Post #2609445

Due to the ever present buzz in software development relating to LLMs and "agents" over the last while, I've occasionally downloaded and tested a local model to see what we're dealing with. Disclaimer: I'm not a fan of AI, it's creation is done in a highly destructive way and has many societal issues in tow, I just want to be real for a moment about reality. The models I've toyed with up to a few months ago were largely rambling broken messes that required all of my hardware to run at a snails pace. On occasion they'd spot something, but on average wasted all of my time. As of a few weeks ago, I was told Qwen3.6 released, and I won't lie, for something that can run on my own hardware... It's admittedly impressive. I'm not going to glaze it, it has many issues and has no ability to do large sweeping rewrites or architectural decisions, but they definitely are able to power an "agent" that looks through files and applies changes for small enough features or fixes. Back when I first messed with this new model a week or two ago I crashed out on my Discord server, hard. It's insane how quickly they went from useless to semi functional. ... Idk what to do with any of this...