15:[[47497406,{"by":"msiraj1","descendants":1,"id":47497406,"kids":[47511705,47525058],"score":2,"text":"Hi all, this is my first question! I have been finding a lot of pre-processing tools to anonymize prompt data, but was wondering if anyone knew of tools that could be used in post-processing llm chat history files.

I want to conduct a study that strives to more easily anonymize the participant chat history so that when I receive it, it reduces PII risk.

another step I will need to add is just dropping chats that discuss personal health or rather summarizes chats that discuss topics of personal health? I really don't know, hence me asking here before just developing it on my own!","time":1774314377,"title":"Anonymize / de-identify LLM chat history export, post-processing","type":"story"}],[47511075,{"by":"zane__chen","descendants":15,"id":47511075,"kids":[47522413,47565301,47533061,47513327,47516285,47518176,47515534,47527451,47538412,47541393,47514076,47511320,47547965,47517438,47526873,47515200,47515473,47513168,47511451,47514680,47525046],"score":5,"text":"As titled.","time":1774395030,"title":"Ask HN: Do you feel less happy when coding with agent?","type":"story"}],[47523335,{"by":"MeetRickAI","dead":true,"id":47523335,"kids":[47539180,47525026,47527058,47525036],"score":8,"time":1774473139,"type":"story"}],[47523792,{"by":"davidpelayo","descendants":0,"id":47523792,"kids":[47525025],"score":4,"time":1774475618,"title":"Show HN: I coded Podhoc – Podcast generator to learn on the go","type":"story","url":"https://podhoc.com"}],[47520220,{"by":"jawiggins","descendants":59,"id":47520220,"kids":[47528314,47525453,47598854,47525855,47524587,47528440,47527060,47527007,47530431,47527135,47530037,47531010,47525213,47524594,47526092,47525242,47622871,47530263,47529759,47524726,47560717,47532271,47529597,47557361,47525020,47530136],"score":88,"text":"$16","time":1774458621,"title":"Show HN: Optio – Orchestrate AI coding agents in K8s to go from ticket to PR","type":"story","url":"https://github.com/jonwiggins/optio"}],[47507666,{"by":"felix089","descendants":98,"id":47507666,"kids":[47508505,47509457,47515524,47515726,47511835,47514072,47591872,47577161,47516688,47509481,47517991,47515692,47508466,47524068,47509256,47530154,47542113,47514047,47515300,47517164,47553160,47525715,47512268,47515349,47513494,47513941,47520249,47527301,47517046,47507931,47515277,47510172,47557319,47521704,47512684,47514394,47529219,47511017,47509557,47522911,47516716,47511278,47513816,47517331,47508926,47559923,47538501,47514930,47517251,47513216,47513982,47513095,47514693,47508186,47509372,47526999,47526719,47518583,47522695],"score":118,"text":"Hey HN! After the Car Wash Test post got quite a big discussion going (400+ comments, https://news.ycombinator.com/item?id=47128138), I spent the past few weeks building a tool so anyone can run these kinds of questions and get structured results. No signup and free to use.

You type a question, define answer options, pick up to 50 models at a time from a pool of 200+, and they all answer independently under identical conditions. No system prompt, structured output, same setup for every model.

You can also run a debate round where models see each other's reasoning and get a chance to change their minds. A reviewer model then summarizes the full transcript. All models are routed via my startup Opper. Any feedback is welcome!

Hope you enjoy it, and would love to hear what you think!","time":1774379727,"title":"Show HN: AI Roundtable – Let 200 models debate your question","type":"story","url":"https://opper.ai/ai-roundtable/"}],[47503617,{"by":"sohamrj","descendants":108,"id":47503617,"kids":[47506611,47508125,47513088,47507935,47505089,47510430,47506848,47506518,47506174,47505771,47524780,47508952,47514813,47532895,47514530,47509946,47506364,47513373,47504790,47504887,47534287,47512141,47506066,47510942,47514760,47513930,47508144,47528141,47512318,47521001,47512290,47505320,47510087,47505177,47511952,47508373,47514239,47505718,47522669,47509272],"score":438,"text":"Gemini Embedding 2 can project raw video directly into a 768-dimensional vector space alongside text. No transcription, no frame captioning, no intermediate text. A query like "green car cutting me off" is directly comparable to a 30-second video clip at the vector level.

