[{"data":1,"prerenderedAt":324},["ShallowReactive",2],{"project-46769752":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":27,"sentiment":28,"stance_mix":29,"scored_comments":34,"aspects":35,"strip":44,"post_text":46,"what":47,"said":60,"maker_in_thread":173,"shape":185,"grid":189,"attempts":286,"same_week":287,"maker_shows":33,"maker_hidden":316,"reader_version":317,"card":318},"46769752","TetrisBench","Gemini Flash reaches 66% win rate on Tetris against Opus","title","Show HN: TetrisBench – Gemini Flash reaches 66% win rate on Tetris against Opus","https:\u002F\u002Ftetrisbench.com\u002Ftetrisbench\u002F","tetrisbench.com","ykhli",1769452960,2026,111,41,611,4366,14,"notable","ai","llm benchmark","web_app",[21,24,25,26],"tetris","gemini flash","ai game playing","curious",-0.297,{"positive":30,"neutral":31,"mixed":32,"critical":33},0,13,1,6,20,{"execution":36,"design":37,"performance":38,"privacy":39,"docs":40,"pricing":41,"idea":42},{"positive":30,"critical":32},{"positive":30,"critical":30},{"positive":30,"critical":30},{"positive":30,"critical":32},{"positive":30,"critical":30},{"positive":30,"critical":30},{"positive":30,"critical":43},4,[45,32,30,30,45,30,30,30,45,32,30,45,45,30,30,30,30,45,45,32,32,30,30,30,45,45,45,30,30,30,45,30,32,30,30,30,45,32,45,45],3,null,{"problem":48,"audience":46,"pricing":46,"motivation":46,"time_to_build":46,"stack":58,"openness":59,"stage":59,"team_size":59},{"value":49,"source":50,"evidence":51},"Compares how well different LLMs play Tetris.","inferred",[52,54],{"id":53},"post",{"id":55,"by":56,"text":57},"46770702","burkaman","It's actually 80% against Opus, 66% average against the 5 models it's tested with.",[],"unknown",{"praise":61,"objections":74,"requests":116,"questions":136,"alternatives":153},[62],{"value":63,"source":64,"evidence":65},"Fun concept and useful feedback for the game mechanics","commenter",[66,70],{"id":67,"by":68,"text":69},"46770670","bubblesorting","Very cool! I am a good Tetris player (in the top 10% of players) and wanted to give brick yeeting against an LLM a spin. Some feedback: - Knowing the scoring system is helpful…",{"id":71,"by":72,"text":73},"46774471","kinduff","Stackflow looks nice! I'm a Balatro fan and I didn't know about this variant. By the way, kudos on your feedback. If I was OP, I would've been honored to get that type of…",[75,86,97,108],{"value":76,"source":64,"evidence":77},"Results don't show useful transfer to real tasks",[78,82],{"id":79,"by":80,"text":81},"46774677","augusteo","LLMs playing Tetris feels like testing a calculator's ability to write poetry. Interesting as a curiosity, but the results don't transfer to the tasks where these models actually…",{"id":83,"by":84,"text":85},"46772215","segmondy","... and what does this prove? what can you decide to use one LLM to solve over another based on this tetrisbench besides play tetris?",{"value":87,"source":64,"evidence":88},"Comparing against Stockfish or humans is not a fair test",[89,93],{"id":90,"by":91,"text":92},"46771396","gpm","Comparing against stockfish isn't fair. That's comparing against enormous amounts of compute spent experimenting with strategies, training neutral nets, etc. It will lose so…",{"id":94,"by":95,"text":96},"46776327","mikkupikku","It's not. Maybe if you used old versions of stockfish that predate the neural net methods used by current versions, because otherwise you'd be comparing the hand-rolled (by an…",{"value":98,"source":64,"evidence":99},"Tetris is solvable with simple algorithms, so LLMs add little",[100,104],{"id":101,"by":102,"text":103},"46772007","p0w3n3d","Guys, I don't know how to tell you but... Tetris can web solved without LLM...",{"id":105,"by":106,"text":107},"46776195","brookman64k","Why run something in a few CPU cyles on a 40 year old home computer when you can do the same (but worse) on a billion-dollar GPU cluster?",{"value":109,"source":64,"evidence":110},"Latency makes LLM real-time play questionable",[111,115],{"id":112,"by":113,"text":114},"46775529","dakom","How does it deal with latency? Afaict remote LLMs need seconds to process, but Tetris can move much faster..",{"id":79,"by":80,"text":81},[117,125,129],{"value":118,"source":64,"evidence":119},"Fix rotation, RNG and DAS to match standard Tetris",[120,121],{"id":67,"by":68,"text":69},{"id":122,"by":123,"text":124},"46771073","qsort","The worst thing is that the delayed auto shift is slightly off and it messes my finesse. (I used to play competitive tetris as well, but between getting older -> worse reflexes…",{"value":126,"source":64,"evidence":127},"Show scoring system and allow remappable keys",[128],{"id":67,"by":68,"text":69},{"value":130,"source":64,"evidence":131},"Explain how the models receive game state",[132],{"id":133,"by":134,"text":135},"46771367","vunderba","Interesting but frustratingly vague on details. How exactly are the models playing? Is it using some kind of PGN equivalent in Tetris that represents a on-going game, passing an…",[137,143],{"value":138,"source":64,"evidence":139},"How are models given the board state and how