[{"data":1,"prerenderedAt":546},["ShallowReactive",2],{"project-49524447":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":48,"post_text":49,"what":50,"said":83,"maker_in_thread":177,"shape":190,"grid":194,"attempts":504,"same_week":509,"maker_shows":44,"maker_hidden":538,"reader_version":539,"card":540},"49524447","slotstream","Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok\u002Fs","title","Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok\u002Fs","https:\u002F\u002Fgithub.com\u002Fcarloslfu\u002Fslotstream","github.com","carloslfu",1788280966,2026,240,118,210,4366,5,"notable","ai","local llm inference","cli",[24,25,26,27,28],"mlx","apple silicon","expert offloading","local llm","qwen","mixed",-0.339,{"positive":32,"neutral":33,"mixed":34,"critical":35},3,18,2,17,40,{"docs":38,"performance":41,"idea":43,"design":46,"execution":47},{"positive":39,"critical":40},0,6,{"positive":39,"critical":42},7,{"positive":44,"critical":45},1,4,{"positive":39,"critical":39},{"positive":39,"critical":39},[44,44,44,44,32,39,44,39,18,44,44,39,39,39,32,32,44,39,32,39,39,32,32,32,44,32,44,39,34,44,44,18,39,32,44,39,18,32,39,44],"I built slotstream, a way to run Qwen3.8-Flash-Next 4-bit on a low-memory mac starting from 16GB, a 125B parameter model that would need 100GB+ memory\u002FRAM, thanks to expert-offloading\u002Fssd-streaming. Easy to install\u002Fupdate, and mac-native using MLX and Swift.\n\nIt ships with auto-mode, which makes a good tradeoff between memory usage and speed. I'll be implementing and porting the MTP module for speculative decoding next",{"problem":51,"audience":57,"pricing":66,"motivation":67,"time_to_build":66,"stack":71,"openness":80,"stage":81,"team_size":82},{"value":52,"source":53,"evidence":54},"Running very large models on Macs with limited memory","maker",[55],{"id":56},"post",{"value":58,"source":59,"evidence":60},"Mac users with limited unified memory who want to run large local models","inferred",[61,62],{"id":56},{"id":63,"by":64,"text":65},"49536413","baristaGeek","I just used it, went through the whole installation (took like 1 hour approx). Long but straightforward. If I put my computer to sleep will it continue? I started the server…",null,{"value":68,"source":53,"evidence":69},"Make a large model run on low-memory Macs by streaming expert weights from SSD",[70],{"id":56},[72,76],{"value":73,"source":53,"evidence":74},"MLX",[75],{"id":56},{"value":77,"source":53,"evidence":78},"Swift",[79],{"id":56},"unknown","beta","solo",{"praise":84,"objections":97,"requests":131,"questions":152,"alternatives":164},[85],{"value":86,"source":87,"evidence":88},"Interest in local models running on constrained hardware","commenter",[89,93],{"id":90,"by":91,"text":92},"49525192","whartung","I'm hoping to see progress in this space. Folks talking about how 32G is not enough for local use, but then there's been work like this to empower it. My hope is that the new 32G…",{"id":94,"by":95,"text":96},"49527422","jacquesm","I love these efforts to get proper models running on lower cost hardware and I think this is where the next real breakthrough will come from. The more efficient this sort of…",[98,109,120],{"value":99,"source":87,"evidence":100},"Redundant with existing similar projects",[101,105],{"id":102,"by":103,"text":104},"49524757","AmazingTurtle","There