This is bad.
-
@xgranade @ireneista Huh, maybe even https://brython.info/ ?
I thought that project had fizzled out, but no, it supports 3.14
@xgranade @ireneista Well, MicroPython has an advantage in that it's still written in C, and it is therefore possible to port C extension modules to it, though there aren't that many which really support it yet
-
This is bad. This is very, very bad.
I'm not trying to pick on Python here, I pick it because Python is something I'm actively using, and so I have a vested interest in the project *not* being AI-vulnerable.
But it's not good, chat. It's very far from good, in fact.
@xgranade
Huh, back to perl then I guess?
-
@SnoopJ @theorangetheme No, absolutely. I see this as the leading indicator rather than the damage itself, if that makes sense?
I keep using the term "AI-vulnerable" to try and point to that there isn't necessarily an actual direct impact, so much as a dramatically increased vulnerability surface area.
@xgranade @SnoopJ @theorangetheme I'm curious--how is Claude directly able to do commits? Why is it not "Claude on behalf of Dave Alvarado"? I understand somebody ran an agent against the code base, but someBODY ran the agent against the code base. Somebody prompted it saying "go find security vulnerabilities in Python".
It sure would be nice to know who, not just "Claude".
-
@ireneista @glyph I hope it doesn't, if only because I want to be focusing on my specfic and screenplays, but if it does come to that, I very very much so appreciate your support.
@xgranade @ireneista @glyph *quickly scribbles out a short story involving a fantastical run for the PSF*
-
@xgranade @SnoopJ @theorangetheme I'm curious--how is Claude directly able to do commits? Why is it not "Claude on behalf of Dave Alvarado"? I understand somebody ran an agent against the code base, but someBODY ran the agent against the code base. Somebody prompted it saying "go find security vulnerabilities in Python".
It sure would be nice to know who, not just "Claude".
@dave @xgranade @theorangetheme I'm not sure I really understand the question. In the commits above, it's a co-author rather than a primary author.
But in the general case, it's able to do it by running the command that adds a commit, in a context where the configured name/email for use with `git` will be the name/email associated with the model (the author metadata includes the specific model as well)
Creating such commits without indication of the human involvement (wherever it originated, since Rube Goldberg contraptions are all the rage right now) is IMO unethical but far from unimaginable.
-
@xgranade
Huh, back to perl then I guess?@srtcd424 If that's what's useful to you? But I don't personally recommend moving away from Python, nor do I think that's an effective tactic for dealing with the problem.
As mentioned, this is a broad problem in OSS *in general*, and Python is now in the blast radius of that problem. Trying to create a dependency path that doesn't include any AI-vulnerable code is very difficult right now.
-
@dave @xgranade @theorangetheme I'm not sure I really understand the question. In the commits above, it's a co-author rather than a primary author.
But in the general case, it's able to do it by running the command that adds a commit, in a context where the configured name/email for use with `git` will be the name/email associated with the model (the author metadata includes the specific model as well)
Creating such commits without indication of the human involvement (wherever it originated, since Rube Goldberg contraptions are all the rage right now) is IMO unethical but far from unimaginable.
@SnoopJ @xgranade @theorangetheme gotcha. On second look, I see that you were grepping, I misunderstood what I was reading there.
As I've thought about it some more, I think I'm standing by my take. IMO the fact that you contributed with Claude is barely more interesting than the fact that you contributed with VS Code. I think that "oh I used an LLM/Agent" is not a defense against, well, anything.
-
@SnoopJ @xgranade @theorangetheme gotcha. On second look, I see that you were grepping, I misunderstood what I was reading there.
As I've thought about it some more, I think I'm standing by my take. IMO the fact that you contributed with Claude is barely more interesting than the fact that you contributed with VS Code. I think that "oh I used an LLM/Agent" is not a defense against, well, anything.
@SnoopJ @xgranade @theorangetheme I don't think we should be personifying LLMs by calling them "co-authors". Claude didn't author, it recursively autocompleted.
-
@SnoopJ @xgranade @theorangetheme gotcha. On second look, I see that you were grepping, I misunderstood what I was reading there.
