[{"data":1,"prerenderedAt":721},["ShallowReactive",2],{"project-48169874":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":49,"post_text":50,"what":51,"said":91,"maker_in_thread":245,"shape":258,"grid":262,"attempts":676,"same_week":684,"maker_shows":46,"maker_hidden":713,"reader_version":714,"card":715},"48169874","Semble","Code search for agents that uses 98% fewer tokens than grep","title","Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep","https:\u002F\u002Fgithub.com\u002FMinishLab\u002Fsemble","github.com","Bibabomas",1779032227,2026,445,151,64,4366,2,"legend","dev_tools","code search for agents","cli",[24,25,26,27,28],"code search","mcp server","semantic code search","bm25","claude code","curious",0.024,{"positive":32,"neutral":33,"mixed":34,"critical":35},16,43,3,14,76,{"performance":38,"execution":41,"idea":43,"docs":45,"design":47,"privacy":48},{"positive":39,"critical":40},4,6,{"positive":42,"critical":18},0,{"positive":44,"critical":44},5,{"positive":42,"critical":46},1,{"positive":42,"critical":42},{"positive":42,"critical":42},[46,34,42,46,44,34,42,42,42,34,46,42,34,34,34,39,44,44,34,44,42,44,39,46,34,46,34,34,34,46,46,39,34,34,34,34,46,34,39,39],"Hey HN! We (Stephan and Thomas) recently open-sourced Semble. We kept running into the same problem while using Claude Code on large codebases: when the agent can't find something directly, it falls back to grep, reading full files or launching subagents. This uses a lot of tokens, and often still misses the relevant code. There are existing tools for this, but they were either too slow to index on demand, needed API keys, or had poor retrieval quality.\n\nSemble is our solution for this. It combines static Model2Vec embeddings (using our latest static model: potion-code-16M) with BM25, fused vi",{"problem":52,"audience":58,"pricing":62,"motivation":63,"time_to_build":62,"stack":67,"openness":88,"stage":89,"team_size":90},{"value":53,"source":54,"evidence":55},"Agents fall back to grep and reading full files, using many tokens and often missing relevant code","maker",[56],{"id":57},"post",{"value":59,"source":54,"evidence":60},"Developers using coding agents like Claude Code on large codebases",[61],{"id":57},null,{"value":64,"source":54,"evidence":65},"Reduce token use and misses when agents search code with grep",[66],{"id":57},[68,80,84],{"value":69,"source":70,"evidence":71},"Python","commenter",[72,76],{"id":73,"by":74,"text":75},"48173229","vikeri","very curious to give it a spin but why write a cli in python? would surely be faster and more portable with go or rust?",{"id":77,"by":78,"text":79},"48177986","flossly","I feel this should be a mode on ripgrep. a disk search-lib in python? really? Burntsushi (author of ripgrep), please chime in!",{"value":81,"source":54,"evidence":82},"Model2Vec static embeddings (potion-code-16M)",[83],{"id":57},{"value":85,"source":54,"evidence":86},"BM25",[87],{"id":57},"open_source","launched","unknown",{"praise":92,"objections":111,"requests":157,"questions":198,"alternatives":217},[93,104],{"value":94,"source":70,"evidence":95},"Fast indexing and queries, good retrieval quality versus existing tools",[96,100],{"id":97,"by":98,"text":99},"48182190","bluejay2387","So about a year ago I wrote my own attempt at something like this using vector indexing and BM25 (the latest version uses CocoIndex, I had a custom coded solution using ChromaDB…",{"id":101,"by":102,"text":103},"48175025","aadishv","Seems like a cool idea so I decided to play with it a bit. The test I ran was in the browsercode (https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowsercode) repo with the following prompt:…",{"value":105,"source":70,"evidence":106},"Useful for humans as well as