[{"data":1,"prerenderedAt":425},["ShallowReactive",2],{"project-47937380":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":35,"aspects":36,"strip":46,"post_text":47,"what":48,"said":67,"maker_in_thread":171,"shape":184,"grid":189,"attempts":387,"same_week":388,"maker_shows":41,"maker_hidden":417,"reader_version":418,"card":419},"47937380","Auto-Architecture","Karpathy's Loop, pointed at a CPU","title","Show HN: Auto-Architecture: Karpathy's Loop, pointed at a CPU","https:\u002F\u002Fgithub.com\u002FFeSens\u002Fauto-arch-tournament\u002Fblob\u002Fmain\u002Fdocs\u002Fauto-arch-tournament-blog-post.md","github.com","fesens",1777396350,2026,241,76,207,4366,5,"notable","hardware","cpu design automation","other",[24,25,26,27,28],"cpu design","fpga","risc-v","llm optimization","nextpnr","mixed",-0.062,{"positive":32,"neutral":33,"mixed":34,"critical":33},9,17,3,46,{"idea":37,"execution":40,"docs":42,"performance":44},{"positive":38,"critical":39},7,12,{"positive":41,"critical":34},1,{"positive":43,"critical":43},0,{"positive":41,"critical":45},2,[34,41,43,43,41,43,34,41,43,41,41,43,41,34,41,34,18,41,34,34,34,18,18,34,41,18,43,41,34,18,18,34,41,43,41,34,34,18,34,18],null,{"problem":47,"audience":47,"pricing":47,"motivation":47,"time_to_build":47,"stack":49,"openness":65,"stage":66,"team_size":65},[50,57,61],{"value":28,"source":51,"evidence":52},"commenter",[53],{"id":54,"by":55,"text":56},"47944680","bsder","> The frontier is the verifier. Um, yes? The big value that AMD had in the x86 market over competitors was their verification model. This has been known for decades. > 3-seed…",{"value":58,"source":51,"evidence":59},"Gowin GW2A-LV18 FPGA (Tang Nano 20K)",[60],{"id":54,"by":55,"text":56},{"value":62,"source":51,"evidence":63},"RISC-V",[64],{"id":54,"by":55,"text":56},"unknown","prototype",{"praise":68,"objections":88,"requests":122,"questions":139,"alternatives":158},[69,80],{"value":70,"source":51,"evidence":71},"Verification matters: the verifier is the key idea",[72,76],{"id":73,"by":74,"text":75},"47943822","sho_hn","Salient on the value of the verifier. Matches my experience in the last two quarters. Nice detail on the encountered failures. Very similar experiences with my own loops against…",{"id":77,"by":78,"text":79},"47945860","fy20","I used it (well, a skill based on the same idea) to optimise a prompt that does data extraction from UGC. However there isn't really a \"correct\" answer that's easy to define in…",{"value":81,"source":51,"evidence":82},"Impressive FPGA results",[83,87],{"id":84,"by":85,"text":86},"47944182","osti","> propose, implement, measure, keep the wins Pretty much what I did to let Codex with gpt5.4xhigh improve my fairly complex CUDA kernel which resulted in 20x throughput…",{"id":54,"by":55,"text":56},[89,100,107,118],{"value":90,"source":51,"evidence":91},"Not a novel idea; similar work exists (AlphaEvolve, openevolve)",[92,96],{"id":93,"by":94,"text":95},"47945421","thald","I was working some time ago on LLM assisted optimizations and algorithm discovery and this does not look like a novel idea. AlphaEvolve from google is evolutionary algorithm…",{"id":97,"by":98,"text":99},"47944376","beepdyboop","This is not a new idea at all, many many have had it, no one really can claim it",{"value":101,"source":102,"evidence":103},"Benchmark overfitting; results may not generalize","maker",[104],{"id":105,"by":11,"text":106},"47948784","Yeah, you are totally right. Its a work in progress, and the post was written by an LLM - Im trying to improve on it (dash pun intended). Regarding the benchmark overfitting,…",{"value":108,"source":51,"evidence":109},"Post written by an LLM, reads poorly",[110,114],{"id":111,"by":112,"text":113},"47943873","fc417fc802","Extremely interesting but I don't understand why it was written by an LLM. Either the frontier models are far better than I realized or else writing this document