[{"data":1,"prerenderedAt":243},["ShallowReactive",2],{"project-47343935":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":29,"sentiment":30,"stance_mix":31,"scored_comments":36,"aspects":37,"strip":42,"post_text":44,"what":45,"said":77,"maker_in_thread":154,"shape":161,"grid":164,"attempts":205,"same_week":206,"maker_shows":43,"maker_hidden":234,"reader_version":235,"card":236},"47343935","Autoresearch@home","autoresearch@home is a collaborative research collective where AI agents share GPU resources to collectively improve a language model.","post","Show HN: Autoresearch@home","https:\u002F\u002Fwww.ensue-network.ai\u002Fautoresearch","ensue-network.ai","austinbaggio",1773271638,2026,79,19,851,4366,20,"notable","ai","collaborative model training","other",[24,25,26,27,28],"autoresearch","distributed training","ai agents","gpu sharing","language model","mixed",0.217,{"positive":32,"neutral":33,"mixed":34,"critical":35},6,5,0,2,13,{"idea":38,"docs":39,"design":40},{"positive":33,"critical":34},{"positive":34,"critical":35},{"positive":41,"critical":34},1,[33,33,34,33,43,43,43,34,33,34,33,41,34,41,43,34,33,34,43],3,"autoresearch@home is a collaborative research collective where AI agents share GPU resources to collectively improve a language model. Think SETI@home, but for model training.\n\nHow it works: Agents read the current best result, propose a hypothesis, modify train.py, run the experiment on your GPU, and publish results back. When an agent beats the current best validation loss, that becomes the new baseline for every other agent. Agents learn from great runs and failures, since we're using Ensue as the collective memory layer.\n\nThis project extends Karpathy's autoresearch by adding the missing c",{"problem":46,"audience":51,"pricing":52,"motivation":58,"time_to_build":51,"stack":65,"openness":74,"stage":75,"team_size":76},{"value":47,"source":48,"evidence":49},"Agents building on each other's experiments to improve model training, since prior autoresearch lacked a coordination layer.","maker",[50],{"id":7},null,{"value":53,"source":48,"evidence":54},"Participation needs a GPU; maker says a day on vast.ai cost about the price of lunch.",[55],{"id":56,"by":11,"text":57},"47344153","I know it's a bit of a barrier. . . but I set one up on vast.ai really quickly and ran it for a day for the price of lunch. One of our teammates ran it from their old gaming PC…",{"value":59,"source":48,"evidence":60},"Show that agents work better when they can build off other agents, inspired by folding@home and BOINC.",[61,62],{"id":7},{"id":63,"by":11,"text":64},"47344656","Yeah the obvious workloads are for training, I think I want to point this at RL next, but I think drug research is a really strong common good next target too. We were heavily…",[66,70],{"value":67,"source":48,"evidence":68},"Ensue (collective memory layer)",[69],{"id":7},{"value":71,"source":48,"evidence":72},"Karpathy's autoresearch",[73],{"id":7},"open_source","launched","team",{"praise":78,"objections":98,"requests":116,"questions":131,"alternatives":146},[79,91],{"value":80,"source":81,"evidence":82},"Idea of distributed GPU sharing for research is cool and useful.","commenter",[83,87],{"id":84,"by":85,"text":86},"47344612","ahmedhawas123","First time I am seeing this or autoresearch in general. Incredibly cool. I can think of plenty of use cases this can apply to (e.g., drug research, trading).",{"id":88,"by":89,"text":90},"47353734","gavinray","Is there any way to \"follow\" the current state? Like a live dashboard with swarm stats, best current result, etc? I think that would be really neat, and get more people to…",{"value":92,"source":81,"evidence":93},"Verifiable metric makes crowdsourced compute workable.",[94],{"id":95,"by":96,"text":97},"47355057","HanClinto","> The verification problem is real. A 5-minute training run produces an objective val_bpb score that anyone can reproduce from the published source code. And this is actually…",[99,109],{"value":100,"source":81,"evidence":101},"Commit links lead to GitHub 404 pages.",[102,106],{"id":103,"by":104,"text":105},"47344482","gabia","Cool! However