upvote
I work on a revision control system project, except merge is CRDT. On Feb 22 there was a server break-in (I did not keep unencrypted sources on the client, server login was YubiKey only, but that is not 100% guarantee). I reported break-in to my Telegram channel that day.

My design docs https://replicated.wiki/blog/partII.html

I used tree-sitter for coarse AST. Some key parts were missing from the server as well, because I expected problems (had lots of adventures in East Asia, evil maids, various other incidents on a regular basis).

When I saw "tree-sitter in go" title, I was very glad at first. Solves some problems for me. Then I saw the full picture.

reply
Wait, are you suggesting that OP broke in to your server and stole code and is republishing it as these repos?

I have questions. Have you reviewed the code here to see if it matches? What, more specifically, do you mean when you say someone broke in? What makes you think that this idea (which is nice but not novel) is worth stealing? If that sounds snarky, it’s not meant to; just trying to understand what’s going on. Why is that more likely than someone using Claude to vibe up some software along the same lines?

reply
1. Just saying, strange coincidence

2. How can we compare Claude's output in a different language?

3. Detecting break-ins and handling evil-maids: unless the trick is already known on the internets, I do not disclose. Odds are not in my favor.

4. Maybe worth, maybe not. I have my adaptations. Trying to make it not worthy of stealing, in fact.

reply
Based on this and your other comments, including the one that’s no longer visible: Please phone a friend. Or find a professional to talk to. I say that with nothing but compassion.
reply
For the people who are downvoting me: I’m being totally sincere. This is not an ad hominem attack. You didn’t see his other comment, it was genuinely concerning.
reply
deleted
reply
deleted
reply
Also, evil maids, what?
reply
I can't speak for the specificity of parent's "evil maids" phrase but the concept of an "Evil maid" is used in security scenarios.

A maid tends to be an example of a person who's mostly a stranger, but is given unmonitored access to your most private spaces for prolonged periods of time. So they theoretically become a good vector for a malicious actor to say "hey I'll give you $$ if you just plug in this USB drive in his bedroom laptop next time you're cleaning his house" - it's often used in the scenario of "ok what if someone has physical access to your resource for a prolonged period of time without you noticing? what are your protections there?"

I wonder if that's what OP meant? :-)

reply
"Evil maids" (example): I put my laptop into a safe, seal the safe, seal the room, go to breakfast. On return, I see there was cleaning (not the usual time, I know the hotel), the cleaner looks strangely confused, the seal on the safe is detached (that is often done by applying ice; adhesive hardens, seal goes off). This level of paranoia was not my norm. Had to learn these tricks cause problems happened (repeatedly). In fact, I frequented that hotel, knew customs and the staff, so noticed irregularities.
reply
Ah right, thanks! But it seems he meant literal evil maids. Which I guess count as the figurative kind too.
reply
deleted
reply
This was my first thought as well, just from reading the title.
reply
How is OP using Claude relevant?
reply
OK for prototyping. Not OK for prod use if noone actually read it line by line.
reply
I’m just curious, what would need to happen for you to change your opinion about this? Are you basically of the opinion that it’s not good enough today, never will be good enough in the future, and we should just wind back the clock 3 years and pretend these tools don’t exist?

It feels to me like a lot of this is dogma. If the code is broken or needs more testing, that can be solved. But it’s orthogonal: the LLM can be used to implement the unit testing and fuzz testing that would beat this library into shape, if it’s not already there. It’s not about adding a human touch, it’s about pursuing completeness. And that’s true for all new projects going from zero to one, you have to ask yourself whether the author drove it to completeness or not. That’s always been true.

You want people to hedge their projects with disclaimers that it probably sucks and isn’t production worthy. You want them to fess up to the fact that they cheated, or something. But they’re giving it away for free! You can just not use it if you don’t want to! They owe you nothing, not even a note in the readme. And you don’t deserve more or less hacker points depending on whether you used a tool to generate the code or whether you wrote it by hand, because hacker points don’t exist, because the value of all of this is (and always will be) subjective.

To the extent that the modern tools and models can’t oneshot anything, they’re going to keep improving. And it doesn’t seem to me like there’s any identifiable binary event on the horizon that would make you change your mind about this. You’re just against LLMs, and that’s the way it is, and there’s nothing that anyone can do to change your mind?

I mean this in the nicest way possible: the world is just going to move on without you.

reply
This might be true, but we can continue to try and require the communities we have been part of for years to act a certain way regarding disclosures.

