How this actually operates
AI education has a credibility problem. These are the specific things done differently here — several of them enforced by the build system rather than by promises.
Most people arrive at an AI course having already been burned by one. So rather than ask you to trust us, here is exactly how this operates — including the parts enforced by code rather than by good intentions.
Every number is generated, not written
The "9 of 16 released" figure on this site is produced by a script that inspects the actual course files at build time. Nobody types it. If a course regressed or went missing, the count would drop on the next build whether we liked it or not.
No testimonials until they are real
This catalog is new. Any wall of five-star reviews here would be fabricated, so there is not one. When genuine reviews exist they will appear with the reviewer’s permission and without editing — and this paragraph will be replaced by them.
No income or outcome claims
We will not tell you this replaces your job, doubles your rate, or saves you a specific number of hours per week. Nobody can know that about you. Each course lists what you will be able to do when it ends — a claim we can actually stand behind.
You own the file, not permission to view it
A course is a single HTML file you download. There is no account to lose, no DRM, no server that can revoke it, and no company that can fold and take your library with it. Back it up like any other file you own.
The free preview is genuinely the course
Preview files contain the complete first four chapters — full narration, working demos, exercises. The paid chapters are physically removed rather than hidden behind a script, which also means there is nothing to bypass. What you hear is exactly what the rest sounds like.
We cannot sell an unfinished course
The build script validates every course file — chapter count, structure, narration length — and refuses to publish a purchase option for anything that fails. Unfinished courses show as in production with no way to pay. That is a property of the code, not a policy someone has to remember.
One price, no theatre
No countdown timers, no launch pricing that ends tonight, no price that shifts depending on how long you have lingered or which link brought you here. No subscription renewing while you are not looking. You pay once.
Simulated output is labelled as simulated
Where a demo shows a canned response instead of calling a live model, the page says so, visibly, next to the output. You are never shown a scripted result dressed up as a real one — which matters most in exactly the courses teaching you to be sceptical.
Refunds without an interrogation
Ask and it is done. No form, no exit survey, no "what could we have done better" before the money moves. A course that did not help you should not have cost you anything, and we would rather refund you than argue about it.
Written by the person teaching it
These are not licensed slide decks or outsourced scripts read by someone who has never used the tools. The material and the demos were built by the same hands, which is why the courses admit when something is unsolved instead of glossing over it.
We say when something is not worth buying
The fine-tuning course spends most of its length explaining why you probably should not fine-tune. The local-models course is candid about when the hosted API simply wins. Courses that only ever validate the purchase are marketing, not education.
Your data stays yours
The course files contain no analytics, no trackers, and make no network calls home. Where a course offers an optional bring-your-own-key demo, the key lives in memory for that page session only — never stored, never logged, never sent anywhere except the provider you chose.
If any of this turns out not to be true in practice, that is a bug and we want to hear about it. Tell us.