are already a handful of repos doing essentially exactly this: `mlx-moe-offload`, `streamlx`, `mlx-moe`, `mlx-flash`, and `deepseek-v4-flash-mlx` - i.e. keep the resident…",{"id":106,"by":107,"text":108},"49530355","c0rruptbytes","so many inference project, omlx already supports all of this and has a 1000 people trying to optimize it constantly",{"value":110,"source":87,"evidence":111},"README reads like AI-generated text",[112,116],{"id":113,"by":114,"text":115},"49525120","embedding-shape","> Hugging Face is the bottleneck, not your link. README could clearly make use of a cleanup, seems to be more like a session log dump now than a good introduction to the project…",{"id":117,"by":118,"text":119},"49525309","Eufrat","I hate this AI style writing because since it doesn’t really understand flow, it’s being inserted in irrelevant places and it is extremely irritating to read.",{"value":121,"source":87,"evidence":122},"Disk reads may wear the SSD",[123,127],{"id":124,"by":125,"text":126},"49525577","jonplackett","Is this going to destroy my SSD?",{"id":128,"by":129,"text":130},"49527908","mrob","Reading causes insignificant wear (\"read disturb\") that likely isn't a problem, but I don't think it's possible to issue pure reads to modern SSDs. The NVMe spec mandates…",[132,136,146],{"value":133,"source":87,"evidence":134},"Lower memory use per process",[135],{"id":63,"by":64,"text":65},{"value":137,"source":87,"evidence":138},"Benchmarks against other inference tools",[139,143],{"id":140,"by":141,"text":142},"49548870","securecloudgrou","Would love to see metrics of model performance, comparison with oMLX\u002FOLLAMA\u002Fothers. Any tooling for local optimization on hardware.",{"id":144,"by":11,"text":145},"49524974","I'm sorry this makes it seem like I didn't do my research. I did a TON. To fix it I'll add a benchmark\u002Fcomparison table. Also, I wouldn't call it market research since this is…",{"value":147,"source":87,"evidence":148},"Native desktop app",[149],{"id":150,"by":64,"text":151},"49536475","Also I don't know if people like this... but could this be wrapped around a nice desktop app?",[153,160],{"value":154,"source":87,"evidence":155},"Energy consumption",[156],{"id":157,"by":158,"text":159},"49525005","ErenayDev","how much energy does it consume?",{"value":161,"source":87,"evidence":162},"Real use case",[163],{"id":63,"by":64,"text":65},[165,169,173],{"value":166,"source":87,"evidence":167},"oMLX",[168],{"id":106,"by":107,"text":108},{"value":170,"source":87,"evidence":171},"mlx-moe-offload, streamlx, mlx-flash",[172],{"id":102,"by":103,"text":104},{"value":174,"source":87,"evidence":175},"Ollama",[176],{"id":63,"by":64,"text":65},{"comments":178,"answered":179,"key":180},27,8,[181,184,187],{"id":182,"text":183},"49525286","thanks! I'll do!",{"id":185,"text":186},"49525327","I feel you! fix incomming",{"id":188,"text":189},"49526125","interesting! Yes, thermal is important. Pretty cool project man! Starred and checking it