agents",[107],{"id":108,"by":109,"text":110},"48173163","singpolyma3","Semantic code search seems like a useful tool for a human too. Not just for agents.",[112,124,135,142],{"value":113,"source":70,"evidence":114},"Token savings may not translate to lower cost or faster end-to-end agent work",[115,119,123],{"id":116,"by":117,"text":118},"48174550","esperent","I did some evals with pi and GPT 5.5. I tested RTK on \u002F headroom on \u002F both on \u002F both off (all with the standard pi system instructions and no AGENTS.md). I forget the exact tests…",{"id":120,"by":121,"text":122},"48175406","freakynit","What I have personally observed with such tools is that they make the AI's dumb, similar to how it makes coders dumb when relying more on AI tools. These agentic AI's are already…",{"id":101,"by":102,"text":103},{"value":125,"source":70,"evidence":126},"Agents still fall back to ripgrep and do not trust results",[127,131],{"id":128,"by":129,"text":130},"48172599","jerezzprime","I'd be interested in seeing actual agent benchmarks (eg CC or Copilot CLI with grep removed and this tool instead). For example, I have explored RTK and various LSP…",{"id":132,"by":133,"text":134},"48176916","cityofdelusion","Feedback: codex-cli hangs when calling this through the MCP. The semble process even sticks around as a zombie, forever stalled out. No idea why, logs have nothing. When called…",{"value":136,"source":70,"evidence":137},"Probabilistic search may miss important results",[138],{"id":139,"by":140,"text":141},"48174555","onoesworkacct","fantastic token savings and performance... but unlike grep it's probabilistic search on search terms. is that an issue? the tiny model might not surface something important",{"value":143,"source":70,"evidence":144},"98% savings baseline is unclear",[145,149,153],{"id":146,"by":147,"text":148},"48176396","pu_pe","> uses 98% fewer tokens than grep So are we supposed to believe that grep is so wasteful that models are reading 98% useless garbage every time they call it? Either this claim is…",{"id":150,"by":151,"text":152},"48171482","ludicrousdispla","grep doesn't need tokens, so what is 98% fewer than zero?",{"id":154,"by":155,"text":156},"48176075","PufPufPuf","The savings are calculated against the assumption that agents read the matched files in their entirety. In my experience, they are smart enough to use grep to bring up a few…",[158,167,175,194],{"value":159,"source":70,"evidence":160},"Agent-level benchmarks and end-to-end savings",[161,162,166],{"id":128,"by":129,"text":130},{"id":163,"by":164,"text":165},"48176126","cagz","Would be nice to see the actual % token\u002Ftime saving across end-to-end coding sessions over these time periods https:\u002F\u002Fgithub.com\u002FMinishLab\u002Fsemble#savings My observation is that…",{"id":116,"by":117,"text":118},{"value":168,"source":70,"evidence":169},"Better install and usage docs",[170,174],{"id":171,"by":172,"text":173},"48175522","zhxiaoliang","The instructions on how to install and use it could use some work. I did eventually install it. Will try it later and report back.",{"id":132,"by":133,"text":134},{"value":176,"source":70,"evidence":177},"Comparison with other tools such as LSPs, colgrep, codebase-memory-mcp and cs",[178,182,186,190],{"id":179,"by":180,"text":181},"48173657","AussieWog93","Better than grep obviously, but how does this compare to existing LSPs?",{"id":183,"by":184,"text":185},"48173521","nextaccountic","How does this compare with colgrep? https:\u002F\u002Fgithub.com\u002Flightonai\u002Fnext-plaid\u002Ftree\u002Fmain\u002Fcolgrep",{"id":187,"by":188,"text":189},"48173704","_ink_","Would this replace something like codebase-memory-mcp[1] or improve when both is being