adamddev1

One of my favorite memes shows a kid having a conversation with his father in the car:

Kid: Dad, why is my sister named Rose?

Dad: Because your mother really like roses.

Kid: Oh, ok, thanks Dad!

Dad: No problem, Costco Hot-Dog!

> "Exponentially growing costs"

I don't think people thing don't think enough about the exponentially growing costs with AI. The cost of tokens, the cost of bugs, the cost of technical debt, the cost of software bloat and slowdown, the cost on the RAM market, on the environment, the costs of the little lies and hallucinations that creep into our codebases, blogs, books, and search results.

What seems like a manageable margin suddenly starts growing and destroying everything.

I moved to a self-hosted Forgejo, with a self-hosted runner on another cheap Hetzner server. It's amazing how clean, snappy, and reliable it is.

Here's a guy who does more than make simple, bloated end-user apps. He tries to dig into the code and make novel performance optimizations etc. He sees that AI can't code everything.

> by writing N parsers (“readers”) and M renderers (“writers”), one could support N × M conversions.

Beautiful writeup for a wonderful project. In an age of vibe-coding hype it's also so nice to see how things can be extended and snowball in usefulness when things are built correctly, by hand, from basic principles.

> Perhaps, then, in the future, people will no longer have a need for tools like pandoc.

I think we will need wonderful things like pandoc more and more. As mentioned there is a huge ecological and practical difference. Even if LLMs could get infintisamally close to deterministic-level reliability, it's still so many more orders of magnitude better in efficiency, especially with big batch jobs etc.

One book probably doesn't do it. I know the LLMs have been trained on one particular book that explains some crucial things about the grammar of a low-resource language. The LLMs will sort of use the information, terms, and concepts in the book to explain things but they fail to understand and continue to confidently mangle the language and plagerize the content of that book.

It almost feels like alchemy.

> In the same way high-level languages are a layer on top of writing machine code manually.

Another layer, yes, but NOT "in the same way." It is a new, different kind of layer. The other layers were deterministic, not probabilistic.

This looks great. It seems crazy how we got to this point where this is a revolutionary, unique product.

But is the information really scanned and preserved for direct access? Or is it just trained into an LLM so that we can only get fuzzy answers about the information?

This is an excellent analogy that shows how LLMs are fundamentally different than linters, auto-complete, or transpilers.

Fair point, but I mean perhaps it's not fair for them to have to impose usage limits (which would limit everybody) when they could just make a decision to avoid certain types of projects (which they feel strongly about.)

They're not obligated to deal with the higher load brought on by people's LLM coded projects, CI pipelines etc. In order to keep things freed up and flexible for others, I think it's fair for them to say, "in this community, we're going to save the space for more hand-coded projects."

It's not fair for everyone to have to eat the costs of AI (like communities getting bulldozed for data centers) and just accept that it has to be used.

Making a principled ban against LLM would seem to allow for more freedom and flexibility with usage limits. People who are creating more by hand would still be allowed to have more breathing room. With a hard limit of resources everyone would be more restricted. That's also a solution, but it's just a totally different approach. The LLM ban could help to allow for freedom without the resources getting overwhelmed.

> In addition, Codeberg conflates “having a community” with “being legitimate software worth hosting”

I don't think this is a fair summary. I don't see this conflation in Codeberg's blog post. The words in quotation marks don't actually exist in Codeberg's blog post.

It seems that what they are saying is: In a free hosting platform where we share all the resources, it's not fair that some people using LLMs are using up disproportionately huge hosting resources to handle the unnaturally large amounts of resources they can churn out with these tools.

I'm a solo dev, and I couldn't find the judgement against solo developers or any slight against people who don't have a team. The concern is that some people don't gobble up an unfair amount of resources.

I personally have been moving to self-hosted Foregejo for ultimate freedom. But I also think it's fair for some providers have the freedom to take a bit more of a harsh (or based - depending on your perspective) stance against LLM generated code. I mean, a lot of people have been bemoaning the fact that the human-generated material has been increasingly hard to find in a sea of AI-generated material. It could be nice to have a corner of curated, human-made code still available on the web.

