Note by @fastfinge

🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca

You can, in fact, learn things from #AI. I asked AI the question "But how can I do this without using an LLM or hardcoding everything myself?" Now I know more than I did before about Finite-state machines. Plus used an LLM to rip an #LLM out of a tool I use.

🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca

I would, of course, guess that about 10 percent of the things I now know about Finite-state machines are wrong. But finding that out doesn't cost me anything, really.

x0 @x0@dragonscave.space

@fastfinge I actually learned about finite state machines, pushdown automata, and the like as part of my degree. A deterministic FSM is pretty simple. I personally implement them as an enum and a function which is called repeatedly with a switch/case statement on a continuously changing state variable. About the most direct form of such a thing.

🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca

@x0 Yeah, I don't have a degree, I think that's the issue. So I just kind of do stuff, even though I've worked in the industry for 20 years.

x0 @x0@dragonscave.space

@fastfinge Meh, honestly it isn't necessary if you can learn by other means. Half the theoretical shit I got a basis in and promptly forgot only surfaced much later in the middle of something unrelated. I had also effectively reinvented the finite state machine years before I even went to college and just learned the formal name for a thing I built when I was 15.

x0 @x0@dragonscave.space

@fastfinge I used that whole computer languages and automata theory more than most I think because as I lived in BGT land I had to hand-roll way more parsers than anyone ought to be doing. And a parser is an automaton that accepts a language of the correct form.

x0 @x0@dragonscave.space

@fastfinge I wrote, from first principles in BGT, a fully functional JSON parser sans Unicode with multiple escaping modes, two orthogonal dialects in one system which allowed me to construct a string specification in order to parse a user inputted string, refined multiple textual map dialects where delimiters weren't enough, and when I moved to NVGT then wrote a msgpack parser.