From the article: engineers get to keep access
And of course the post picture is ai slop
Not even good slop. One of the dudes has two fucking arms on one side.
I feel like they explicitly prompted it to give that guy a fucked up arm, which is a nice touch if so
I don’t think so.
A lot of these websites only popped up to ride the hype/hate train to get clicks. Almost all are AI slop farms rehashing content for specific audience.
Ah, fair point, didn’t notice the suspicious qualities of the rest of the article
The whole article has been “improved” with a chatbot
I think it might be too late. Anthropic probably has all of their clients info already.
As recently as July, chief information officer Marco Argenti was talking up “vibe coding” by Goldman’s non-technical employees. He told Bloomberg that traders and bankers had built tools with AI and passed them to the firm’s engineers, of whom there are about 12,000, to be improved.
That arrangement let the people closest to a problem build a first version without waiting for engineering time. It also meant thousands of employees without engineering training were running software that writes and executes code, on the same machines they use to handle client information
That’s so stupid holy shit. I get AI adoption, but giving non dev/eng people privileges to write and execute code is so dumb.
That’s the whole point though. Companies don’t want software engineers, those cost money. If the business segment that needs the software can write it with plain English then that solves the need and saves them from paying engineers.
Right, but the more sensible ones still keep senior developers/engineers to manage the agents and act as the human-in-the-loop. They also know to keep the coding features to those people only. Like a marketing person that uses AI for presentations and graphics has no business writing, pushing, and running code. That’s also why fresh grads are getting the short end of the stick because companies are shifting hiring to senior/experienced people only.
Totally agree with the current state of LLM’s. They are sadly getting “good enough” that the code they shit out mostly works. I don’t know if, as we get close to the theoretical cap on LLM’s abilities, if that will hold.
I don’t see llm babysitting as a skilled career going forward if progress continues. Even when the LLM gets stuff wrong I’m still just using the model to discover why these days. I hate it.
In my company they’re trying to make it so only non-dev/eng people can write code, so they can let all of us developers go. Managers already vibe-coded a replacement for our whole system in secret, and we only found out by accident. So at least Goldman Sachs is ahead of many in learning from their own idiocy.
Before you go, you should helpfully vibe port and vibe refactor the original codebase to make it forwards-compatible with the new one.
You should vibe code a backup version management system too, and give an agent control over it. Prompt it with an extensive and detailed list of all the ways it could lose the original data and tell it to be sure never to do any of them. Unless the data got corrupted somehow. How would we know if the data’s corrupted? Running a suite of vibe-coded tests, of course.
Practice your shocked face for when the agent does an oopsie and accidentally overwrites the only pre-slop versions.
It would have to wipe 15 years of source control, which would be impressive even for a “rogue” agent. Trust me, I’ve turned over a lot of ideas in my head during angry moments, but source control, backups and audit trails are sometimes there to protect us from our worse impulses.