used? [1] - https:\u002F\u002Fgithub.com\u002FDeusData\u002Fcodebase-memory-mcp",{"id":191,"by":192,"text":193},"48175368","boyter","Interesting. I too have been working in this space, though I took a different approach. Rather than building an index, I worked on making a \"smarter grep\" by offering search over…",{"value":195,"source":70,"evidence":196},"Fix for codex-cli MCP hang",[197],{"id":132,"by":133,"text":134},[199,206,210],{"value":200,"source":70,"evidence":201},"Does it support multiple languages?",[202],{"id":203,"by":204,"text":205},"48175065","gslepak","Does this support any language or is it limited to a specific set of languages?",{"value":207,"source":70,"evidence":208},"How does it compare with LSPs?",[209],{"id":179,"by":180,"text":181},{"value":211,"source":70,"evidence":212},"Security and supply-chain risk",[213],{"id":214,"by":215,"text":216},"48175878","handonam","The one thing I'm a bit nervous about: security. Thoughts of supply-chain \"what-ifs\" gives me a bit of pause here. Would like to hear security-minded folks give their thoughts on…",[218,223,228,234,238],{"value":219,"source":70,"evidence":220},"grep and ripgrep",[221,222],{"id":128,"by":129,"text":130},{"id":132,"by":133,"text":134},{"value":224,"source":70,"evidence":225},"codebase-memory-mcp",[226,227],{"id":120,"by":121,"text":122},{"id":187,"by":188,"text":189},{"value":229,"source":54,"evidence":230},"Serena",[231],{"id":232,"by":11,"text":233},"48175752","Serena does a lot more than semble (I actually used serena before building this and didn't like how much it does by default). That also made it hard to see if it was actually…",{"value":235,"source":70,"evidence":236},"colgrep",[237],{"id":183,"by":184,"text":185},{"value":239,"source":70,"evidence":240},"context-mode",[241],{"id":242,"by":243,"text":244},"48173270","smcleod","How does it compare to context-mode or serina that are both well established now?",{"comments":246,"answered":247,"key":248},22,11,[249,252,255],{"id":250,"text":251},"48175665","Hey, codebase-memory-mcp and semble are not exactly the same, but it's an interesting comparison, I'll put it on the todolist to check that out and add it to our benchmarks if…",{"id":253,"text":254},"48175792","Sorry to hear about the MCP integration, that's definitely something we'll look into. If you have any info about your system or how to reproduce it please let me know. Very nice…",{"id":256,"text":257},"48189691","Hey, this skepticism is fair and we share it, which is why we don't claim end-to-end agent improvements since we haven't measured those (yet). The benchmark we published measures…",{"commenters":259,"max_depth":40,"maker_replies":246,"first_comment_mins":260,"lifespan_hours":261},92,109,171,[263,265,267,270,272,276,280,283,286,289,291,294,296,299,300,302,305,307,309,312,314,316,320,323,325,328,331,334,336,339,342,344,347,350,353,355,358,361,364,367,369,370,373,376,379,381,383,386,389,391,393,396,399,402,405,408,410,413,415,418,420,423,427,429,431,434,437,441,444,448,450,452,454,456,459,463,466,468,471,474,476,478,479,481,486,490,494,498,500,502,505,507,510,513,515,517,520,522,526,528,530,532,534,536,540,542,544,545,549,552,554,556,558,561,564,567,569,571,575,578,580,585,587,591,593,597,599,603,605,610,614,616,618,622,624,628,629,631,634,637,640,642,645,649,651,653,657,661,665,669,671,674],[120,42,46,264,121,42],"project\u002Fcritical\u002Fperformance",[250,42,42,266,11,46],"maker\u002Fneutral\u002Fnone",[268,42,34,269,121,42],"48175777","project\u002Fmixed\u002Fexecution",[253,42,42,271,11,46],"maker\u002Fneutral\u002Fexecution",[273,42,34,274,275,42],"48179212","project\u002Fneutral\u002Fexecution","stavros",[277,42,42,278,279,42],"48176193","other\u002Fneutral\u002Fnone","parasti",[281,42,