is latency handled?",[140,141,142],{"id":133,"by":134,"text":135},{"id":112,"by":113,"text":114},{"id":79,"by":80,"text":81},{"value":144,"source":64,"evidence":145},"Why Gemini Flash beats Opus and what drives the difference",[146,150],{"id":147,"by":148,"text":149},"46772446","indigodaddy","Is there a tl;dr on why this is? Does it just make faster decisions?",{"id":151,"by":11,"text":152},"46776027","my unvalidated theory is that this comes down to the coding model’s training objective: Tetris is fundamentally an optimization problem with delayed rewards. Some models seem to…",[154,160,167],{"value":155,"source":64,"evidence":156},"Stackflow",[157],{"id":158,"by":68,"text":159},"46771252","Hello fellow Tetris nerd with a -sort username :) On the topic of reflexes decaying (I'm getting there, in my late 30s): Have you played Stackflow? It's a number go up roguelite…",{"value":161,"source":64,"evidence":162},"Stacky",[163],{"id":164,"by":165,"text":166},"46774514","Keyframe","cool to hear about more randomizers. I am not a great tetris player but I absolutely love the game. I put NES, 7-bag and my own take on randomizer (crap) in my online tetris-like…",{"value":168,"source":64,"evidence":169},"Brian Moore's AI clocks",[170],{"id":171,"by":134,"text":172},"46772614","Thanks for the clarification! Kind of reminds me of the Brian Moore's AI clocks which uses several LLMs to continuously generate HTML\u002FCSS to create an analog clock for…",{"comments":33,"answered":174,"key":175},2,[176,179,182],{"id":177,"text":178},"46820005","Thanks so much for the amazing feedback!!! Will update the app to incorporate these",{"id":180,"text":181},"46771752","Thanks for all the questions! More details on how this works: - Each model starts with an initial optimization function for evaluating Tetris moves. - As the game progresses, the…",{"id":183,"text":184},"46771967","answered this in a comment above! It's not turn or visual layout based since LLMs are not trained that way. The representation is a JSON structure, but LLMs plug in algorithms…",{"commenters":186,"max_depth":43,"maker_replies":33,"first_comment_mins":187,"lifespan_hours":188},27,69,80,[190,192,194,196,197,200,201,203,205,207,211,215,217,220,221,223,225,228,231,233,235,237,238,240,242,246,249,253,254,256,257,259,262,266,269,272,274,278,280,281,282],[67,30,45,191,68,30],"project\u002Fmixed\u002Fexecution",[122,30,32,193,123,30],"project\u002Fcritical\u002Fexecution",[158,30,30,195,68,30],"commenter\u002Fneutral\u002Fnone",[71,30,30,195,72,30],[198,30,45,199,134,30],"46771309","project\u002Fneutral\u002Fdesign",[164,30,30,195,165,30],[177,30,30,202,11,32],"project\u002Fpositive\u002Fexecution",[180,32,30,204,11,32],"project\u002Fneutral\u002Fidea",[112,32,45,206,113,30],"project\u002Fneutral\u002Fperformance",[208,174,32,209,210,30],"46772832","project\u002Fcritical\u002Fprivacy","bityard",[212,45,30,213,214,30],"46770669","other\u002Fneutral\u002Fnone","OGEnthusiast",[133,43,45,216,134,30],"project\u002Fneutral\u002Fdocs",[218,43,45,206,219,30],"46771687","storystarling",[183,43,30,204,11,32],[171,43,30,222,134,30],"other\u002Fneutral\u002Fidea",[224,43,30,195,11,32],"46773101",[226,43,30,195,227,30],"46773153","mhh__",[229,43,45,230,219,30],"46777500","project\u002Fneutral\u002Fpricing",[55,232,45,206,56,30],5,[79,33,32,234,80,30],"project\u002Fcritical\u002Fidea",[101,236,32,234,102,30],7,[105,236,30,195,106,30],[239,236,30,195,102,30],"46778530",[241,236,30,213,91,30],"46782648",[243,244,45,204,245,30],"46771104",8,"esafak",[247,244,45,199,248,30],"46771312","bogtog",[250,251,45,204,252,30],"46770623",9,"akomtu",[90,251,30,195,91,30],[255,251,30,195,252,30],"46774805",[94,251,30,195,95,30],[258,251,45,204,134,30],"46771845",[260,251,30,261,11,32],"46773158","maker\u002Fpositive\u002Fnone",[263,264,32,234,265,30],"46770686",10,"arendtio",[267,264,30,213,268,30],"46771507","i_cannot_hack",[270,264,30,213,271,30],"46771945","westurner",[273,264,30,213,271,30],"46786714",[275,276,45,204,277,30],"46771669",11,"tiahura",[83,279,32,234,84,30],12,[147,31,45,204,148,30],[151,31,30,204,11,32],[283,18,45,284,285,30],"46772178","project\u002Fneutral\u002Fnone","purplecats",[],[288,292,297,302,306,311],{"id":289,"name":290,"points":291,"category":20},"46721802","isometric.nyc",1325,{"id":293,"name":294,"points":295,"category":296},"46793693","HN Arcade",352,"consumer_social",{"id":298,"name":299,"points":300,"category":301},"46779522","one-agent-one-browser",324,"dev_tools",{"id":303,"name":304,"points":305,"category":296},"46802254","Moltbook",287,{"id":307,"name":308,"points":309,"category":310},"46731432","Whosthere",278,"infrastructure",{"id":312,"name":313,"points":314,"category":315},"46753301","Bonsplit",245,"languages",false,"v2",{"id":319,"name":5,"line":6,"ai":316,"cat":20,"year":13,"points":14,"comments":15,"rec":27,"s":320,"top":18,"kw":321,"maker":11,"domain":10,"tier":19,"attempts":30,"strip":322,"url":9,"mix":323,"praise":63,"objection":76,"sub":21},46769752,-0.3,[21,24,25,26],"3100300031033000033110003330003010003133",[30,31,32,33],1793991141846]