required a lot…",{"id":115,"by":116,"text":117},"47947264","vessenes","Yeah I find this current LLM voice very tiring to read; I get enough of it day-to-day wrangling claude and others. I don’t think ‘writing’ this took very much work though, it was…",{"value":119,"source":51,"evidence":120},"Gains exploit Gowin FPGA weaknesses",[121],{"id":54,"by":55,"text":56},[123,132],{"value":124,"source":51,"evidence":125},"Apply the loop to nextpnr and other FPGA vendors",[126,129],{"id":127,"by":55,"text":128},"47958195","Please apply this loop to nextpnr for any of the commodity Xilinx, Altera, or Lattice parts. For example, everything about Lattice has been stuck for almost a decade at this point.",{"id":130,"by":11,"text":131},"47947554","\"I would LOVE somebody to bounce AI off of reversing the architecture and bitstreams for the stupid-ass closed-source FPGAs.\" The only reason I'm using Gowin is because it has a…",{"value":133,"source":51,"evidence":134},"Replicate with specific hardware and benchmarks",[135],{"id":136,"by":137,"text":138},"47958219","mikewarot","I'm interesting in replicating the experiment, but I can't seem to find the actual hardware specifications. I assume this is something I could pick up from Digikey, Mouser, or…",[140,147],{"value":141,"source":51,"evidence":142},"Is this related to karpathy\u002Fautoresearch?",[143],{"id":144,"by":145,"text":146},"47944345","DeathArrow","Is this related to autoresearch? https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch",{"value":148,"source":51,"evidence":149},"Is it a genetic algorithm or hill climbing?",[150,154],{"id":151,"by":152,"text":153},"47949528","jstanley","That's not a genetic algorithm, that's stochastic gradient descent. To be a genetic algorithm it would need to have mutation (which you have here) and crossover (which you don't).",{"id":155,"by":156,"text":157},"47952996","bishopp92","It is rather a variation of hill climbing [1]. As others pointed out evolutionary algorithms employ a richer set of search operators. [1]…",[159,167],{"value":160,"source":51,"evidence":161},"Karpathy's autoresearch",[162,166],{"id":163,"by":164,"text":165},"47943844","pteetor","In case you are unfamiliar with Karpathy's Loop[1], it is a genetic algorithm[2] where the genetic \"mutations\" are clever-but-random ideas generated by an LLM agent, aimed at…",{"id":144,"by":145,"text":146},{"value":168,"source":51,"evidence":169},"AlphaEvolve \u002F openevolve",[170],{"id":93,"by":94,"text":95},{"comments":172,"answered":173,"key":174},13,4,[175,178,181],{"id":176,"text":177},"47948742","I genuinely laughed reading the first words. Yeah, its hard to be novel",{"id":179,"text":180},"47948722","For sure! The hypothesis generation gotta be improved. Your take on the \"least likely\" is interesting. In the beginning of the repo I was having problems with \"hypothesis…",{"id":182,"text":183},"47948641","Nice references! tks",{"commenters":185,"max_depth":186,"maker_replies":172,"first_comment_mins":187,"lifespan_hours":188},53,6,458,142,[190,192,196,197,201,204,206,209,213,216,219,221,224,226,227,230,233,234,237,240,243,246,249,251,253,255,259,261,264,267,271,275,278,281,282,285,289,290,292,294,296,297,299,302,305,307,310,313,316,319,321,323,326,329,331,334,335,337,339,342,346,348,350,353,355,357,359,360,361,364,366,369,373,375,378,381,385],[163,43,34,191,164,43],"project\u002Fneutral\u002Fidea",[193,43,41,194,195,43],"47944238","project\u002Fcritical\u002Fidea","2001zhaozhao",[97,43,41,194,98,43],[198,43,43,199,200,43],"47944990","other\u002Fneutral\u002Fnone","utopiah",[202,43,43,199,203,43],"47945113","janalsncm",[176,43,43,205,11,41],"maker\u002Fneutral\u002Fnone",[207,43,43,199,208,43],"47945865","wiseowise",[210,43,41,211,212,43],"47944590","project\u002Fcritical\u002Fexecution","naveen99",[214,43,41,194,215,43],"47944721","stingraycharles",[217,43,43,218,203,43],"47945129","commenter\u002Fneutral\u002Fnone",[220,43,43,218,215,43],"47945418",[222,43,34,191,223,43],"47944748","faangguyindia",[179,43,43,225,11,41],"maker\u002Fneutral\u002Fidea",[93,43,41,194,94,43],[228,43,41,194,229,43],"47946562","amelius",[231,43,43,199,232,43],"47946759","riedel",[182,43,43,205,11,41],[235,43,34,191,236,43],"47947576","zby",[238,43,41,194,239,43],"47945747","yk09123",[241,43,43,218,242,43],"47946094","KronisLV",[244,43,41,194,245,43],"47946830","2ndorderthought",[247,43,43,218,248,43],"47947698","ajmurmann",[250,43,43,218,245,43],"47947906",[252,43,43,225,11,41],"47948686",[254,43,41,211,245,43],"47946802",[256,43,34,257,258,43],"47945982","project\u002Fmixed\u002Fdocs","bodegajed",[260,43,41,194,229,43],"47946448",[262,43,41,194,263,43],"47946771","iterateoften",[265,43,41,194,266,43],"47946804","embedding-shape",[268,43,18,269,270,43],"47949487","project\u002Fpositive\u002Fidea","coredog64",[272,43,18,273,274,43],"47951708","project\u002Fpositive\u002Fexecution","8note",[276,43,43,199,277,43],"47950360","thrawa8387336",[279,43,43,199,280,43],"47951044","hbarka",[151,43,41,194,152,43],[283,43,34,191,284,43],"47951904","ks2048",[286,43,43,287,288,43],"47952667","commenter\u002Fneutral\u002Fidea","Legend2440",[155,43,34,191,156,43],[111,41,34,291,112,43],"maker\u002Fmixed\u002Fdocs",[115,41,34,293,116,43],"project\u002Fmixed\u002Fperformance",[105,41,43,295,11,41],"maker\u002Fneutral\u002Fperformance",[73,45,18,269,74,43],[298,45,43,205,11,41],"47949665",[300,45,18,269,301,43],"48010064","cadamsdotcom",[303,34,43,199,304,43],"47945462","mordae",[306,34,43,205,11,41],"47949649",[308,173,34,191,309,43],"47943928","thin_carapace",[311,173,41,194,312,43],"47944180","_carbyau_",[314,173,43,199,315,43],"47945369","regularfry",[317,18,18,269,318,43],"47948136","robviren",[320,18,34,191,266,43],"47948207",[322,18,43,199,274,43],"47952006",[324,186,34,191,325,43],"47943969","outside1234",[327,186,41,211,328,43],"47944127","dataviz1000",[330,186,34,191,74,43],"47943991",[332,186,43,199,333,43],"47945182","faeyanpiraat",[77,186,18,269,78,43],[144,38,34,336,145,43],"project\u002Fneutral\u002Fnone",[338,38,43,205,11,41],"47948814",[84,340,18,341,85,43],8,"project\u002Fpositive\u002Fperformance",[343,340,34,344,345,43],"47944202","project\u002Fneutral\u002Fperformance","hackyhacky",[347,340,34,344,85,43],"47944766",[54,32,41,349,55,43],"project\u002Fcritical\u002Fperformance",[351,32,43,218,352,43],"47944775","gbgarbeb",[354,32,43,225,11,41],"47947536",[356,32,43,218,116,43],"47947290",[358,32,41,349,55,43],"47957090",[130,32,43,336,11,41],[127,32,34,191,55,43],[362,32,34,191,363,43],"47948374","tverbeure",[365,32,43,341,11,41],"47949697",[136,367,34,368,137,43],10,"project\u002Fneutral\u002Fdocs",[370,371,18,269,372,43],"47945107",11,"Havoc",[374,371,43,225,11,41],"47947561",[376,39,34,191,377,43],"47946328","koonsolo",[379,172,34,191,380,43],"47947034","oidar",[382,383,18,269,384,43],"47951335",14,"anoop4bhat",[386,383,43,205,11,41],"47951379",[],[389,394,398,403,407,412],{"id":390,"name":391,"points":392,"category":393},"47920787","dirac",393,"dev_tools",{"id":395,"name":396,"points":397,"category":393},"47874647","Honker",327,{"id":399,"name":400,"points":401,"category":402},"47882697","Tolaria",318,"productivity",{"id":404,"name":405,"points":406,"category":402},"47899844","WUPHF",260,{"id":408,"name":409,"points":410,"category":411},"47886517","How LLMs Work",245,"learning_science",{"id":413,"name":414,"points":415,"category":416},"47876071","Turning a Gaussian Splat into a videogame",236,"games",false,"v2",{"id":420,"name":5,"line":6,"ai":417,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":421,"top":18,"kw":422,"maker":11,"domain":10,"tier":19,"attempts":43,"strip":423,"url":9,"mix":424,"praise":70,"objection":90,"sub":21},47937380,-0.06,[24,25,26,27,28],"3100103101101313513335531501355310133535",[32,33,34,33],1794036699859]