As I've thought about it some more, I think I'm standing by my take. IMO the fact that you contributed with Claude is barely more interesting than the fact that you contributed with VS Code. I think that "oh I used an LLM/Agent" is not a defense against, well, anything.
@dave @SnoopJ @theorangetheme It's not interesting, but it is important as part of understanding the vulnerability surface introduced by that code. There are many things about code that are simultaneously boring as fuck and also critically important.
-
@SnoopJ @xgranade @theorangetheme I don't think we should be personifying LLMs by calling them "co-authors". Claude didn't author, it recursively autocompleted.
@dave @SnoopJ @theorangetheme I don't even disagree, but that's the signal that Claude gives us, and there's no Git metadata for "this code was extruded by $x slop machine."
-
@xgranade oh, it looks like the warning is complicated?
https://github.com/python/cpython/commits?author=claude shows no commits
so idk what exactly the warning is saying
@astraluma @xgranade If you search for 'claude' you can find the commits where Claude is a "co-author" https://github.com/search?q=repo%3Apython%2Fcpython+claude&type=commits
-
@astraluma @xgranade If you search for 'claude' you can find the commits where Claude is a "co-author" https://github.com/search?q=repo%3Apython%2Fcpython+claude&type=commits
@nausicaa @astraluma As @joelle pointed out, Claude is also a name that real people have. @SnoopJ's cantrip is going to be less susceptible to false positives by filtering on "anthropic.com" as well.
SnoopJ (@SnoopJ@hachyderm.io)
@theorangetheme@en.osm.town @xgranade@wandering.shop here are the commits on `main` where it's explicitly a co-author: (Edit: I missed a few commits because I hadn't pulled :picardfacepalm:) ``` $ git log --oneline -i --grep "Co-authored-by: Claude.*anthropic\.com" 300de1e98ac gh-86519: Add prefixmatch APIs to the re module (GH-31137) ac8b5b68900 gh-143650: Fix importlib race condition on import failure (GH-143651) 9b8d59c136c gh-72798: Add mapping example to str.translate documentation (#144454) 34e5a63f145 gh-141444: Replace dead URL in urllib.robotparser example (GH-144443) 59f247e43bc gh-115952: Fix a potential virtual memory allocation denial of service in pickle (GH-119204) 5b1862bdd80 gh-87512: Fix `subprocess` using `timeout=` on Windows blocking with a large `input=` (GH-142058) cc6bc4c97f7 GH-134453: Fix subprocess memoryview input handling on POSIX (GH-134949) 532c37695d0 gh-137134: Update SQLite to 3.50.4 for binary releases (GH-137135) ```
Hachyderm.io (hachyderm.io)
-
@srtcd424 If that's what's useful to you? But I don't personally recommend moving away from Python, nor do I think that's an effective tactic for dealing with the problem.
As mentioned, this is a broad problem in OSS *in general*, and Python is now in the blast radius of that problem. Trying to create a dependency path that doesn't include any AI-vulnerable code is very difficult right now.
@xgranade
Yeah, sorry, it was dark humour. I'm honestly terrified about where all this heading Not personally a python fan probably due to my vintage but it's used for a frightening proportion of software I rely on. -
@dave @SnoopJ @theorangetheme It's not interesting, but it is important as part of understanding the vulnerability surface introduced by that code. There are many things about code that are simultaneously boring as fuck and also critically important.
@xgranade @SnoopJ @theorangetheme yeah I've been thinking about that, and I'm not sure I agree. The outputted code is the outputted code. "y = x + 1" doesn't gain additional attack surface because Claude autocompleted it.
I think there are all sorts of *human* exploits that can happen and are happening, but those are all based on our laziness checking Claude's work, not Claude's output itself. Things like maintainers going "Jesus take the wheel" when Claude writes commits because it's easier
-
@xgranade
Yeah, sorry, it was dark humour. I'm honestly terrified about where all this heading Not personally a python fan probably due to my vintage but it's used for a frightening proportion of software I rely on.@srtcd424 No need to apologize, I just want to be clear about my own views on this rather than inadvertently implying criticism of Python *in particular* that I neither mean nor want to make.