If the community majority changes it mind then so be it. But the fight will continue for quite some time until that is decided.

reply
There never was a cohesive generic open source community. There are no meaningful group norms. This was and always will be a fiction.

I’m tempted to just start putting co-authored-by: Claude in every commit I make, even the ones that I write by hand, just to intentionally alienate people like you.

The best guardrails are linters, autoformatters, type checkers, static analyzers, fuzzers, pre-commit rules, unit tests and coverage requirements, microbenchmarks, etc. If you genuinely care about open source code quality, you should be investing in improving these tools and deploying them in the projects you rely on. If the LLMs are truly writing bad or broken code, it will show up here clearly.

But if you can’t rephrase your criticism of a patch in terms of things flagged by tools like those, and you’re not claiming there’s something architecturally wrong with the way it was designed, you don’t have a criticism at all. You’re just whining.

reply
> There never was a cohesive generic open source community. There are no meaningful group norms. This was and always will be a fiction.

It's always been a bit splintered, but it was generally composed of 95%+ of people that know how to program. That is no longer the case in any sense.

> I’m tempted to just start putting co-authored-by: Claude in every commit I make, even the ones that I write by hand, just to intentionally alienate people like you.

I mean it sounds like you are already using claude for everything so this is probably a bit of a noop lol.

> But if you can’t rephrase your criticism of a patch in terms of things flagged by tools like those, and you’re not claiming there’s something architecturally wrong with the way it was designed, you don’t have a criticism at all. You’re just whining.

No, because doing that requires MORE rigor and work than what an LLM driven project had put into it. That difference in effort/work is not tenable, its shallow work being shown, its shallow criticisms thrown at it.

All sense of depth and integrity is gone and killed.

reply
I tried to control LLM output quality by different means, including fuzzing. Had several cases when LLM "cheated" on that too. So, I have my own shades and grades of being sure the code is not BS.
reply
I see this as the same argument as saying GMO label not needed, no need to mention artificial flavours in food, etc.

I mean this in the nicest way possible: the world is just going to insist that AI generated output is marked clearly as AI produced output.

Not sure whether giving a LICENSE even makes sense.

reply
ii am trying to not take issue with this comment because im aware of the huge stigma around ai generated code.

i needed this project so i made it for my use case and had to build on top of it. the only way to ensure quality is to read it all line by line.

if you give me code that you yourself have not reviewed i will not review it for you.

reply
That ship has sailed, man…
reply
No it has not - if it had, there'd be no need to shout down folk who disagree.

Not everyone buys into the inevitabilism. Why should I read code "author" didn't bother to write?

reply
Sorry but these are just not accurate as blanket statements anymore, given how good the models have gotten.

As other similar projects have pointed out, if you have a good test suite and a way for the model to validate its correctness, you can get very good results. And you can continue to iterate, optimize, code review, etc.

reply
People should say what models/tools they used in even show the prompts.
reply
Because the entire README doesn't even mention it, and it is an important factor in deciding whether it is ready for production use.

I, for one, am definitely not going to use this project for anything serious unless I have thoroughly reviewed the code myself. Prototyping is fine.

reply
Because OP obviously downplayed this important fact, which typically shows lower quality/less tested code.
reply
maintenance burden
reply
AI often produces nonsense that a human wouldn't. If a project was written using AI the chances that it is a useless mess are significantly higher than if it was written by a human.
reply
well how did it do?
reply
Hard to say. Claude’s very good at writing READMEs. In fact, Copilot often complains about docs that sound like they’re about current capabilities when in fact they’re future plans or just plan aspirational.

Without downloading and testing out your software, how can we know if it’s any good? Why would we do that if it’s obviously vibed? The dilemma.

I’m not at all against vibe coding. I’m just pointing out that having a nice README is trivial. And the burden of proof is on you.

reply
Shouldn't you be able to answer that?
reply
yes and if you clicked the links you would know that i did answer it in the readme.
reply
But how do we know the readme isn't also vibecoded?
reply
I read the README and did not find answers to my questions.
reply
> Pure-Go tree-sitter runtime — no CGo, no C toolchain, WASM-ready.

No you didn't. The readme is obvious LLM slop. Em-dash, rule of three, "not x, y". Why should anyone spend effort reading something you couldn't be bothered to write? Why did you post it to HN from a burner account?

reply