Elissa

Ron and I independently ended up doing a similar thing recently - spending a few weeks kicking the tires of various current LLM models for coding.

And we both came to basically the same conclusions. His post is so close to my feelings my thoughts on writing one... yeah, no, I'll just link to Ron's post instead.

https://www.grumpygamer.com/my_dinner_with_ai/

My Dinner With AI

Ron Gilbert's often incoherent and bitter ramblings about the Game Industry

Grumpy Gamer
Mar 19 at 3:56amWeb
Show threadBrian Swetland20h ago

@vampiress Strong agreement with his closing thought:
"I’m going to enjoy programming and not being a baby sitter to an AI programmer."

(I remain underwhelmed by these tools and even if they were much better, they still sort of entirely miss the point, so...)

Show threadBoyd Stephen Smith Jr.20h ago
@vampiress I'm just going to pretend this was my experience, so I don't have to ignore the moral, ethical, and environmental issues to do my own experiments. 
Show threadmahadevank16h ago

@vampiress

Nice read!

I still use it only in "chat mode" locally.

llama-cli + jan-code-4b for coding, qwen-3-4b for general stuff

No integration with editors and such, more out of privacy concerns than anything else.

I don't think AI is inherently bad, but the current custodians of the technology are horrible people and I do not want to support them.

Show threadAdam Kent16h ago
@vampiress Completely agree on train wreck points -- ethically, morally, financially, environmentally. I've been a massive sceptic. But having used a bit of Claude on C# and typescript and even terraform over the last couple months -- I'm starting to see why people are saying it's got some use cases.
Show threadBe Kinder Than Necessary12h ago

@vampiress

🤔 I wonder how hard it is to type things like:

"Setting aside the moral and ethical concerns around how these LLMs were built and trained . . ."

(from the comments to the article)