the standard libraries to be dangerous, and then build a project applying a model to something with a social angle, which in practice means healthcare, education, climate, law or accessibility. The teaching is the strong part. Because the groups are small, students get real attention, and the instructors are close enough in age to be relatable while still knowing the material properly. Having a graduate student who is currently doing research explain why a model is overfitting is a genuinely different experience from watching a recorded lecture, and for a curious sixteen year old that access is the thing worth paying for.
Inspirit has also built out sensible progression around the core offering, with a middle school version for grades 4 to 8, a one to one research mentorship for students who want to take a project further, and an in person Silicon Valley option, so a student who catches the bug has somewhere to go next. Now the part the marketing will not tell you. At around $1,400 for 25 hours you are paying roughly $56 an hour for small group instruction, which is not outrageous for live teaching but is a great deal of money for a family when the same student could work through Kaggle Learn, the Google crash course and Karpathy's videos for nothing. The honest question is whether your child will actually do that alone.
Most will not, and structure has real value, but that is the trade you are making rather than the one advertised. The bigger issue is the admissions story that hovers around programmes like this. Parents sign up because it sounds like a differentiator. It was, briefly.
It is now common enough that an admissions reader has seen many versions of the same two week AI project, and a project built in ten sessions tends to look like what it is. Students who benefit are the ones who treat the cohort as a starting point and keep building afterwards, which is exactly what the research mentorship track is for. The other limitation is depth. Twenty five hours gets you a working intuition, some competent code and a demo.
It does not get you a solid grasp of the maths, and any claim that a teenager finishes as an AI researcher should be read with a raised eyebrow. Instructor experience also varies because staff rotate between cohorts, so two families can have noticeably different experiences of the same programme. My rating of 3.6 reflects a well run programme at a price that assumes you value the branding. If your child is interested and you can afford it without straining, they will enjoy it and learn something real.
If the motivation is the application, save the money, have them build something genuinely their own over six months, and let that be the story instead.