I've seen some extremely unnatural ones, like börek that looks overstuffed or Afghan Kabuli Palau with what looked like chocolate squares instead of meat chunks! It feels like looking at sci-fi horror.

The craziest offense I have ever seen was (in Germany) an add for a Turkish Kokoreç sandwich where they had a picture of the entire spit of the kokereç being cooked, unchopped, sitting in a bun!!

Exactly. I've been trying to say this. Some people think that because people can write CRUD APIs and end-user apps with AI, no-one will need to write code anymore.

There is so much to be built in terms of tooling and libraries where we need tight, solid, well-crafted, original, unique code.

It's sad now to see a university library where students sit among endless shelves of amazing books, sitting on laptops, almost all of them with ChatGPT open. The library has become just a place to sit and open an LLM.

And then the AI gets trained on the advice and information subreddits, and we see an exponential growth downwards in quality?

I see people talking about the negative experiences, and I'm sure they are real, but I also had some very positive experiences.

I remember being amazed that some internet stranger took the time to understand my question and provide some great solution where I was stuck. It gave me a little encouragement and social connection as I plugged away on a project alone.

I miss that feel good human connection when an LLM can quickly get me an answer.

The book takes you through simpler examples, but I found that the exercises did an excellent job of drilling a kind of recursive problem solving skill into your head. You represent the data in types, and then keep breaking the problem down into smaller functions until it is solved. It's almost magic. After being armed with this way of thinking (through lots of practice with the many exercises) I found that problems I didn't know where to begin with just melted away.

Natural language phrase generation in a very complex foreign language that follows recursive syntax trees. And making a GUI to handle and edit these syntax trees visually.

How to Design Programs with Racket was life changing.

https://htdp.org

That helped me to think about recursion, functional programming, and type driven development. After going through HTDP I was able to breeze through complex problems that were unsolvable before.

> Using LLM is like hiring an unreliable translator when I already know the language.

Fantastic analogy. And when you don't really know the language that well you're just amazed at the translator, like "wow you can speak really well!"

Yes! People judge the "usefulness" of the AI generated material by how much they can get away with using it themselves, now. But what happens when more and more is built on top of these generated falsehoods. The errors and chaos bubbles up exponentially.

I remember when I was just learning how to code and making some web app, I had to do a lot more blind guessing and running. "Ok let me try this... Will that work now?" I remember staying up really late, feeling stuck to the computer in that slot machine mode.

Then when I learned more I got less and less of that guessing feeling. I understood what I was building and what would work, I began using typed languages and could keep on track with the compiler/LSP. This brought me more into a satisfying flow state, and I had less of that addicting "wait let me see if this will work" magic.

It seems like coding with Claude etc is a lot like a trip back to the guesswork stage, and I don't want to go back there.

(Sometimes, when I'm doing some dev-opsy type stuff of stringing a bunch of messy components together or working with a pile of complex APIs, I can feel myself back in the blind guessing territory, and incidentally this is where I find a chat with an LLM most helpful.)

I really wish PWAs were more well-known among average users. If people knew and expected they could install certain websites as apps, that would simplify things so much and really balance the power of the app stores.

All those things (handsaws, tracksaws, CNC saws) are analogous to deterministic, clearly designed tools (assemblers, compilers, language servers). The jump to LLMs is a jump to a completely different category.

Thanks for the answer. From what I can see, most people who are enthusiastic and optimistic about AI use are producing end-user apps, internal tools (limited use) or just hobby libraries. The fuzziness may be tolerable on the edges of immediate use.

Creating something that is solid enough for widespread, reliable building is just in another category. And I wish people recognized this distinction more when they say we don't need to look at code anymore.

I still think internal tools are kind of in a category closer to end-user applications. They are limited in how widely they are used, and therefore they don't have to be as solid as say libraries used by millions.