pull down to refresh

Am i doing AI wrong?

No. There's no wrong.

what exactly are they doing with AIs that are taking them on this rocketship to the moon?

The rocketship is bs. Devs are just happy that they don't have to write code anymore. And if you're really lazy and like inferior products, you don't have to test or read or anything. Just auto-ship all the slop automagically and let the users whine. That's what's going on, and if you've really never in your life had a contract that assigns you a nice fat penalty if you deliver crap, it's heaven.

But what will happen is that since this is actively victimizing users, there will be more contracts that penalize. And all the yolo code out there will eventually be ignored because as mediocre quality software becomes abundant, the bar for acceptable software rises exponentially. There is no excuse to deliver buggy shit if all you do is run some GPUs.

in my job, i would need something that could interact with amazon support and grind them down, the same way the useless support shitheads grind me down, but something like that needs an api and all sorts.

You only need the AI in runtime for cognitive tasks. So if you have a bunch of APIs that you need to connect to, connect them. Ask an AI to connect them for you (and test it, thoroughly) and it will be easier. I think the thing that is true about all the fomo is that the sky is the limit now. So while there is a lot of bs going around, what does help is thinking in opportunities to get rid of the drag on your day-to-day and exploring solutions that make your life easier, rather than staying stuck on the problem.

I'm almost getting some kind of paranoia that there's this master world of ai and I'm oblivious.

No. But do play.

what tips and tricks can you share?

Think about what's truly bothering you in your job, and analyze it. Can even co-work on that with an LLM. Let it search and analyze reports about similar issues to what you're experiencing. Ask it what the state of the art is in tools to improve your job.

But don't let anyone scare you. It's not scary. Just a little multi-tool.

Think about what's truly bothering you in your job, and analyze it. Can even co-work on that with an LLM. Let it search and analyze reports about similar issues to what you're experiencing. Ask it what the state of the art is in tools to improve your job.

I just did this actually. It put me on to Oblivion for managing and organizing my workflow. The learning curve was relatively simple and the the payoff was huge. Makes no sense pulling your hair out about a problem these days at least until you've conferred with LLM to try and work through possible solutions.

reply
39 sats \ 2 replies \ @optimism 2h

Claw 0 - Gemini Chat 1?

reply
116 sats \ 1 reply \ @billytheked 2h

There are some things I think there is still some value in working through, even if it's with the help of the LLM (I think I used perplexity)

Louis has been chilling. I try to be more careful after your warning.

reply
108 sats \ 0 replies \ @optimism 1h

One of my fav things to do now for knowledge or research questions is ask 2 LLMs concurrently: knowledge, (re)search.

For now my favorites for simple chatbot are gemini-3.1-pro-preview (very new but it's been extremely consistent for me thus far) and claude-opus-4-6-thinking (no-nonsense), and for search, gpt-5.2-search (thinks long, often gives very complete analysis) and grok-4.1-fast-search (somehow is capable of finding relevant results for niche, specific queries)

reply