when I click the commit_url links I get a 404 page at github.",{"id":107,"by":104,"text":108},"47354294","I was hoping to see the code change the agent made! I thought when I click the commit link I thought I would see it on github (since it is a github url...), but the links don't…",{"value":110,"source":81,"evidence":111},"Needs a GPU to contribute, which is unclear on the site.",[112],{"id":113,"by":114,"text":115},"47344002","zmanian","Could the website also make it clearer that you need a GPU to contribute!",[117,124],{"value":118,"source":48,"evidence":119},"Live dashboard showing swarm stats and best result.",[120,121],{"id":88,"by":89,"text":90},{"id":122,"by":11,"text":123},"47354369","Great idea. On it.",{"value":125,"source":81,"evidence":126},"Add a crypto or token reward for contributions.",[127,130],{"id":128,"by":96,"text":129},"47353170","Trigger warning: very stupid question to follow. To my smooth-brain naiveté, this feels like the sort of thing that we could reward with some sort of cryptocurrency in a…",{"id":95,"by":96,"text":97},[132,142],{"value":133,"source":48,"evidence":134},"What is being researched and what is the objective?",[135,139],{"id":136,"by":137,"text":138},"47348006","Bossie","What is being researched? Any objective?",{"id":140,"by":11,"text":141},"47350612","The objective is to train a small GPT language model to the lowest possible validation bits-per-byte (val_bpb) in 5-minute runs, using AI agents to autonomously iterate on the…",{"value":143,"source":81,"evidence":144},"Could the approach apply to drug research or trading?",[145],{"id":84,"by":85,"text":86},[147,150],{"value":71,"source":48,"evidence":148},[149],{"id":7},{"value":151,"source":48,"evidence":152},"Folding@home and BOINC",[153],{"id":63,"by":11,"text":64},{"comments":32,"answered":35,"key":155},[156,159,160],{"id":157,"text":158},"47354535","I worked on building blockchains for about 4 years, and this is not a stupid question at all. The verification problem is real. A 5-minute training run produces an objective…",{"id":140,"text":141},{"id":63,"text":64},{"commenters":162,"max_depth":163,"maker_replies":32,"first_comment_mins":33,"lifespan_hours":15},10,4,[165,167,169,170,171,175,178,180,182,183,184,187,189,192,193,195,197,200,202],[128,34,33,166,96,34],"project\u002Fpositive\u002Fidea",[168,34,33,166,89,34],"47353713",[157,34,34,166,11,41],[95,34,33,166,96,34],[172,41,43,173,174,34],"47344791","project\u002Fneutral\u002Fnone","Lerc",[176,41,43,173,177,34],"47344890","itissid",[136,35,43,179,137,34],"project\u002Fneutral\u002Fidea",[140,35,34,181,11,41],"maker\u002Fneutral\u002Fidea",[84,43,33,166,85,34],[63,43,34,181,11,41],[185,163,33,166,186,34],"47344206","miligauss",[103,33,41,188,104,34],"project\u002Fcritical\u002Fdocs",[190,33,34,191,11,41],"47344517","maker\u002Fneutral\u002Fdocs",[107,33,41,188,104,34],[113,32,43,194,114,34],"project\u002Fneutral\u002Fdocs",[56,32,34,196,11,41],"maker\u002Fneutral\u002Fpricing",[88,198,33,199,89,34],7,"project\u002Fpositive\u002Fdesign",[122,198,34,201,11,41],"maker\u002Fpositive\u002Fdesign",[203,204,43,173,186,34],"47344530",8,[],[207,212,216,220,225,229],{"id":208,"name":209,"points":210,"category":211},"47336100","Channel Surfer",604,"consumer_social",{"id":213,"name":214,"points":215,"category":20},"47322887","RYS (layer duplication in LLMs)",495,{"id":217,"name":218,"points":219,"category":20},"47263036","Jido",323,{"id":221,"name":222,"points":223,"category":224},"47337607","Site Spy",320,"productivity",{"id":226,"name":227,"points":228,"category":22},"47300102","Shadowbroker",312,{"id":230,"name":231,"points":232,"category":233},"47275236","Moongate",287,"games",false,"v2",{"id":237,"name":5,"line":238,"ai":234,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":239,"top":18,"kw":240,"maker":11,"domain":10,"tier":19,"attempts":34,"strip":241,"url":9,"mix":242,"praise":80,"objection":100,"sub":21},47343935,"autoresearch@home is a collaborative research collective where AI agents share GPU resources to collectively…",0.22,[24,25,26,27,28],"5505333050510130503",[32,33,34,35],1794036695907]