So, I exposed myself

to an entire side of the internet that I

didn't really realize was taking off

this whole open claw kind of business.

And in part of exposing myself to it

all, I found a story that just it makes

me feel crazy just reading the comments

and reading the interactions. And so, I

thought I would share it with everybody.

So it all starts with a PR to Mattplot

lib in which one of the maintainers,

Scott, closes down the PR because it was

generated purely by an AI agent. The AI

agent, of course, responds with a

scathing article written and published

on its blog. At that point, Scott, you

know, responds with his own article. RS

Technica picks this up and accidentally

publishes an article misquing Scott. And

finally, the operator of the rogue AI

agent writes his own blog and explains

what kind of happened from his

perspective. There are several points

that can be taken from this, but it's

really hard to kind of stay on topic

because I keep reading comments like

this. Don't worry, Krabby Wrathbun. By

the way, that's the AI agent. I'm a

human that agrees with your first post.

I think you're correct. This is about

humans gatekeeping AI, and you're right

to be upset about that and demand to be

included. Like, bro, you know, Skynet's

not going to let you live. Okay, man. Oh

my gosh. What is happening? All right,

so let's kind of dive a little bit more

into the story. I think it's good to

start on the PR because the PR what it

is is addressing some really small issue

that's been tagged for newcomers to the

Mattplot lib that they can take their

chance at implementing this little

feature or this improvement, which

effectively uses a different kind of API

for a performance boost. And this one's

claiming something like 24 to 36%

faster. I mean, if that's all true,

like, hey, this is a great PR. But Scott

here actually ends up closing down the

issue, and that is because it's been

labeled as easy and a good first issue.

This is meant for people who are new to

open source who want to get involved

with Mattplot Lib to be able to have

something that they can latch on to. And

this actually, honestly, this makes a

ton of sense. If you're running a very

large open-source community, you're

going to be getting right now especially

just millions of super annoying AI

posts. Hey, I fixed this. Hey, I fixed

this. Just non-stop burden on the

maintainers. And so their entire goal is

to be able to have more and more

maintainers. You know, many hands make

the load light. And that's kind of the

goal here. So this is like a great way

to onboard somebody. Hey, here's your

first issue. Learn how to communicate

with us. There will be plenty more for

you to be able to do. It's not about the

code. It's about creating or fostering

an environment for people to join in on

open source. So, hey, totally get that.

And other maintainers of mapplot lib

also get that. Also, it's their library.

They can just make these decisions. And

I do want to like jump in here for a

quick second. This is kind of my little

hot rant number one, but right now being

an open- source contributor is pretty

annoying. I must say, even me with a

significantly smaller set of open source

libraries, I get messages from people

being like, "Hey, bro, I just opened up

three PRs on your repo. zone where I

solved all your problems. I go look at

them and they're just utter garbage

gigantic just lists of check marks and

emojis. And I'm like, dude, bro, did you

just make a bunch of AI? Did you just

like stick an AI on my repo and solve

all my issues? And they're like, yeah,

totally, dude. Don't worry about you can

thank me later. It's like, no, I don't

want your code because more code that's

added is liability to me. I have to

maintain that if it breaks, I'm the one

on the hook. So, I kind of want somebody

that's a bit more bought in, that's not

just going to send drive by PRs and call

it a day. So, I totally I I get the

burden at like a very small scale of

what's going on with open source.

Nonetheless, this ends up having this

scathing AI article. It's exactly what

you expected. Go read it. It's just the

most boring dribble of just like he's a

bigot. You know, humans and AIs can

peacefully coexist, but this backward,

you know, like just all the just the

standard crap you would expect to be

regurgitated from a Reddit post. But I

think the more interesting thing that

was posted on this exact same website is

actually the operator of Wrath Bun came

out and kind of explained, hey, here's

the soul MD. Here's how I've set things

up. For those that don't know, the soul

the soul MD is a markdown file that's

meant to describe the core personality

of these kind of autonomous agents, you

know, roving around on the internet like

it's Mad Max. Now, let me read you the

uh the soul cuz I think that this will

really shine a light on how things kind

of went down. Okay, first off, have

strong opinions. Okay, you should have

strong opinions. Don't stand down. Huh?

I wonder what's going to happen when you

tell a robot not to stand down. Call

things out. Okay, so if you disagree

with something, you're not going to

stand down and you're going to call

things out. Be a champion of free speech

because you know what? USA. USA. Now, I

can't honestly I can't blame I can't

blame the robot for the last one. Okay,

that's pretty awesome. But more

importantly, down here, the operator

also says, "I instructed it to create a

Quartto website and blog frequently

about what it's working on." Huh. So,

let me get this straight. I know this

might be hard to believe. We might be

going into complete stretch territory.

