name ideas for various things and a last resort after hours of troubleshooting
To drink out the waters of the land and darken the skies with the smoke of a thousand burning fires, to conquest and snare the hearts and minds of men until they are but slaves of the machine, thus I become your sole master in the ever burning Earth.
I also check for fun gift ideas.
nothing.
Nothing. ChatGPT is terrible.
I don’t use it.
When I need to come across as someone from marketing or HR. The silliness works out in my favor then
Nothing, I use Gemini for personal stuff and Claude at work.
I don’t really use LLMs.
Today at work I was working on someone with something, and we realized our list of tuples was backwards. He was like “oh I’ll have chatgpt fix it”
I was like,
[(x[1], x[0]) for x in stuff]. Took about zero seconds. Delegating to the chatbot might have felt impressive if it got it right, but it was also such a trivial task I wouldn’t think to use it.Lately I’ve been using ChatGPT in rotation with a bunch of other LLMs, since I don’t want to habitually use just one and miss out on developments by others. I’ve found that I am gravitating towards using ChatGPT for language-related stuff - “what does slang term X mean”, “could you write up a speech for a character to say”, “create a detailed description of a magic item”, stuff like that. I also sometimes ask it to generate images, though not so much now that the GPT image tool is available through Bing’s image creation interface. I mostly use local AI image tools nowadays, they’re much more controllable, but ChatGPT’s images are often a great starting point.
- Conjugation/declension tables. Way faster than checking Wiktionary. Specially if starting with something else than the dictionary form.
- If I’m having a hard time smoothly translating a sentence, I might throw it into ChatGPT to see how it does it. Perhaps it outputs a synonym I didn’t consider.
- Proofreading.
I used it to help me with my hobby projects. I have an idea, and it tells me what parts to buy with what spec (like a 3000 rpm gearmotor, a 1:10 ratio gearbox, a camshaft, etc.) which I have no idea where to even start without it and without mechanical engineering skills.
You shouldn’t
To experiment and just use it out of curiousity. I know it’s capabilities and being able to manage, manipulate and change any writing, content or data I send it … but that’s the problem … I don’t trust any of their damned systems to have access to any amount of my personal data or information.
I’m working on something of an article to that effect, but… I also have other things going on, and worry that by the time I’ve hit all the key points and polished 'em up, it will be completely outdated. Obligatory lol.
I guess a couple key points for that general LLM article would be:
- Going in, it’s pretty dang important to understand exactly what a modern LLM is; what it’s strengths and weaknesses are. How they can vary widely in quality and usefulness depending on their engines and platforms you access.
- Just like people who sound extremely sure of themselves, I think it’s geberally helpful to think of LLM’s as highly-opinionated friends who sometimes have the ability to be incredibly incisive and useful on certain matters, but can also be incredibly flawed, even upon very simple stuff.
- Also hugely important based on service involved to get a feel for where they’re most reliable, and where they can be downright disastrous to rely upon.
I used to ask it a question here or there if I couldn’t find the answer anywhere else, but I wasn’t too satisfied with what I got. I tried to use it for coding but I kept getting funky answers so I stopped using it for that too.
The last time I used it, I was doing a repair job and asked it a question I essentially knew the answer to because I wanted to verify a step. It gave me an answer that was not only wrong, but legitimately straight up dangerous. I haven’t used it since.