out!",{"commenters":191,"max_depth":179,"maker_replies":178,"first_comment_mins":192,"lifespan_hours":193},68,21,122,[195,197,199,200,202,205,208,212,215,217,221,224,228,232,235,237,240,244,248,250,254,255,258,260,262,265,267,270,273,275,279,282,285,288,291,293,296,299,300,303,305,308,312,314,317,321,323,327,331,334,337,339,341,343,345,347,349,353,356,358,361,363,365,369,373,377,381,384,387,388,390,392,394,396,399,401,404,406,408,411,413,415,416,419,422,425,428,431,433,436,440,441,445,447,450,453,455,458,460,462,465,467,470,472,474,476,478,480,483,485,487,489,492,494,496,499,501],[113,39,44,196,114,39],"project\u002Fcritical\u002Fdocs",[182,39,39,198,11,44],"maker\u002Fneutral\u002Fnone",[117,39,44,196,118,39],[185,39,39,201,11,44],"maker\u002Fneutral\u002Fdocs",[203,39,39,204,118,39],"49525951","commenter\u002Fneutral\u002Fnone",[206,39,39,204,207,39],"49527495","cyanydeez",[209,39,32,210,211,39],"49527888","project\u002Fneutral\u002Fdocs","makira",[213,39,44,196,214,39],"49528312","trollbridge",[216,39,39,204,114,39],"49534369",[218,39,39,219,220,39],"49576711","commenter\u002Fneutral\u002Fdocs","brookst",[222,39,44,196,223,39],"49529273","ricardobeat",[225,44,32,226,227,39],"49527160","project\u002Fneutral\u002Fperformance","mulemisterX",[229,44,39,230,231,39],"49528004","commenter\u002Fneutral\u002Fperformance","ig0r0",[233,44,39,230,234,39],"49528166","metadat",[236,44,39,230,231,39],"49532273",[238,44,39,230,239,39],"49528749","pram",[241,44,39,242,243,39],"49529684","other\u002Fneutral\u002Fnone","hadlock",[245,44,39,246,247,39],"49530124","other\u002Fneutral\u002Fperformance","kamranjon",[249,44,39,230,125,39],"49558655",[251,34,44,252,253,39],"49525218","project\u002Fcritical\u002Fperformance","prometheus1992",[188,34,39,204,11,44],[256,34,39,230,257,39],"49526511","Balooga",[259,34,39,204,253,39],"49527297",[261,34,39,246,214,39],"49528397",[263,34,39,246,264,39],"49532686","monster_truck",[90,32,18,266,91,39],"project\u002Fpositive\u002Fidea",[268,32,39,269,11,44],"49525356","maker\u002Fpositive\u002Fidea",[271,32,44,252,272,39],"49527085","tyre",[274,32,44,252,214,39],"49528323",[276,45,39,277,278,39],"49530561","other\u002Fneutral\u002Fidea","jmward01",[280,45,39,204,281,39],"49534355","MayeulC",[283,18,44,196,284,39],"49525114","drcongo",[286,18,39,242,287,39],"49525322","drums8787",[289,18,39,242,290,39],"49525542","bogzz",[292,18,39,242,284,39],"49526195",[294,18,39,204,295,39],"49526707","brailsafe",[297,18,39,242,298,39],"49526120","thirtygeo",[94,40,39,277,95,39],[301,40,39,277,302,39],"49528211","cosmic_cheese",[304,40,39,277,95,39],"49528529",[306,40,39,277,307,39],"49589256","softwarewright",[309,42,32,310,311,39],"49525564","project\u002Fneutral\u002Fidea","atif089",[313,42,39,198,11,44],"49526090",[315,42,32,310,316,39],"49527186","red_hare",[318,179,32,319,320,39],"49545169","project\u002Fneutral\u002Fdesign","nerdsnipe",[322,179,39,198,11,44],"49545785",[324,325,44,252,326,39],"49545904",9,"jedbrooke",[328,329,39,246,330,39],"49533837",10,"jwr",[332,329,39,230,333,39],"49534227","nixon_why69",[335,329,39,230,336,39],"49534672","ch_sm",[338,329,39,230,333,39],"49535139",[340,329,39,230,330,39],"49584609",[157,342,32,226,158,39],11,[344,342,39,198,11,44],"49525155",[150,346,32,319,64,39],12,[348,346,39,198,11,44],"49536631",[350,351,39,277,352,39],"49525061",13,"karmakaze",[354,351,39,277,355,39],"49525099","0x457",[357,351,39,198,11,44],"49525216",[359,351,39,277,360,39],"49528757","rzzzt",[362,351,39,242,355,39],"49529035",[364,351,39,242,11,44],"49525210",[366,367,39,242,368,39],"49527207",14,"nikanj",[370,371,32,226,372,