Here’s an impressive Show HN project: a comprehensive, reverse-engineered, LLM-written technical documentation of how a calculator’s operating system works. What a resource!

Insert a standard paragraph about how we live in interesting times. I keep seeing projects like this increasingly more often: complex, working beasts – that would previously have been nigh impossible to implement by one person in their spare time – are now springing up like mushrooms after the rain. Here an Apple Lisa emulator; there a modern implementation of a long-forgotten language from the 1960s.

Someone asked a simple question in the comments under that Show HN. Sadly, it seems to be gone now; I can’t find it anymore. It went something like this: “Okay, so you’ve built this. Now what?”

Excellent question! And one, I think, that we need to be asking ourselves more often these days. I know I want to.

It’s all too easy to focus on the (very real) excitement that arises when you see a working result of your prompt appear out of nothing in front of your eyes. But that’s just a dopamine rush that will pass soon. Imagine it passed already; imagine you have your creation in front of your eyes. Now what?

Is it because someone is actually going to use your creation, in anger? Have you actually solved someone’s problem? Is that someone you, perhaps?

That reverse-engineered documentation – is anyone going to read it? Will it facilitate someone’s learning? Will it make them more knowledgeable? If there are any hallucinations inaccuracies in it, will they be able to point them out?

Have you learned something by doing this? Will you be able to carry that knowledge forward to your future experience?

Or are you in it just because of that dope rush? Will it just bit-rot on GitHub, unused, undiscovered, after you announce it to the world and the trickle of upvotes inevitably dries up? Will it just be a bookmark, one of thousands soon to become millions, in the Babel library of all imaginable software? Or will you keep prompting for more rushes?

* * *

Let me be clear: I’m not picking on any project in particular. I won’t be judging you if your honest answer is “dope rush”. It’s fine to desire one; it’s fine to experience one once in a while. It’s a value in itself.

But – need I remind you? – there are ethical and societal hazards to using LLMs: using them carries a cost, and it’s our responsibility to offset that cost in some way. And even LLMs aside, it’s probably not a good idea for your actions to be rush-driven. It’s good to have a conscious purpose.

I have some LLM-facilitated pet projects, so I think it’s only fair to answer “Now what?” publicly, as honestly as I can.

• Edsger: For the most part, I think, I did it for its hack value. But also I taught myself a lot about modding the reMarkable, and I have plans to reuse parts of Edsger for something more practical, something I see myself using: an integration between it and Clerk.
• clj-concraft and JMorfeusz: I’ve touched on it in the blog post already. Long story short, it’s about (1) learning about the underlying mechanisms, (2) planning to use them as building blocks of a project that sees some, if niche, use.
• Smaller things / toys: I admit some of these were rush-only. But there are some that I am regularly using myself (my .emacs.d comes to mind); and, for a more playful example, the one-shot editor for poems in Pilish was posted as an aid to a Facebook group with monthly challenges for poets that I’m a part of.