42,278,282,42],"48176713","spoiler",[284,42,42,278,285,42],"48181915","vorticalbox",[287,42,42,278,288,42],"48179282","zahlman",[290,42,42,278,121,42],"48180539",[292,42,42,278,293,42],"48182334","arcanemachiner",[295,42,42,278,121,42],"48188797",[297,42,46,264,298,42],"48181904","onlyrealcuzzo",[256,42,42,266,11,46],[301,42,42,278,298,42],"48195484",[303,42,42,278,304,42],"48205933","kurtextrem",[191,46,44,306,192,42],"project\u002Fpositive\u002Fidea",[308,46,42,266,11,46],"48175802",[310,46,42,278,311,42],"48177641","zikani_03",[313,46,42,278,192,42],"48187872",[128,18,34,315,129,42],"project\u002Fneutral\u002Fperformance",[317,18,34,318,319,42],"48172742","maker\u002Fneutral\u002Fperformance","stephantul",[321,18,42,278,322,42],"48173096","giancarlostoro",[324,18,42,278,184,42],"48173503",[326,18,42,278,327,42],"48173601","aleksiy123",[329,18,42,278,330,42],"48173915","maille",[332,18,42,278,333,42],"48174992","philipbjorge",[335,18,42,278,11,46],"48175808",[337,18,42,278,338,42],"48175866","oefrha",[340,18,42,278,341,42],"48175997","DeathArrow",[343,18,42,278,180,42],"48173663",[345,18,42,278,346,42],"48173687","gigatexal",[348,18,42,278,349,42],"48173893","yakbarber",[351,18,42,278,352,42],"48176586","turbine401",[354,18,42,318,11,46],"48175701",[356,18,42,278,357,42],"48175902","carlmr",[359,18,34,315,360,42],"48175951","Riany",[362,18,34,363,341,42],"48175974","project\u002Fneutral\u002Fidea",[365,18,34,363,366,42],"48194319","economyballoon",[368,18,42,278,275,42],"48179257",[116,34,46,264,117,42],[371,34,42,278,372,42],"48174613","zobzu",[374,34,42,278,375,42],"48174911","jack_pp",[377,34,42,378,319,42],"48175536","commenter\u002Fneutral\u002Fnone",[380,34,42,278,117,42],"48175743",[382,34,42,378,319,42],"48176101",[384,34,34,385,117,42],"48178671","project\u002Fneutral\u002Fdocs",[387,34,42,278,388,42],"48176108","forgotusername6",[390,34,42,278,304,42],"48205921",[101,39,34,392,102,42],"project\u002Fmixed\u002Fperformance",[394,39,42,395,319,42],"48175443","commenter\u002Fpositive\u002Fnone",[397,44,34,363,398,42],"48177252","adelks",[400,40,34,363,401,42],"48174924","andai",[403,40,42,278,404,42],"48175353","mackenney",[406,40,42,278,407,42],"48176464","Imanari",[409,40,42,278,401,42],"48180133",[411,40,34,412,319,42],"48175436","project\u002Fneutral\u002Fnone",[108,414,44,306,109,42],7,[416,414,34,417,319,42],"48175422","project\u002Fmixed\u002Fdesign",[179,419,34,363,180,42],8,[421,419,34,412,422,42],"48174181","CharlesW",[424,419,44,425,426,42],"48174927","project\u002Fpositive\u002Fperformance","cormacrelf",[428,419,42,378,422,42],"48175469",[430,419,44,425,426,42],"48176604",[432,419,44,425,433,42],"48177381","xeyownt",[435,419,34,266,436,42],"48259634","Runonthespot",[438,439,42,278,440,42],"48178226",9,"jelder",[442,443,42,278,304,42],"48205911",10,[445,247,42,446,447,42],"48173813","other\u002Fneutral\u002Fidea","wrxd",[449,247,42,378,11,46],"48175728",[187,451,34,363,188,42],12,[183,453,34,412,184,42],13,[455,453,42,378,319,42],"48175419",[457,35,34,363,458,42],"48173506","abcdefg12",[460,35,44,461,462,42],"48173555","project\u002Fpositive\u002Fnone","dopidopHN2",[464,35,42,278,465,42],"48173586","molszanski",[467,35,42,363,11,46],"48175733",[469,35,42,278,470,42],"48176279","sincerely",[472,35,42,278,473,42],"48179111","losvedir",[97,475,44,425,98,42],15,[477,475,42,395,11,46],"48189672",[163,32,34,315,164,42],[480,32,42,412,11,46],"48176165",[482,483,46,484,485,42],"48176348",17,"project\u002Fcritical\u002Fidea","sonink",[487,488,44,306,489,42],"48175502",18,"digitaltrees",[491,492,42,278,493,42],"48190303",19,"te0006",[495,496,44,306,497,42],"48177913",20,"luodaint",[146,499,46,264,147,42],21,[501,499,34,315,192,42],"48176452",[503,499,42,278,504,42],"48176