I've currently taken over a vibe coded project. It's not horrible, but they're doing some shit wrong. It's a small codebase so whatever. I will say this, I've taken over some projects written by humans that were far bigger piles of shit than anything I've AI produce. And some of that shit was running $400 million a year companies on literal trash.

So that's why the execs get excited about AI, they've always been fine with shit.

Companies in fintech that prize velocity over security are not the companies who are going to let anyone do a good job. You wouldn’t even be able to save them from themselves. They value think-alike, which is often fatal.

https://www.coops.tech/ (uk)

https://tech-coops.xyz/

Fixing problems created by LLMs seems like a growth industry, as @JustinDerrick pointed out, and it can expose you to potential future employers and human networks if you use it for that. Also tool advice is going to be big as LLMs obsolete big chunks of enterprise.

Wrapping a co-op around a team can make sense for resiliency and one-stop service offerings.

We don’t allow AI near our software stack because our customers care about security. I would be open to using vulnerability discovery models, but so far my experience is that they’re really bad unless someone else is paying the bill and they’re guided by experienced exploit writers.

We don’t allow AI near our hardware except offline models (and then only because someone wanted to run an experiment, which proved to be a waste of time, finding ‘bugs’ of a class that the tooling made impossible by construction) because it’s confidential IP and none of the providers have sufficiently strong privacy policies to be acceptable for corporate use.

Our sales VP doesn’t use AI because his view is that his job is to make customers trust us. That requires personal connection and a track record of telling the truth, not bland text from a bullshit generator.

Our CEO and does seem to like generating marketing pictures with AI. I’m hoping to get him to stop at some point (hopefully soon we can find budget to hire a human to set fire to them and replace them with ones that aren’t terrible).

I am a few years retired from the tech industry and its shiny-new-thing hype cycle.

If most companies (let's be clear: most tech C-suites) have drunk the AI cool-aid, you can ride this out.

Soon enough the promised AI benefits will turn out to be operationally and/or financially negative in most situations. And then we will hopefully see the limited use cases where it actually adds value.

Make a change if you want to, for a better environment/compensation/whatever.

LLMs can be phenomenal at giving hints, granted.

But the reliability simply is lacking, so I have to verify each single step, which is sooo frustrating. So I found it too exhausting to use general LLMs.

Plus it degrades my own skill in a subversive way (use it or loose it).

@GossiTheDog

but not before the bubble pops

not sure when that will be

grocery stores also tend to do well during downturns as well, as everyone pivots from eating out.

seek out anti-AI folks to work with and for