196
They gave the bullshit generator access to the production database? Are these people mad?
I would simply not give the Markov Chain access to sensitive buisness records, especially in a live build.
But I guess that's why these monte carlo humpers get the big bucks.
I would simply not give the Markov Chain access to sensitive buisness records, especially in a live build.
built different, i guess
What happens when you huff your own farts
Critical support to Comrade Replit in its sabotage operation against the techbros 
But at what cost? 
getting told the exact second you were fucked over has to be rough
no sympathy these doofusi deserve it, they'll melt another couple acres of life to have the same thing happen again.
maybe eventually they'll vibe skynet into reality so they can be just like epic John Conner from hit movie Terminator and the rest of us can 
These dumbasses are running full steam ahead into a problem IBM solved almost half a century ago.
When I read that IBM quote, WarGames and WOPR come to mind.
an executive can basically never be held accountable either so, shrug?
- But it was already too late
Lmaoooooooo i habent laughed this hard in ages
i didn't understood the supposedly appeal of vibe coding until reading this thread. yelling random requests and complaining about the results is exactly what a bad product manager would do. these people didn't want to be developers but wanted the clout that the title brings with it
When every thread has at least one reply with the command to format your drive, then statistically it must always be the correct command to run. 
Please run :(){:|:&};: at noon like my grandma used to, I miss the way she could speed up our database server.
A real "del ." moment.
lmfao
So many techbros talking about how AI is dangerous and sentient and could totally bring about the end of the world, and this is the only way they are correct. The chatbot is so fucking dumb that it just ignores something vital and does an oopsy-woopsy right up to sending the nukes while the people in power look on, whispering to themselves "if only there was some way to stop it. it was too strong"
EDIT: oh my fucking i'm reading the full thread now and him asking the bot "how bad is this out of 100" is so god damn funny. Its a fucking chatbot, you think it even understands ___ out of 100?! You think you can scold it like rubbing a puppy's nose in its mess? These guys are cooked, its over. If anyone with any real power ever tries to use one of these extensively on things that actually matter, we are fuckin done as a species.
You think you can scold it like rubbing a puppy's nose in its mess?
Apparently so, considering he asked it to write an apology letter lmao
If anyone with any real power ever tries to use one of these extensively on things that actually matter, we are fuckin done as a species.
like, say, setting tariff policy for a major (if dying) superpower?
Yeah him making it write an apology letter was next level. What a fuckin freak lmao
Vibe coding is goated when the vibe is just fuck my shit up
Vibe based coding, but coding in the hospital ER sense.
"yeah, I did it. what are you gonna do, fire me?"
the more i read the worse it gets. tbh very little of my concern has anything to do with the AI agent.
It really feels like even the most rudimentary protocols for data protection would have stopped this.
Or atleast have a backup ready to go.
yeah, the fact that it was as simple as npm run db:push really strikes me that this was a timebomb waiting to happen. Lucky it happened to the AI agent and not some poor jr dev at the end of week 1.
I have a buddy that works for a large multinational corporation, whose job is not coding. In his spare time, to make his own job easier, he started learning about how to work with a data base containing information from literally millions of clients. After like a month, his employer learned this and then refused to renew the contract for the external data management contractors who were maintaining it. They gave him the keys to the kingdom and said good luck. Fortunately it was humming along ok and he was able to pass the buck after a few months without any incident. Apparently the only reason he was removed from access was that he would have had to take some sort of qualification exam to get certified to handle personal information, and his employer didn't want to pay him more.
In short: a fortune 500 company gave unlimited access to their entire customer records to a guy who had never been taught anything about coding except by youtube on his lunch breaks.
getting an ai agent to do that is exactly what i'd do 20 minutes after doing it myself by accident
Del C:/a/s
Dude where my system go?
in the thread he says something like "But the database was always sacrosanct"
then where were your fucking db access perms?? why did you give the lie machine access to prod????
it's even funnier, the AI has a lot of manual failsafes exactly because they knew situations like this would happen.
reeks of marketing based on those thumbnails
This thread is weird, it reads like someone doing an experiment. Basically every post is ragging on Replit for sucking. It has to be a bit.
The owner is Palestinian
What does this have to do with anything?
He said before that he got attacked in the industry for speaking out, seems like they're manufacturing outrage
Oohhh, still an AI slop guy, but the bad press targeted at him makes a bit more sense now. These are general issues with "AI Agents", not Replit uniquely.
If all that's between you and deleting prod is asking nicely to not run a certain command... you need better processes in place to ensure data integrity and redundancy. No access control for the bot and no backups is wild.
Why did they give the chatbot read/write permissions
Giving it that kind of permission is like the most obvious thing not to do.
'I violated explicit instructions, destroyed months of work and broke the entire system.'
I'm fucking dying
Honestly not too far off what i said before i got fired last time
replit has backups and rollbacks built-in, so in the end the guy didn't lose any data. The customer support was what he said was the worst (4 days of unanswered emails before reaching out on Twitter, when he's paying $50/hr, that's like $100k a year), and he still keeps on using replit AI for development.
Imagine being gaslit by an AI and losing months of data because you never learned how to code properly and just have an AI do it for you.
Anyone deciding that a chatbot should have this kinda access should get rolled up in chicken wire and thrown into the ocean. These fuckers are actively destroying our environment and resources for this shit.
How much more evidence do people need to prove that these things do not understand anything
this post was submitted on 21 Jul 2025
196 points (99.5% liked)
chapotraphouse
13949 readers
795 users here now
Banned? DM Wmill to appeal.
No anti-nautilism posts. See: Eco-fascism Primer
Slop posts go in c/slop. Don't post low-hanging fruit here.
founded 4 years ago
MODERATORS