626","pas",[506,499,42,378,11,46],"48177156",[508,499,42,278,509,42],"48181400","ac29",[511,246,34,315,512,42],"48181151","jrflo",[514,246,42,315,11,46],"48189698",[516,246,34,315,512,42],"48193247",[132,518,46,519,133,42],23,"project\u002Fcritical\u002Fexecution",[521,518,42,266,11,46],"48177188",[523,524,34,315,525,42],"48170972",24,"esafranchik",[527,524,34,315,319,42],"48171050",[529,524,34,315,525,42],"48171250",[531,524,34,315,319,42],"48171278",[533,524,34,315,525,42],"48171416",[535,524,34,315,319,42],"48171711",[537,538,34,363,539,42],"48171856",25,"mrpf1ster",[541,538,34,385,319,42],"48171927",[242,543,34,412,243,42],26,[232,543,42,363,11,46],[546,547,34,412,548,42],"48173726",27,"jahala",[550,547,42,278,551,42],"48173835","Escapade5160",[553,547,42,412,11,46],"48180350",[139,555,46,484,140,42],28,[557,555,34,363,319,42],"48175622",[171,559,46,560,172,42],29,"project\u002Fcritical\u002Fdocs",[562,559,34,563,319,42],"48175526","maker\u002Fneutral\u002Fdocs",[565,566,42,278,497,42],"48177905",30,[154,568,46,264,155,42],31,[570,568,42,378,11,46],"48176154",[572,573,44,461,574,42],"48176235",32,"florians",[214,576,34,577,215,42],33,"project\u002Fneutral\u002Fprivacy",[579,576,42,577,11,46],"48176196",[581,582,34,583,584,42],"48176203",34,"maker\u002Fneutral\u002Fidea","derrickrburns",[586,582,42,378,11,46],"48176246",[588,589,34,363,590,42],"48175739",35,"therealdrag0",[592,589,42,363,11,46],"48175780",[594,595,34,363,596,42],"48176621",36,"shermantanktop",[598,595,42,363,11,46],"48177204",[600,601,34,385,602,42],"48173477",37,"porker",[604,601,34,412,319,42],"48175399",[606,607,34,608,609,42],"48181447",38,"project\u002Fneutral\u002Fdesign","ssalka",[611,612,44,306,613,42],"48176929",39,"michal_lola2",[203,615,34,412,204,42],40,[617,615,34,412,319,42],"48175416",[619,620,46,264,621,42],"48180998",41,"cush",[623,620,42,461,11,46],"48189706",[625,626,44,461,627,42],"48178864",42,"paolomon",[150,33,46,484,151,42],[630,33,42,378,319,42],"48171680",[632,33,42,278,633,42],"48171990","djaboss",[635,33,42,636,319,42],"48172031","commenter\u002Fneutral\u002Fperformance",[638,33,46,484,639,42],"48175104","hparadiz",[641,33,42,378,319,42],"48175387",[643,33,42,378,644,42],"48178687","mrweasel",[646,647,44,461,648,42],"48174827",44,"jasonli0226",[77,650,46,519,78,42],45,[652,650,42,274,11,46],"48178169",[654,655,44,461,656,42],"48179113",46,"JasonBuildAI",[658,659,44,461,660,42],"48175026",47,"vemulasukrit",[662,663,46,484,664,42],"48175233",48,"ind-igo",[666,667,44,461,668,42],"48173950",49,"ramsono",[73,670,34,274,74,42],50,[672,670,42,378,673,42],"48173819","skeledrew",[675,670,34,274,319,42],"48175393",[677,680,683],{"id":678,"t":679,"points":414,"comments":42},"47910885",1777215627,{"id":681,"t":682,"points":419,"comments":42},"47997629",1777820528,{"id":4,"t":12,"points":14,"comments":15},[685,690,695,700,704,709],{"id":686,"name":687,"points":688,"category":689},"48179677","Files.md",730,"productivity",{"id":691,"name":692,"points":693,"category":694},"48192383","Forge",687,"ai",{"id":696,"name":697,"points":698,"category":699},"48191602","Gaussian Splat of a Strawberry",530,"creative",{"id":701,"name":702,"points":703,"category":689},"48215979","Phosphene",428,{"id":705,"name":706,"points":707,"category":708},"48178184","Auto-identity-remove",325,"security_privacy",{"id":710,"name":711,"points":712,"category":694},"48146369","whichllm",283,false,"v2",{"id":716,"name":5,"line":6,"ai":713,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":717,"top":18,"kw":718,"maker":11,"domain":10,"tier":19,"attempts":34,"strip":719,"url":9,"mix":720,"praise":94,"objection":113,"sub":21},48169874,0.02,[24,25,26,27,28],"1301530003103334553505413133311433331344",[32,33,34,35],1793991154444]