39],"49529004",15,"amelius",[374,371,32,375,376,39],"49529208","project\u002Fmixed\u002Fperformance","pornel",[378,379,32,310,380,39],"49528769",16,"siris9476",[382,379,39,383,11,44],"49549739","project\u002Fpositive\u002Fperformance",[385,379,32,386,380,39],"49550803","maker\u002Fneutral\u002Fperformance",[140,35,32,226,141,39],[389,35,32,310,141,39],"49548888",[106,33,44,391,107,39],"project\u002Fcritical\u002Fidea",[393,33,39,198,11,44],"49530960",[395,33,39,310,11,44],"49531004",[63,397,32,398,64,39],19,"project\u002Fmixed\u002Fexecution",[400,397,39,198,11,44],"49536772",[402,397,32,403,64,39],"49540733","project\u002Fneutral\u002Fnone",[405,397,39,204,11,44],"49545807",[124,407,44,252,125,39],20,[409,407,39,204,410,39],"49525647","cromka",[412,407,39,226,11,44],"49525870",[414,407,39,226,11,44],"49525887",[128,407,32,226,129,39],[417,407,32,226,418,39],"49525673","egorfine",[420,407,44,252,421,39],"49530084","ElectricalUnion",[423,407,44,252,424,39],"49529049","Gigachad",[426,192,32,310,427,39],"49527514","kethinov",[429,192,44,391,430,39],"49532584","kzrdude",[102,432,44,391,103,39],22,[434,432,39,204,435,39],"49524836","api",[437,432,39,438,439,39],"49524956","commenter\u002Fneutral\u002Fidea","docheinestages",[144,432,39,201,11,44],[442,432,18,443,444,39],"49525070","maker\u002Fpositive\u002Fnone","EyMaddis",[446,432,39,443,11,44],"49525195",[448,432,39,204,449,39],"49525115","genxy",[451,432,39,204,452,39],"49525214","oceanplexian",[454,432,39,204,11,44],"49525011",[456,432,32,310,457,39],"49524846","Barbing",[459,432,39,201,11,44],"49524953",[461,432,39,204,457,39],"49529240",[463,432,44,391,464,39],"49524906","dofm",[466,432,39,204,11,44],"49524960",[468,432,39,204,469,39],"49525004","noir_lord",[471,432,39,310,11,44],"49525180",[473,432,39,204,464,39],"49525313",[475,432,39,226,11,44],"49525582",[477,432,39,242,464,39],"49527184",[479,432,39,310,11,44],"49524922",[481,432,18,443,482,39],"49525550","xlayn",[484,432,32,310,430,39],"49524969",[486,432,39,204,449,39],"49525095",[488,432,39,242,295,39],"49526757",[490,432,39,242,491,39],"49527117","brcmthrowaway",[493,432,39,198,11,44],"49530943",[495,432,39,204,295,39],"49531895",[497,432,39,277,498,39],"49527822","mannyv",[500,432,39,204,11,44],"49530947",[502,432,44,196,503,39],"49533194","sersi",[505,508],{"id":506,"t":507,"points":32,"comments":44},"49510441",1788187614,{"id":4,"t":12,"points":14,"comments":15},[510,514,519,524,528,533],{"id":511,"name":512,"points":513,"category":20},"49461817","load-bearing",710,{"id":515,"name":516,"points":517,"category":518},"49489376","Laser Graffiti",253,"creative",{"id":520,"name":521,"points":522,"category":523},"49471965","OpenTIE",231,"games",{"id":525,"name":526,"points":522,"category":527},"49478426","wilsonharper.net","hardware",{"id":529,"name":530,"points":531,"category":532},"49471407","Experiential",222,"dev_tools",{"id":534,"name":535,"points":536,"category":537},"49528895","Weedout",185,"consumer_social",false,"v2",{"id":541,"name":5,"line":6,"ai":538,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":542,"top":18,"kw":543,"maker":11,"domain":10,"tier":19,"attempts":34,"strip":544,"url":9,"mix":545,"praise":86,"objection":99,"sub":21},49524447,-0.34,[24,25,26,27,28],"1111301051100033103003331310211503105301",[32,33,34,35],1794036709178]