-
@xgranade @SnoopJ @theorangetheme yeah I've been thinking about that, and I'm not sure I agree. The outputted code is the outputted code. "y = x + 1" doesn't gain additional attack surface because Claude autocompleted it.
I think there are all sorts of *human* exploits that can happen and are happening, but those are all based on our laziness checking Claude's work, not Claude's output itself. Things like maintainers going "Jesus take the wheel" when Claude writes commits because it's easier
@xgranade @SnoopJ @theorangetheme please don't read any of this as my endorsement of slop, I can't stand it. I'm just trying to pick apart how code autocompleted by Claude is different from the moral hazard of trusting Claude in the first place.
-
@xgranade @SnoopJ @theorangetheme yeah I've been thinking about that, and I'm not sure I agree. The outputted code is the outputted code. "y = x + 1" doesn't gain additional attack surface because Claude autocompleted it.
I think there are all sorts of *human* exploits that can happen and are happening, but those are all based on our laziness checking Claude's work, not Claude's output itself. Things like maintainers going "Jesus take the wheel" when Claude writes commits because it's easier
@dave @SnoopJ @theorangetheme My views here are complicated, but let me try and give a somewhat accurate condensed version?
First, to your `y = x + 1` example, if the code is simple enough, that vulnerability can be mitigated by human review — the problem is still there, I contend, but was contained by review. The problem is that humans *suck* at scanning for that kind of problem. Take the TSA looking for guns in x-ray scans... they keep failing at that, and incredibly badly.
-
@srtcd424 No need to apologize, I just want to be clear about my own views on this rather than inadvertently implying criticism of Python *in particular* that I neither mean nor want to make.
@xgranade
Yeah, fair. It feels like we're fish trapped in a pool of trustworthy software that's rapidly drying up & shrinking -
@nausicaa @astraluma As @joelle pointed out, Claude is also a name that real people have. @SnoopJ's cantrip is going to be less susceptible to false positives by filtering on "anthropic.com" as well.
SnoopJ (@SnoopJ@hachyderm.io)
@theorangetheme@en.osm.town @xgranade@wandering.shop here are the commits on `main` where it's explicitly a co-author: (Edit: I missed a few commits because I hadn't pulled :picardfacepalm:) ``` $ git log --oneline -i --grep "Co-authored-by: Claude.*anthropic\.com" 300de1e98ac gh-86519: Add prefixmatch APIs to the re module (GH-31137) ac8b5b68900 gh-143650: Fix importlib race condition on import failure (GH-143651) 9b8d59c136c gh-72798: Add mapping example to str.translate documentation (#144454) 34e5a63f145 gh-141444: Replace dead URL in urllib.robotparser example (GH-144443) 59f247e43bc gh-115952: Fix a potential virtual memory allocation denial of service in pickle (GH-119204) 5b1862bdd80 gh-87512: Fix `subprocess` using `timeout=` on Windows blocking with a large `input=` (GH-142058) cc6bc4c97f7 GH-134453: Fix subprocess memoryview input handling on POSIX (GH-134949) 532c37695d0 gh-137134: Update SQLite to 3.50.4 for binary releases (GH-137135) ```
Hachyderm.io (hachyderm.io)
@xgranade @nausicaa @astraluma @SnoopJ
Also sometimes it's in the *commit message* that Claude helped, rather than in the user or first line of the commit, so --oneline probably isn't what you want either.
-
@dave @SnoopJ @theorangetheme My views here are complicated, but let me try and give a somewhat accurate condensed version?
First, to your `y = x + 1` example, if the code is simple enough, that vulnerability can be mitigated by human review — the problem is still there, I contend, but was contained by review. The problem is that humans *suck* at scanning for that kind of problem. Take the TSA looking for guns in x-ray scans... they keep failing at that, and incredibly badly.
@dave @SnoopJ @theorangetheme As code changes grow, it's even harder to do that mitigation, especially when those code changes interact with a highly complex code base. There's times where `y = x + 1` would be a catastrophic error due to someone else doing pointer math and whatnot, say.
Beyond that, though, it's not clear to what degree *if any* extruded code can be copyrighted. If it can't be, what impact does that have on the project.
