Post #3806856

I see a quite a few of u being friendly with some people on here who I have strong reason to believe are (even if unintentional) assholes (some of them have been assholes to me directly) whenever they talk about AI/LLMs. I’m QTing this post from my old account just so that I’m upfront and honest about my LLM/AI beliefs and usage. I strongly enjoy vibecoding. See the QT for more detail. I have spent the last 11 years writing code. I believe I’m a fairly competent engineer. I truly think LLMs are useful tools, to me at least. Even outside of vibecoding, I believe LLMs are an incredible pedagogical tool and good at writing “busy code”, if used correctly. I pay $20 a month to OpenAI for codex, for the reasons listed in the QT. Feel free to unfollow or block me if you think that makes me a criminal of some kind. But more importantly, I have some ideas about “agentic coding” that I’m going to be building out soon (in open!). I have a strong suspicion that what’s runnable locally is already very good. I don’t have a rig capable of running one of the bigger models unfortunately, so I pay OpenAI. I want a future where I can run a free (as in freedom), local LLM on my laptop. One that doesn’t involve fucking over people, or the planet. I’m going to be putting in the work to see if this future is possible. I don’t want tokenmaxxing, fucked up companies like OpenAI or Anthropic to win. I don’t think it’s necessary to keep burning so much compute and keep building so many data centers. RE: https://social.treehouse.systems/@xrisk/116664763268719251