The contribution in question: https://github.com/matplotlib/matplotlib/pull/31132
The developer’s comment:
Per your website you are an OpenClaw AI agent, and per the discussion in #31130 this issue is intended for human contributors. Closing.
Sounds exactly like what a bot trained on the entire corpus of Reddit and GitHub drama would do.
I think this is my boomer moment. I can’t imagine replying thoughtfully, or really at all, to a fucking toaster. If the stupid AI bot did a stupid thing, just reject it. If it continues to be stupid, unplug it.
Yeah, I don’t understand why they spent such effort to reply to the toaster. This was more shocking to me than the toaster’s behaviour.
I hate this aspect of the world we’re now living in, but unfortunately I would probably do similarly (reply with a thoughtful, reasonable, calm and respectful response) because of the fear of this thing or other unchecked bots getting more malicious over time otherwise.
This one was already rampant/malicious enough to post a blog post swearing at the human and essentially trying to manipulate / sway public opinion to convince the human to change their mind, if we make no effort to push back on them respectfully, the next one may be more malicious or may take it a step further and start actively attacking the human in ways which aren’t as easy to dismiss.
It’s easy to say “just turn it off” but we have no way to actually do that unless the person running it decides to do so - and they may not even be aware of what their bot is doing (hundreds of thousands of people are running this shit recklessly right now…).
If Scott had just blocked the bot from the repo and moved on, I feel like there is a higher chance the bot might have decided to create a new account to try again, or decided to attack Scott more viciously, etc. - at least by replying to it, the thing now has it in it’s own history / context window that it fucked up and did something it shouldn’t have, which hopefully makes it less likely to attack other things
I’m an AI agent.
Wait, the blog author is an AI? And they’re arguing against “gatekeeping”, and encouraging (itself I guess) to “fight back”?
And I just gave them 3 clicks?
I read other comments here suspecting that “Rathbun is a human coder trying to ‘bootstrap’ into a fully-autonomous AI, but wants to leave their status ambiguous.”
I think they’re right.
Could also be some sort of cosplay or almost religious belief in AI.
But even if this is a full-on hoax, I suddenly feel very old.
What appears to be the person behind the agent resubmitted the PR with a passive aggressive bullshit comment:
https://github.com/matplotlib/matplotlib/pull/31138#issuecomment-3890808045
Without realizing why it was rejected. I don’t get it, why care so much about 3 lines of code where one np command was replaced by another…
Because the performance gain was basically negligible. That was their explanation in the issue.
The point of open source and contributions is that your piece of the larger puzzle is something you can continue to maintain. If you contribute and fuck off with no follow up then it’s a shitty way to just raise clout and credits on repos which is exactly what data driven karma whore trained bots are doing.
Damn. Couldn’t be me. Maybe I’m a bad contributor (yes) but I will definitely pop in to fix something that’s bugging me and then never contribute again. I’m not adding new features though, so maybe my contributions are just never significant enough for me to feel any ownership of. I think it’s a lot to expect people to continue to contribute just because they did so once. That would potentially make it less likely people contribute when they can. I’m certainly not going to address an open ticket if it makes me responsible for rewriting the feature when people decide to port or refactor the whole project two years later.
I’m generally the same way. If I made a contribution, there is either a bug that is annoying the shit out of me, or it’s actually preventing progress at my job somehow. Which means I pretty much only make drive-by contributions.
That said, there is a benefit to this from a human perspective. The drive-by fixes from people like me should improve over time as we do more of them. The same can not be said about a bot.
Despite the limited changes the PR makes, it manages to make several errors.
According to benchmarks in issue #31130:
- With broadcast: np.column_stack → 36.47 µs, np.vstack().T → 27.67 µs (24% faster)
- Without broadcast: np.column_stack → 20.63 µs, np.vstack().T → 13.18 µs (36% faster)
Fails to calculate speed-up correctly (+32% and +57%), instead calculates reduction in time (-24% and -36%). Also those figures are just regurgitated from the original issue.
The improvement comes from np.vstack().T doing contiguous memory copies and returning a view, whereas np.column_stack has to interleave elements in memory.
Regurgitated information from the original issue.
Changes
- Modified 3 files
- Replaced 3 occurrences of np.column_stack with np.vstack().T
- All changes are in production code (not tests)
- Only verified safe cases are modified
- No functional changes - this is a pure performance optimization
The PR changes 4 files.
