This is why we can't have nice things.AI is destroying open source, and it's not even good yet | 3:36
Jeff Geerling | 1.04M subscribers | 209,667 views | February 16, 2026
Posted on 02/18/2026 9:02:32 AM PST by SunkenCiv
This is why we can't have nice things.
AI is destroying open source, and it's not even good yet | 3:36
Jeff Geerling | 1.04M subscribers | 209,667 views | February 16, 2026
(Excerpt) Read more at youtube.com ...
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
YouTube transcript reformatted at textformatter.ai follows.
TranscriptOver the weekend, tech news site Ars Technica retracted an article because the AI a writer used hallucinated quotes from an open-source library maintainer. I'll leave some links if you want to know the details, but the irony here is the maintainer, Scott Schambaugh (and I'm sorry if I'm butchering your name here), he was harassed by someone's AI agent over not merging its AI slop code. It's likely that the bot was running through someone's local agentic AI instance running OpenClaw. And the guy who built that was just hired by OpenAI to "work on bringing agents to everyone." You'll have to forgive me if I'm just not feeling enthusiastic about that. It feels like every AI company is channeling Russ Hanneman vibes lately. "All the sudden I'm 22 years young, and I'm worth 1.2B." Last month, even before OpenClaw was released, curl maintainer Daniel Stenberg dropped bug bounties because AI slop resulted in actual useful vulnerability reports going down from 15% all the way to 5%.
And that's not the worst of it. The authors of these bug reports seem to have a more entitled attitude, despite not having done a lick of real work here. They try to inflate the security impact of the so-called bugs that their AIs encounter because they don't really care about curl, and they don't care about Daniel or other open source maintainers. They just want to grab quick cash using their private AI army.
I manage over 300 open-source projects, and while many of them are more niche than curl or matplotlib, I've seen my own increase in AI slop PRs. And it's gotten so bad, GitHub added a feature to disable pull requests entirely. Pull requests are like the fundamental thing that made GitHub popular, making it easy to contribute back to the software that we use and love. And now we're seeing that closed off.
The problem is, AI slop generation is getting easier, but it's not getting that much smarter; we've hit a plateau where code generation is pretty good. I mean, I used local open models to help me migrate my blog from Drupal to Hugo, and I admit, it's really helpful if you know what you're doing with it. But it's not really improving like it did the past few years, and it can't overcome human skill gaps. The problem is, the humans who review the code that are responsible for the useful software that keeps our systems going, they don't have infinite resources, unlike AI companies.
Some people say in response, like, "We should let AI take over code reviews, too," but that's not the answer here. If you're running a little weather dashboard or building a little toy server for your homelab, fine. But I'm not going to run my production apps, the things that actually make me money or could cause harm if they break, on unreviewed AI slop code. And for every story about this stuff that gets traction, there are a hundred more that don't get reported. If this was a problem already before OpenClaw's release, this hiring by OpenAI is going to make it worse by democratizing it even further.
Right now, this AI craze feels the same as the crypto and NFT booms, with the same signs of insane behavior and weird optimism. The only difference is, there's more useful purposes for LLMs and machine learning, so the scammers can point to those things as they bring down everything good in the name of their AI god. Since my last main channel video on "AI coming for us all," we now have hard drives as the new shortage to watch out for, as Western Digital just they've already sold through all their inventory for this year, and they're already selling through a lot for 2027 and 2028. There are some people out there believing the AI bubble isn't a bubble, but those people are misguided, just like the AI that hallucinated the quotes in that Ars Technica article. People are saying this time it's different, but it's not. The same signs are there from other crashes. The big question I have is, how many other things are we going to let AI companies destroy before they have to pay their dues? Until next time, I'm Jeff Geerling.
My security system stopped me from opening that link. It says it’s extremely dangerous.
AI vs AI [Mad Magazine callback]
Spies like us?
My opinion: China's approach is smarter. If you can get 95% of the capability without having to rebuild the entire power grid and build multi-billion dollar infrastructure, do it. I have yet to see something the big US models does that Chinese small models don't. And often the Chinese do things better (AI Video, as one big example).
Side benefit: China releases the small models as open source, and they can be - and are - run locally on normal PCs. Ironically, China is only place that offers people the ability to use AI locally, with privacy. People like that.
I said this but need to clarify that it's not true, by the way. OpenAI and Meta have released models too, good ones but they don't do it often and they are getting long in the tooth. Other western companies put out small models that are finetuned too, they tend to not be that great in my experience. China puts out new models all the time, pushing the state of the art. That's why I said only them, not really trye but they are the biggest, most innovative player in that domain.
I would think date of posting would be helpful to determine if something was 100% hand-crafted.
What Big Brother USA would love is to be able to identify the relatively few people that will cause big trouble.
That are lots of people around Tucson. One person snatched Nancy.
There are fields where getting pointed in the right direction can be useful such as policing, stock picking and military planning.
Medicine needs somewhat greater accuracy. If AI X-Ray Reader screws up when Dr. Scan Man wouldn’t, that’s a massive liability issue.
Many things have been automated by 100% human effort to a near maximal extent.
This whole “AI” thing is starting to have a tinge of the old ‘everyone will have a thin client PC’ days from Ellison and the 90s. They want you dependent on their network-centric computing. They don’t want you town your own independent machines that you can upgrade and change at will with your own vast collections of data and tools.
There is also the issue of cost.
Putting in $40,000 chips into mortar shells would be too much for Ukraine’s defense budget.
I would think the open source libraries would have Amazon type user feedback commenting, such as:
The module only generated the correct SQL 52% of the time for us.
“They want you dependent on their network-centric computing.”
Consider the case of Walmart automated checkout.
Walmart would like to ensure that a $100 item doesn’t go out the door with a UPC code of an item that is priced at $1.
That means visually coordinating the item with its scanned UPC.
That’s a lot of data going to the data centers.
Checkout systems sometimes have trouble handling relatively simple credit card transactions.
In the past Big Corporation was merely expected to worship at the AI Altar by Wall Street.
Wall Street is now wanting to see a greenback roadway.
So far, I started out rather liking Walmart’s automated checking because of not waing in a long checkout line, but now they seem to be reducing the number of machines, so it’s getting to be a mess again. If it weren’t for the often exorbitant cost at our regular grocers, I would quit going to Walmart for most things. Not hardly worth it any more.
Good catch. Passed it on to three other people.
At under 4 minutes, it’s a quick watch.
I don’t like that it associates each of my credit cards at checkout with my online account, my name, and my phone number.
A guy who came in just before me entered his phone number and didn’t scan or buy anything, then ran out the door and the dang auto checkout for me quickly showed it was going to send my paper receipt to him—AND DID - before I could hit cancel on the texted receipt. I couldn’t even print my paper receipt.
I had to get the manager to print my receipt and go through the security video to see what happened. She found out to disassociate my credit card from HIS cell phone number I had to enter my cell phone number and hit send at the screen that says it’s going to send it, by default. But I don’t want my cell phone number associated with any phone number or credit card.
Walmart self-checkouts SUCK!
The best would be AI v AI v AI....
Glad you liked it, and yup, Jeff’s concision here is easy to appreciate.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.