Asked several to write a c implementation of some basic networking stuff.

ChatGPT: needed to refine my input, got reasonable output. Complete answers, just compile and run.

Google: the output was just a few snippets, nothing to be used as-is.

MSFT: terrible output, and -no suprise here- the compiled code crashed with null pointer references etc. The worst answers ever.

For simple problems (programming low-level microcontrollers), my go to will be ChatGPT everytime.

Google should get it’s act together, Microsoft can exit the stage.

There are no good LLMs.

I’m sure many don’t have the hardware to run local, but for most things that will probably work just as well as the full models, plus you can modify them and experiment. Start with Ollama as the base to run them, and see what works best. I tend to primarily use the edited uncensored versions of llama3 like the Neural Daredevil variations.

But just remember at any model’s base, even the biggest and best, they are at the core a predictor. This works great for some uses, not so well for others. Don’t use a screwdriver for a hammer…at least not until they merge them to be able to do both well.

No one mentioned Phind or Perplexity, both are niiice.

Llama3 local is pretty good

I’ve been using ChatGPT, specialized ones on Huggingface, and a bunch of local ones using ollama. A colleague who is into this deep says Claude is giving him best results.

Thing is, depends on the task. For coding, I’ve found all suck. ChatGPT gets you up to a point, then puts out completely wrong stuff. Gemini, Microsoft, and CodeWhisperer put out half-baked rubbish. If you don’t already know the domain, it will be frustrating finding the bugs.

For images, I’ve tried DALL-E for placeholder graphics. Problem is, if you change a single prompt element to refine the output, it will generate completely different images with no way to go back. Same with Adobe generators. Folks have recommended Stability for related images. Will be trying that next.

Most LLMs are just barely acceptable. Good for casual messing around, but I wouldn’t bet the business on any of them. Once the novelty wears off, and the CFOs tally up the costs, my prediction is a lot of these are going away.

I only use the ones I self host

Eliza

I do a lot of incredibly specific VHDL and 45GS02 asm, so the answer is none.

Even if I didn’t do obscure things with obscure languages, answer’d still be none, because I’d rather spend a few hours learning what the code does and how to use it, instead of “just hope the output runs” while not knowing what and why it’s trying to do what it’s doing.

like any other technology whatever is standard or convenient. Not super wild about them in relation to how the rest of the internet and technology has been going.