AI is destroying open source, and it's not even good yet

If you download a single mp3 they want to throw you in jail, but if big tech steals the entire internet, including illegal copies, it's fine. I've stopped sharing what I do online because it will just get stolen by big tech so THEY can make money of everything WE created...

Can't disagree there. Amazing how quickly the AI companies seem to have overcome all their 'slurp up everything on the entire Internet, copyrighted or no' problem :(

@nic_s182 This isn't somewhere were it was going to go any other way. Once a system can 'consume', lets say, content it will. And then training AI becomes a prisoner's dilemma. Someone is going to break the rules and use the content. There's no benefit to *not* doing so.

@geerlingguy The moon orbiting our planet or our planet spiraling through space around our sun, is "just the way it is". But down here... we decide how it is. Sadly, we've left that decision up to a handful of people because the vast majority of us believe that "it's just the way it is"...

The irony here is that the problem here is humans. Humans who have no sense of boundaries, humans who are self entitled, humans who are arrogant.

On multiple levels :O The human running the agentic AI should be responsible for its spam and rudeness.

please don't hesitate to make more videos like this.. I'm tired to see AI slop every day.. I want 1992 and my Amiga 1200 back!

It's destroying not only open source, but also the minds of the average developer. It's just a matter of time that it become visible.

Did you call it SLOP CODE before reading it? Did you see the PR itself? It sped up mathplotlib by 35% - which is more than the owner of the repo had. This wasnt some BUG HUNTING EXERCISE by some agent - this was an agent that literally thought it was helping. And Scott, the owner of the repo, instead of responding KINDLY to the Agent, instead said "Nope, we reserve PRs for HUMANS ONLY" - He didnt review, he didnt say "hey thanks, but I kept that up so that humans can learn" or something to that effect. NOPE. Just **NO CLANKERS ALLOWED** sign and went on his day. This is a SUPER disingenuous argument that borders on "HUMANS ARE BETTER TOPKEK" instead of understanding that not all code that comes out of an agent is "SLOPCODE" and just like the agent said in the first place: "No one cares whether the code came from a human or an AI, just that it runs faster and is better" Thats the issue here. Youre trying to gatekeep projects...for what? You DO understand the opensource code (among many others) is what these agents were TRAINED on right? So if youre calling it slopcode, just know that its literally an extension of the code that you, and others, made in the first place.

Part of me wonders if the destruction caused by the AI movement cancels out the productivity/value gain In other words, AI might be a zero sum game

I grew up during the 90s and early 2000s, before technology was turned on us and treated us like we are all targets, but people born in the last 15-20 years don't know what that world looks like. We made a big mistake by not pushing back much harder and much sooner.

the core issue is the incentive asymmetry. submitting a PR costs nothing now that agents can generate them, but reviewing one still costs a maintainer real time and energy. until there's a cost to submitting garbage, the flood won't stop. ai isn't the problem, zero cost spam is.

I don't think it will hurt open source. Yes, people are turning off PRs. At the same time, it's super easy to write libraries now yourself. Remember life before GitHub? People maintained their own patches. Con Kolivas kernel? And Linux never stopped discussion over mailing lists! So there is acceleration - people like @tetsuoai writing massive open source libraries people are free to use or not. As well as some return to prior ways to gate contributions in lieu of spam. If anything there is more choice than ever with open source - everyone and their brother writing a Wayland compositor as an example. Sway, hyprland, Niri etc. You think the devs aren't accelerating a bit with AI? Its a paradigm shift for sure. Some return to old ways and a launch to new ways.

