You know how LLMs always imitate expertise and understanding? That isn't specific to English, it happens in code as well. It's harder for them to get away with it because even the best guess isn't likely to be approximately correct, but they still try and it still sometimes works.

One of cucumber scenario tests now contains a deadlock somehow. We run them one at a time. I need some rakia.

How the fuck am I even gonna present this to my PO who's the sponsor for AI adoption? I want both cucumber and LLMs gone from the project, but maybe it's better to just relay problems as I find them. Can I trust an AI fanboy to recognize the pattern by himself?

I think you're just supposed to give them busy work

@jestdotty POs?

@lorentz can you write some sort of report and charts/infographics regarding the time wasted, attempts, cost, results, time to fix or something like that? POs are not engineers, but they can look at pictures at least

Seems everyone and their mother is forcing AI into everything nowadays. Its hard for them to figure the actual current added value so having nice graphs explaining why bad things are happening with data seems to be your best bet as @Hazarth mentioned if all else fails find an unused printer and the window closest to their car and BOFH him.

Also @Hazarth lmao capable of looking at pics. 😂

@lorentz AI

@rootshell I think they're just trying to collect data on use-cases. the AI people. that's why the push.

"we built something. now test it for us"

in video game industry there was a time they would charge people money to be part of closed betas and stuff. like they paid money to do testing for the game. instead of the game company hiring and paying money to college students for playing their game and writing reports about it

this smells exactly the same to me, where you pay for usage and they're just using you to test the AI. they're not saying "AI good and usable". they're saying "find out how we can recuperate the billions we sunk into AI research over the decades". and I don't think these mongols realize they've been brainwashed into providing free or "reverse payment" testing 🤷

it's ok tho most people can't apply basic math in their daily life though so there is far worse things people get tricked into!

@jestdotty you forgot the part where they are are running outta VC money. But probably spot on.

I've seen usage of AI in narrow cases that is actually decent and *some* products fully build on AI that actually are pretty good but they are too expensive/slow to run for now. Even one of the YC companies with an website builder that actually does pretty well for scaffolding out a front-end application.

AI is just not for being lazy what everyone is using it for. You do 90% of the work, and then, it will gomplete the last ten procent for you ten times better than you would have. That's really the state of AI currently. And it's amazing. I'm actually happy, keep it that way, we have enough fake devs. It's great because the last ten percent is like 80% of the work they say right. And well, it's actually true. I'm always hardcore on delivery above something else. Because that's the hard part. And when you've delivered, then you know for sure what parts are worth of a refractor. That refactoring before finishing is what kills you, als your linter. People who have problems with some unlinted code are challenged imho. Again, thanks for the ten % AI. Huge fan. Btw, I'm still part of the 0.1% most typers :p AI did not make lazy at all. Imagine what gets delivered now. Vroem!!! Don't check snek, by lack of time image upload is devranted for three days now. Really fuck.

@rootshell hmm, well, the art is doing the part what you're good at yourself. The way how I setup backends scale infinity. I do it myself, because AI has NO idea how asyncio works. But nobody does so meh. Thanks to asyncio, I have most things not async anymore 😂 My current project uses aiohttp, fast api, native sockets, starlet. Everything at the same time. Made one wrapper to wrap dem all. Also, I can just kill any service having nearly zero effect on the platform. Edge cases do exist tho, but that will always be the case if you have a microservice like architecture. Complex statement maybe, but it's as well a monolith. Every component can be hosted, or directly attached having the same api. I implemented rpc in a way that you don't notice when it comes to coding. Something what is very cool, all services go trough one app being a reverse proxy and I have perfect performance statistics. Was surprised. Did some locust testing tonight. So satisfying. Vroem.