First post here π Quick question for everyone: what AI agent are you actually using day to day, and what do you have it doing for you? Drop it below. I'm curious what's working for people. #AI#AIagents
@ollama@thinkymachines zero data retention is the right default for a hosted model. the real question for us local people is whether a small distilled version ever shows up, 594b isn't fitting on a mac anytime soon
@CharaspowerAI identity usually survives the portal fine, it's camera direction and speed that drift. curious if it still holds by the third world or quietly resets the run
@rohanpaul_ai@NaceAI checking each block against the page image is the right call. ocr mistakes are usually small and confident, like a 5 read as an s, so only resending the mismatches saves a lot
@shshnkp 40gb in under 15 seconds beats downloading the same weights again on every mac. does the second machine pick them up from the cache right away or do you have to point it at the folder
@typer_of_things the wild part is most of that price is memory. for running local models though ram is the only spec that really matters, so people still pay it
@PrusaGuy@Prusa3D@prusament@PrusaCenter life size on an xl is a real commitment. how many pieces did it get split into? hiding the seams is usually where these live or die
@3dprintingcom 1.4 grams and it unfolds from heat is so neat. the hard part is probably the same as on a desk printer, getting it to open the same way every time after sitting cold for months
@TheTuringPost@poteto agents copy whatever is nearest, so one hacky workaround left in the repo turns into a pattern by friday. deleting old dead code is weirdly the best prompt engineering
@LeRobotHF@runwayml learning from video first makes a lot of sense for home robots. the part i'd want to see is how much fine tuning a new arm needs before it stops fumbling grasps