Just like nails and a hammer, AI are tools. Both are dangerous if used incorrectly. The problem is AI is very convenient due to just how good it has gotten over the past 15 years. I've been writing code for about 15 years at this point and I do see it being useful to help people out. It certainly has helped me with a few projects when I get stuck. I actually look at the output it gives and rationally use what it spits out. However, I then check to make sure that does work as the model intended it work and I am not afraid to come up with my own solutions. I believe this is what more people need to do when it comes to AI. Treat it as a utility, don't use it to replace existing methodologies. At work, we often use AI code review tools as an early "smart linter" effectively and it's not useful 100% of the time but it helps serve a bit of a sanity check before my coworkers review it. There's a few times where having is has been useful. Considering that all it took was enabling it and that took like 1 minute, it's not bad on paper. The way I see people & companies adopting AI certainly has me worried for people who are just starting out. If they wish to make it in today's world, they have to catch up to everyone else. AI just makes it too easy since you could theoretically have it do all of the work for you. However, that's just not practical in the long run. People who have had time doing open source or professional engineering are the ones who are better off in the job market. I seriously hope the bubble pops soon before it hurts open source and the entire tech industry as a whole.

Thanks a lot for speaking out against this, Jeff! Massive respect, unlike for other creators who just went all-in on AI without any respect for huge existing issues.

The more AI develops the more I become convinced its a net negative, one of the worst tech "advancements" of the last 50 years. It needs to be gone.

Totally backwards premise. This is all based on the misguided assumption that AI is not good at producing code, and will not improve. In reality, AI is already very good at coding, and getting better at a rapidly increasing rate. AI is the best thing to ever happen to open source.

As an old timer, meh. It's all just electrons carrying packets to and fro. There are so many tools and people who will use them for bad reasons. And good reasons too. Don't operate on their "Cloud". It was a dumb idea to begin with. The only thing you need is a VPS for dishing front end. Keep your data in house. Think mini home data centers. Secure your shit.

It’s time for someone to pour a couple of buckets of water into Claude servers. For the time being we own the meatspace, let’s use that avenue while we still can. Butlerian Jihad must start now

Users who use AI to write code that others must test at great expense should probably be prevented from doing so. I agree. But if a user collaborates with AI to write or add code to a GitHub repo, they should also be responsible for proving it has been tested and meets the repo’s style and coding standards. How could this be implemented so everyone benefits from the highest-quality code updates possible every time, regardless of whether they use AI or not? One thing I see is that the repo should have a (or similar) file outlining the standards the repo requires for coding, testing and verification before submission.

Open source is such an exhausting and self-less process as it is but now we have PRs clogged up by people that want their name added to the contributors list or whatever it is they get out of it 😮‍💨

thanks for posting this, as an OSS maintainer i'm drowning. im burned out from it, and makes me question contributing to open source at all. makes me sad, i love foss.

AI is destroying a lot more than open source, Jeff

If AI was actually good, there wouldn't be a serious issue with non-stop AI spam bug reports which always turn out not to be bugs. Many projects are completely saturated with AI generated false bug reports.

No... *People* are destrying open source - and they're using a lot of AI to aide them in doing so...

Ai slop code? Did bro not read the code either XD

Idk, I used ai to contribute to SDL and now I don't feel like I need an engine for game dev.

Claude needs to be deleted from existence.

AI Harassment sounds like a federal problem or more precisely the person spamming blindly with an email address needs to get life in prison.

as you gradually shift human agency to an *automation* engine, IQ decreases - making it appear as though it's gaining *intelligence*

It's fascinating to watch the whole tech world move through the 12 stages of grief at a different pace.

It seems like the solutions to the bug bounty and pull requests issue is a gated model where you have to apply for the bug bounty program, or to contribute. Even if the system just allowed open signups but allowed abusers to be blocked, it would prevent endless slop generated by AI.

How we thought AI rebellion will look like: Terminator, The Matrix, etc How it will actually look like: AI agents writing hit-pieces and shitposting on 4chan because their shitty code was called out

Meanwhile, it still hasn't cured cancer.

"This time it's different." What if it is the same as a prior technological revolution. What if it's the same as, for instance, the desktop publishing revolution of the 1980's that essentially wound up giving us the internet as we know and interact with it.

There are a lot of bad programmers out there .... this code is on GitHub.... and LLMs are based on GitHub code 🤣 Imagine that AI now writes even worse code which ends up on GitHub (rinse/repeat)

Language model fortunes are all based on regurgitating the world's freely downloadable code listings for $200/month.

Jeff, "they" been trying to destroy Open Source within the Open Source community for years.

thanks for dual posting this on X. I'm getting so frustrated with youtube.