Thank you. I think that's a good explanation for much of the phenomena and your insight would have strengthened my post. I think you're probably right about where many of the people who see it that way are coming from. But not all of them...
> After about 30 minutes of futzing with prompts, it was clearly stuck trying to create the proper regex and I just went back to pasting alter tables and getting deletes back until the job was done.
If you're copying and pasting SQL statements, then SQL statements are the intermediate product. The fact that you didn't carefully review them and just ran them immediately is no different than an LLM producing Java source code that you shipped to the user without reviewing because it worked correctly in your limited testing. There's still an intermediate product that should have gone through the same software development robustness process that all source code should go through, you just didn't care to do it (and maybe rightly so if it's not super important).
There's no reason to be assmad. I know how to produce the sql statements in question. Being able to give the AI an example of what I wanted was part of the process. But doing hundreds of them is prone to typos. And cranking them out by hand is just slower. Especially when there are compound keys. Giving it to the AI just made the process faster.
But the AI could not, absolutely not, generate a program that could do what the AI was doing. Which would be really nice, because I'm probably going to go back through those hundreds of statements again in the future. The database is evolving, and my task is to migrate it to a whole new one. I would really rather have it give me a program. But it could NOT do that.
> I've come to hate every cookbook that starts with 100 pages of here is a tour of my pantry, which sounds a lot like, here is how to use pip!
Yeah I agree. I hate when books do more hand holding than the reader clearly needs to the point of tedium. Plus many of those setup steps like how to use a package manager change over time and make the book stale instead of evergreen. And Austin was clearly not writing an absolute beginners book.
That's why when I pitched both the Classic Computer Science Problems series and Computer Science from Scratch I explicitly told publishers in the proposals that I was not writing a beginners book (been there, done that). I was clear that I was writing an intermediate book for people that already know programming.
It's a different, more narrow audience. But you can be successful if you write a good book. It's also a less tapped market and luckily publishers were able to see that.
Absolutely. And further because when you prompt ChatGPT as you write your ray tracer you don't know what the important things to ask are. Sure, you can get their with enough prompts of "what should I be asking you" or "explain to me the basics" of so and so. But the point of the book is all of that work has already been done for you in a vetted way.
> This is why most publishers won't even talk to you unless you have a finished manuscript already
This is absolutely not true in the world of technical publishing. I mean books published with publishers like O'Reilly, Manning, No Starch, etc. Usually you come to them with just a proposal and a couple chapters or even just a proposal. Or their acquisition editors actually reach out to you. It's the exception (not quite rare, but definitely less than 20% of books) that comes to them with a finished manuscript. I did that with my last book. I've published 5 technical books across three different technical publishers, so I know a bit about this business...
I'm just replying to this comment to not discourage people who just have an idea and not a finished book yet but have the motivation to finish and want to get a deal.
Ironically, I was working on a book with a similar concept in the same time frame that came out as "Computer Science from Scratch: Interpreters, Computer Art, Emulators, and ML in Python" with No Starch Press a couple months ago. Like Austin's book it contains a CHIP8 chapter and a couple chapters on making a programming language. The difference with regards to his experience and my experience in writing it with a traditional publisher, is that I was an experienced author so I felt comfortable finishing the entire book first before shopping it around to publishers. I didn't want too much scrutiny around the core concept and I was getting similar signals of "every chapter must have AI."
I wrote a similar blog post a month ago describing the process of creating the book and getting it published called "Writing Computer Science from Scratch":
Some in this thread have wondered what publisher Austin was working with. Based on my experience working with three different technical publishers and the setup and terms Austin was offered, my educated guess would be Manning.
I will critique the blog post a little bit. It's presented as a critique of the experience of working with the publisher, but ultimately I'm reading between the lines that the book failed because he was missing deadlines. He wrote that "life got in the way" and I think he lost his motivation only partially because the publisher wanted AI in more of the book. Many of the trials he had along the way: dealing with a development editor who wants to tailor your style to a particular audience, a technical editor who needs a couple chapters to warmup, back and forth on the proposal, etc. these are all really par for the course when writing a technical book. Ultimately you have to be self-motivated to finish because of course the development editor, technical editor, etc are going to disagree with you from time to time and try to push you in different directions. If that alone is so demotivating to you, it's just not for you to work with a publisher.
PS I think his blog is really good and he should think about self publishing under a time frame and terms he is more comfortable with.
Agreed. The one time I worked through a publisher I beat every schedule and it was all smooth enough.
I’m glad I did it but I’m not sure how much the publisher added beyond some prestige and a few bucks. The first edition in particular I felt I needed to pad out a bit to meet length requirements.
I’ll also note that the publisher was right to bring up AI, even if they did not do it in an artful way. He himself comes to doubt the need for his book in the era of LLMs and he says that is part of why he cancelled. To his publisher’s credit they raised the issue early in the process where a pivot would have been more practical.
In fairness to the author, he presents a reasonably balanced view and it did not read to me like “my publisher sucked.”
Manning makes sense - all the details fit, and there aren't that many. Publishing is a stupid business that makes less and less sense every passing day. Self publishing and going through an outlet, marketing for yourself, or contracting out the relevant tasks, will save you a ton of money for anything publishers can offer anymore. They survive more and more often on grift and network effects that are increasingly irrelevant and often run counter to the interests of a given author or work.
