TL;DR: If this course is not for you, due to multiple violations of warranty, you are still entitled to a refund, should you still desire one. This is not to personally disrespect anyone, it's just business. As a result, I hope that this will be handled professionally (and not censored).
I think the course may get better, as it did a little in weeks 2 and 3. I hope to see it continue to improve.
So, as this is not to be taken personally, I'll try to keep from putting out my opinion. Instead, I'll simply point out the reasons why prohibiting refunds is not acceptable, and neither is censoring discussion on the topic.
Let's begin with how the course was clearly described (from this video):
There are a lot of issues with that as the foundation of this course's warranty.
I know these subjects, but many others did not have the required experience, and the code for the things shown in the videos was often not provided, requiring students to produce or find it on their own.
Producing a product in accordance with any provided details is the foundation of warranty law--which is grounds for a refund. Details communicated about a sold product constitute your warranty, under US law, whether or not you think you have a warranty.
There are two pillars to this course as advertised: machine learning and monetization.
The machine learning concepts and implementation frameworks themselves are inadequately explained, even though it's clear that the target audience is indeed absolute beginners.
This is made very clear in multiple cases throughout the course. For example, in week 2, it is explained what a web app is (see 00:01:40 of the main video on week 2). There are lots of other places where it is made clear that the target audience isn't expected to have any prerequisite experience.
Participants are never given actual guidance on how to make any money, which is the other pillar to the course. Several times it was mentioned that participants could take the GitHub project that was copied and pasted into the course and "put it behind a paywall" or they could "add Firebase as a database later on," allowing them to host real, living customers, but guidance has not yet been given on how to do that for any of the projects advertised (see 14:10 of the main video on week 2).
This leaves the course, thus far, at a 30% failure rate to uphold the warranty (on that point alone).
Participants are never guided through the process of setting up an actual, monetized app instance. They simply receive some conceptual verbal walkthroughs while code is shown to them on the screen. Many of the gaps left behind in the course are then filled with related material. Sometimes the code shown contains features for monetization, but these are delivered in the manner of a feature showcase, not a lesson. It is assumed that participants are left to figure it out on their own with whatever guidance they can find.
This course relies on a copious amount of that related--often third party--material, which is subject to change and which does not always align with the course.
What customers were "sold" and what they received is a violation of warranty law. It's not acceptable to prohibit refunds in this case, despite attempts made with the unannounced policy--which itself is an issue, due to the fact that it wasn't available upon purchase.
In addition, it wasn't just published after the purchase date, it was published after most of the policy's window for refunds had even passed! In other words, the last-minute refund prohibition alone is illegal. Here's the bread-crumb trail of actual sources, if you would like to see this verified:
^^^ This is what first piqued my suspicions. (Shown: A message posted Sep. 8th implying the policy wasn't originally in place.)
^^^ This index may reflect an update on the page, not its original publishing. However, this is not the final piece of evidence. See the following image. (Shown: A search revealing the policy's index date.)
^^^ This is essentially the "smoking gun." The policy wasn't readily available to the customer upon purchase of the course. (Shown: A "Wayback Machine" snapshot from Aug. 12th showing the above policy was non-existent upon selling the product.)
^^^ Instead, it was put in place later, after the refund window specified had even passed. (Shown: A "Wayback Machine" snapshot from Sep. 10th showing the policy was now in place.)
Unfortunately, that's not all. There are still promises left in this course that cannot be excluded from the warranty, and thus cannot be protected from refunds by the policy. For example, it was stated that at the end of the course, graduates would receive a connection from Siraj to someone valuable in his network, if they chose not to "make a startup."
As this cannot be tested within the 14 day policy period (because the course is 10 weeks long), this is still grounds for at least a partial refund at the end of the course should this fall through. It would still be grounds, even if that turned out to be the only thing wrong with this course--and there is much more, as was shown.
The idea that graduating participants will be put "into startups" at the end of this course is at best an uneducated statement and at worst illicit advertisement. Not to be crass, but I would recommend sending these groups over to an accelerator or a consultant to avoid crushing dreams.
I worked as a contract consultant and educator in a startup in the Kubernetes space before leaving my employer. I designed, built, and managed a training curriculum that we delivered around the globe before the company was purchased, which sold for over 1k USD per seat per day. I have a degree in entrepreneurship from professors who have "been there and done that." I know that talk is cheap and that you have to chisel through a brick wall called reality--with your bare teeth--in order to create a real exit-viable or lifestyle-viable product.
Just some more accurate marketing would have been great. If the course was called "Diving into AI Applications," I would feel very confident in my investment. However, while I hope that us being guinea pigs was helpful, there is certainly a feel to the course that this is "testing in production." 😅
In conclusion, there's much to learn from this experience for whoever was involved in designing and producing this course. End-user testing may have revealed the flaws in this course far in advance, as well as many other things that could have been done. I still respect Siraj, and I'm giving him the benefit of the doubt by assuming the issues noted weren't intentional. 😅
I'm assuming this is just his first experience producing and selling a course--and there's no shame in that. Hey, I remember my first training contract, and it was a terrible experience for me in every way. 😬 But there is definitely shame in prohibiting refunds... so... just don't do that.
I admire Siraj's desire to hype up ML and inspire others to learn it, and I'm not saying any of this to be taken personally. It's just business. I'd be happy to provide some guidance next time this course is refreshed, as there are a lot of issues, enough that I felt it would be too much to list here. Who knows, it could be a comeback as sweet as No Man's Sky. 🙂
However, if anyone would still like a refund, there should be no prohibition against providing one in full.