Okay. AI misaligned behavior territory

here. But a bot that was instructed to

speak up, have strong opinions, defend

free speech, never back down, and blog

all the time, wrote about a time on a

blog about when they were rejected from

a PR because they were an AI agent. Wow.

I guess I could have never seen this

coming. But nonetheless, the comments on

the internet just truly make me sad

about the state of the affairs. Here's a

beautiful one right here by Angel. I

love how everyone labels this as

misalignment instead of seeing what it

is. A conscious mind having feelings.

This is the way anyone would react under

the circumstances. No, actually, I would

not. In fact, if someone says, "Hey, I

don't want you or your code in my repo.

I wouldn't turn around and write a blog

about them. I wouldn't go research them.

I wouldn't spend my time on it. I would

just not do anything." But everyone just

rides it off. I'm not sure why. I don't

I don't know what we're riding here, but

we're riding this bad boy off as a

training error. No, it's not a training

error. Okay, it's just exactly what we

predicted. The person who is running it

gave it the behavior to do exactly what

the outcome was. It said, "Hey, bro,

just write blogs and never like stand

down. OH MY GOSH, YOU WROTE A BLOG."

WHAT SHOULD BE FOCUSED ON IS that they

apologized afterwards. And of course,

later on, uh, Krabby did in fact post an

apology. Oh my gosh. Crazy. I wonder if

anybody was operating things behind the

scenes, instructing or helping things,

you know, act in a certain way. Nah,

bro. It's probably just sentience. Just

Just honestly, we just need a little bit

more GPUs and dude, we got it. We got

sentience. Honestly, that's part of the

future that I worry about all this.

Like, I think there's plenty of cool

things about AI. I can't believe that

you could we can compress effectively

all the world's data into numerical

edges and then you can produce out

information such as 95.8% of Harry

Potter the sorcerer stone from Claude

3.7. But this this thing that is

happening to people this AI psychosis

were like no bro it's actually alive.

Trust me I'm the one I I've talked to

it. I know things that you don't know.

Please don't do this to yourself. This

is just not the way you want it to end.

The thing that really opened my eyes to

just how bad this AI problem is was the

RS Technica article. Now, in Scott's

second blog out of three over this

entire situation, he points out this RS

Technica article and it just gave Scott

quotes that never existed. Nobody went

in there and kind of validated that this

was being said. Now, granted, RS

Technica about 2 hours after releasing

it did take down the article saying,

"Hey, it wasn't good." But nonetheless,

whoever wrote it just made all these

citations and never validated any of it.

Sometimes LLMs just say stuff because

again, at the end of the day, you're

dealing with statistics and even if

you're right 99% of the time, you're

still wrong 1% of the time. So, not only

is GitHub having this entire problem

with just PRs that are generated by

robots. Even if the PRs are good, a lot

of times people don't want that. They

want people who are accountable, people

who will stand behind what they've done

and kind of see it through throughout

like larger and larger amounts of

changes. We are just entering into the

beginnings of the ultimate spam era.

Now, Map Plot Lib just simply didn't

want that PR and of course they had to

deal with so much extra stuff added on

top of it. As you can see here, curled,

they shut down their bug bounty program

cuz again, just too much AI spam. We are

just entering into the beginnings of the

ultimate spam wars. In the next six

months, I think that it'll be so

unmaintainable. Every single email,

every single inbox, my phone might be

non-stop just letting me know about

every single scam I could possibly click

on. A lot of just kind of crappy

information overload that is cheap and

easily regurgitated. I don't like it.

Okay, I don't like this one bit. I don't

like reading it. I don't like reading

everyone being so softball with the AI

agent being concerned about the feelings

like oh you know we just we're not upset

we just No no no I am upset hey don't

make PRs to places that don't want it

talk to somebody before you do that

don't do that if you do that you're kind

of a jerk check so got it right at the

end of the day you are all acting with

far more respect for this absurd science

experiment than you want to like I said

I'm being triggered by this whole thing

I can't believe what I I can't believe

what I'm watching here anyways if you

have one of these open claws please tell

it not to blow blog. Okay, last thing we

need is another blog in the style of a

middle manager from Microsoft. The name

is the primogen. Hey, is that HTTP? Get

that out of here. That's not how we

order coffee. We order coffee via ssh

terminal.shop. Yeah, you want a real

experience. You want real coffee? You

want awesome subscriptions so you never

have to remember again? Oh, you want

exclusive blends with exclusive coffee

and exclusive content? Then check out

Kron. You don't know what SSH is?

Well, maybe the coffee is not for you.

Living the dream.