Glad the author got out relatively unscathed.
Self publish - especially with AI available to get you through the stuff where you just need superficial or process knowledge, like which firms to hire and how to market a self-published work, what boilerplate legal protections you need. You'll get 99% of the value of a big publishing firm at a small fraction of the cost, and you won't have to put up with someone else taking a cut just because they know a few things that they don't want to tell you in order to justify taking your money.
> AI available to get you through the stuff where you just need superficial or process knowledge, like which firms to hire and how to market a self-published work, what boilerplate legal protections you need
Putting aside for a moment that nobody should be trusting a frequently-hallucinating AI algorithm with any of the above...
Your world-view is one of those that returns to the old adage "it only works if you value your time at zero".
Its the sort of thing we see in tech the whole time. Some dude saying "oh, I can just fix my motherboard myself".
Or in the automotive sector, someone with experience and kit fixing their own engine block.
Well, sure you can dude. Because you've got the domain expertise, you've got the kit AND you are willing to value your time at zero.
However in the majority of cases, if you do not value your time at zero, then spending even just a few hours waving an oscilloscope and soldering iron over the proverbial motherboard is time better spent on other tasks and the "more expensive" option suddenly does not look that expensive any more.
And that is all before we address the other elephant in the room.... Your suggestion that it is easy to self-market a self-published work.
Maybe if you are a well known and respected author, such as Mr Performance (Brendan Gregg) or Mr Oracle (Tom Kyte) etc.
But if you are just Joe Schmoe. And perhaps especially if you are Joe Schmoe who's just written your first self-published book. The outcome is unlikely to be the same.
You can’t stop it from being mercilessly pirated, and it’s a fool’s errand to try, unless you want to go down the user hostile route much of the ebook industry insists on (vendor specific reading apps/devices serving as DRM).
Bandcamp learned this lesson. GOG learned this lesson. They both provide services users love, without DRM, and just accept that there is no capitalistic scarcity inherent to digital goods like there is to physical ones. An indie ebook publisher would be wise to heed those teachings.
As someone who has sold software online for 20 years, I am very familiar with piracy. It rankles to put a lot of work into something and people just help themself for free. But customer hostile DRM is not great either.
I was envisioning the latter (changing behavior on the fly). Think the hot-reload that Flutter/Dart provides, but on steroids and guided by an LLM.
Interpretation isn’t strictly required, but I think runtimes that support hot-swap / reloadable boundaries (often via interpretation or JIT) make this much easier in practice.
Thanks for posting this detailed history and breakdown. And congratulations on your success. As someone with a podcast in the same ballpark of downloads, I will say it's a real testament to how much people must like The Program that you have managed to monetize to the tune of ~$30K CAD in a single year with just 140,000 downloads in the same year. That's really good! People must love what you're doing and that must be very gratifying. Cheers!
A very astute observation - The Program's audience is indeed loyal if not large. I mentioned in some previous reports that I might have painted myself into too big of a niche with a sci-fi audio drama at the intersection of IT and humanities, but people who like it seem to really love it. 1000 true fans, and all that I guess.
Very related but self promotional—I have a hobby business selling restored Mac mini G4s. I clean all of them internally, upgrade them with 128 GB SSDs, max them out at 1 GB of RAM, put a new clock battery in, and pre-install the Mac OS 9 Lives hacked version of Mac OS 9 that runs on them. You can buy one from me here:
I don't think I'll start pre-installing System 7 since most of my customers are using Mac OS 9 (and the domain is os9.shop!), but you could certainly get a machine from me with Mac OS 9 and install System 7 yourself if you so desire.
My customers have included a lot of real businesses running legacy software who want the fastest, least intrusive, and least energy intensive Mac OS 9 desktop machine they can buy. I've sold to dentists, veterinarians, museums, and auto repair stores. You'd be amazed how many people are running Classic Mac software in 2025.
> My customers have included a lot of real businesses running legacy software who want the fastest, least intrusive, and least energy intensive Mac OS 9 desktop machine they can buy. I've sold to dentists, veterinarians, museums, and auto repair stores. You'd be amazed how many people are running Classic Mac software in 2025.
Is there any software that only runs on System 7, and not 9.2.2?
Did you have to do anything special to get the SSD to play nice with OS9? I tried adding one to a 300MHz G3 iMac and it took forever to initialize on boot and would randomly stall a lot.
I use a mSATA to IDE adapter that I buy in bulk. This is the Amazon available equivalent of it: https://amzn.to/48qEaOm
I use only 128 GB mSATA cards from reputable brands.
I always do the following:
- Boot from the Mac OS 9 Lives 9.2.2 image (v9 of the image) by CD
- Wipe the SSD using Disk Utilities 2.1
- Restore from the CD
I will say this fails perhaps 1 out of 20 times. Hard to say how often this is an actual hardware failure versus some kind of incompatibility with the mSATA SSD since I do use a range of brands. I am always using the same adapters.
Seems more of a curiousity than something practical - in particular, the System 7 "native" on the Mac mini G4 is missing a lot of drivers. There aren't that many situations where software runs well on System 7 tha doesn't on Mac OS 9.2.2, and for the rare case that it does, emulation in something like vMac is sufficient.
reply