Feature Creep 2026, 08 - The Problems Database - A Database Schema and Reading Up On AI Applications

Something slightly different this time, mostly because I just finished working through a book on QCD, and I decided that now that I'm not in uni anymore (or for now, we'll see, ambitions, ambitions), I should keep up with practice and do more of the problems in the books as I read them. I don't feel like including them in the weekly physics/math posts. I want to do practice, I don't want to return to weekly homework. I think. I've wanted to set up a database for fun for a while now, but I would like to get some use out of the whole thing, once it's done. I'm imagining something that can present me with a problem, tells me where to find it, and can give me a solution in markdown or LaTeX. It would be great if I could make an import script that could pull the problems directly out of the books I've read, and do the sourcing, but I don't currently believe in coding that by hand. PDFs are nice to read, but nice for very little else. I do imagine that this is what AI is tailor-made for, and despite generally being a little bit picky about what I use it for, I figure I might try my hand at making the AI pull the problems out of the books and put them into some markdown string that I could then write into the database. First things first though, I do need to sketch an UML, or I'm going to have to do a bunch of migrations because of oversight, so this week might not be as much actual code work, and more modelling and tooling, I guess?

After testing a few tools, I can probably safely rule out markdown as a solution. There's a few PDF to LaTeX converters out there, but they struggle really badly with the kerning of the PDF I'm using, and that doesn't exactly inspire much confidence in its accuracy in equations. I'm fearing that I'll have to move the data by hand, which is going to be a whole production on its own each time. I suppose it's very doable as I'm reading a book, but it's less cool, if I want to include the books that I've already read into the database as well, which was the use case I had envisioned. I suppose the backlog's also pretty large for that at this point, and the PDFs don't go anywhere - where I can't find them, anyways. I might populate the database as I go starting now, and when I feel like I'm in the flow, I might add more stuff from previous reads into it.

I'm pretty sure this is what I'm angling for, except that the problem_text and solution_text fields will be blobs containing LaTeX. I used dbschemasmith.com, and didn't find the blob option, but I'll change the SQL Script by hand, when I do the setup.

Upon setup, nevermind that export function, either I didn't read the relationships correctly, or those point the wrong way, so I had to futz around in the .sql file a bunch, but it's not the end of the world, since it's really not a complicated database. That's really all I want to do for this project for now. Now all I need to do, is insert the problems I come across. Getting them back out is going to be something for another point in time, when I'm further along with the math-timer.