@sudoingX Where does cursor fit in the local LLM paradigm? I'm using local as my main starting point, Claude for dev work, haven't explored the local-drives-Claude idea yet
@nickreese Ugh it's absolutely terrible. I've often wondered if I could pipe LLM output through a secondary transformer to strip out the word vomit ๐คฎ
@wesbos I was thinking of this yesterday ๐คฏ you know how you look at search queries to your site to help with SEO? (been out the game a while, may not be applicable now?) I was thinking that about OSS agent searches, it could help me shore up docs, if I could get summaries
I'm working on a paper "Autonomous Skill Acquisition in Adversarial Networks" to show a better path forward for AI Agents. Very inspired by the GEPA paper (2507.19457) and some original research on memory categorization across disciplines.
@danielkhunter@whoiskatrin I built a chat tree app as a TUI with "blessed" because I didn't know Tauri enough ๐ is there a good template to start with?
@lucasmeijer Then it deletes planning docs and other task docs, capturing all lessons into ADR. Then I manually review and curate (often fully delete) the final ADR. Then feature branch merges into main nice and clean, but I had the benefit of the extra docs while it was going.
@GergelyOrosz@lumitech_co Good on you for blocking. All these AI replies seem to go with some sort of "parrot the OP in one sentence" prompt. I suspect someone is selling a "get more followers" bot/plugin that just idles doing this. Such a waste.
@sudoingX In the fight for local have you figured out a search that doesn't egress? Have been thinking recently about something like Kiwix piped into some vector search and an MCP interface, so that at least a good chunk of knowledge is fully local.
@sudoingX Still making my flow better, but usually several sessions open and work through them in a review phase, so slowdown doesn't change much for some workflows at least. The interrogation phase is where speed matters and that's been painful now today.
@sudoingX Just getting back to work after a week+ offline and wondered why Claude was so much slower than I remember. Any guesses on when the change happened?
My mental model of AI Agents shifted: I was about to make a 1-liner PR and realized AI tools = "double check against 5 docs, 3 internal APIs, and at least 1 external API" whereas my PR was "fleshy human memory and intuition". Uncomfortable, and also powerful feeling.
@DimitrisPapail To be clear: I've got no stake in arXiv, I'm not in academia, I just care about the quality of academic papers and think self-publishing of them (your directional argument) is not the solution to this increasing nightmare of LLM slop.
@DimitrisPapail GitHub is just a free, indiscriminate, content hosting service. Literally anyone, human or AI, can post "papers" of any kind, amazing or slop. If all you're looking for is a public hosting service then sure, use GitHub.
@inkdrop_app Hey Takuya, at my work we use 1Password Connect which runs in Docker, so it caches things locally. I just did a deep dive, happy to share details. I guess that's not for dev on the terminal though, it's standing up Docker container groups, so it might not help you?
@inkdrop_app I think it's human nature to hold on to old mental models. I've built a lot of amazing solid stuff with MongoDB over my career and people still parrot a 15 year old meme video for why it's a bad choice. Usually I just roll my eyes and ignore it. ๐คทโโ๏ธ