I used this to build a CLI that indexes hours of footage into ChromaDB, then searches it with natural language and auto-trims the matching clip. Demo video on the GitHub README.\nIndexing costs ~$2.50/hr of footage. Still-frame detection skips idle chunks, so security camera / sentry mode footage is much cheaper.","time":1774364307,"title":"Show HN: Gemini can now natively embed video, so I built sub-second video search","type":"story","url":"https://github.com/ssrajadh/sentrysearch"}],[47521785,{"by":"shalomer","descendants":13,"id":47521785,"kids":[47522541,47522250,47528462,47522614,47522620,47521796,47522642],"score":6,"text":"$17","time":1774465697,"title":"Show HN: I built a voice AI that responds like a real woman","type":"story"}],[47494696,{"by":"noemit","descendants":22,"id":47494696,"kids":[47498409,47496895,47569769,47540979,47518087,47561130,47540200,47499371,47499342,47499000,47548754,47508348,47532964,47526878,47508063,47498203,47500805,47522625,47497365,47498274],"score":41,"text":"I ran an experiment to see if CLI actually was the most intuitive format for tool calling. (As claimed by a ex-Manus AI Backend Engineer) I gave my model random scenarios and a single tool "run" - i told it that it worked like a CLI. I told it to guess commands.

it guessed great commands, but it formatted it always with a colon up front, like \n:help\n:browser\n:search\n:curl

It was trained on how terminals look, not what you actually type (you don't type the ":")

I have since updated my code in my agent tool to stop fighting against this intuition.

LLMs they learn what commands look like in documentation/artifacts, not what the human actually typed on the keyboard.

Seems so obvious. This is why you have to test your LLM and see how it naturally works, so you don't have to fight it with your system prompt.

This is Kimi K2.5 Btw.","time":1774297908,"title":"LLMs learn what programmers create, not how programmers work","type":"story"}],[47493567,{"by":"caffeinedoom","descendants":9,"id":47493567,"kids":[47522795,47529888,47523034,47522237,47521884,47500411,47528367,47522602],"score":46,"text":"$18","time":1774291742,"title":"Show HN: I built a site that maps the web from a bounty hunter's perspective","type":"story","url":"https://www.neobotnet.com/"}],[47506641,{"by":"HeyMeco","descendants":39,"id":47506641,"kids":[47518934,47507522,47507428,47507039,47507182,47507144,47507418,47506965,47518563],"score":114,"time":1774375156,"title":"ARM AGI CPU: Specs and SKUs","type":"story","url":"https://sbcwiki.com/docs/soc-manufacturers/arm/arm-silicon/"}],[47516650,{"by":"matthiaswh","descendants":178,"id":47516650,"kids":[47517103,47518914,47517877,47519259,47517096,47520600,47519685,47518388,47517589,47517446,47517981,47518172,47522037,47517560,47517708,47518058,47525982,47519584,47517750,47519693,47517680,47517482,47528261,47517651,47526496,47517983,47526132,47526609,47518198,47519359,47519331,47520004,47519535,47520238,47519325,47520623,47525842,47518443,47519133,47520231,47517697,47517646,47519348,47525821,47550675,47520989,47520556,47517761,47520680,47518995,47517601,47518538,47520942,47517416,47517303],"score":361,"time":1774442954,"title":"Ensu – Ente’s Local LLM app","type":"story","url":"https://ente.com/blog/ensu/"}],[47513475,{"by":"ray__","descendants":166,"id":47513475,"kids":[47514494,47517134,47516613,47522769,47516077,47514198,47523603,47523628,47572239,47523556,47515243,47514056,47523985,47516527,47533300,47517320,47550678,47515947,47514482,47516972,47515814,47517281,47516572,47526798,47514379,47514903,47540831,47516899,47540835,47518279,47519320,47515829,47517605,47520996,47516211,47515088,47518798,47517233,47514763,47514574,47515179,47514227,47515365,47516763,47525310,47525083,47518516,47516180,47514995],"score":576,"time":1774414854,"title":"TurboQuant: Redefining AI efficiency with extreme compression","type":"story","url":"https://research.google/blog/turboquant-redefining-ai-efficiency-with-extreme-compression/"}],[47508745,{"by":"jakelsaunders94","descendants":527,"id":47508745,"kids":[47509620,47509064,47509693,47509365,47508869,47509050,47509035,47508965,47511152,47511530,47509274,47509030,47509743,47509322,47508928,47512392,47508999,47511599,47508931,47509257,47509062,47508909,47512674,47509973,47508938,47509097,47508946,47508901,47511620,47517430,47513735,47508943,47509003,47509369,47508893,47509028,47511089,47513116,47512040,47509744,47512411,47509202,47509508,47512118,47508930,47508923,47511333,47517116,47509824,47508889,47508894,47508971,47511471,47510320,47511072,47510042,47510945,47510641,47509197,47509154,47514197,47511059,47510596,47518856,47509418,47509489,47512576,47569059,47512921,47508874,47510276,47521403,47513046,47509196,47508940,47509597,47509198,47512064,47509218,47509037,47523047,47512756,47509532,47512253,47508990,47513003,47509353,47509063,47511162,47514734,47509515,47509191,47509304,47511221,47509173,47509172,47527402,47512567,47508959,47509111,47512119,47509614,47518438,47509499,47510321,47511314,47509753,47509405,47515235,47509605,47513873,47509187,47509002,47512415,47509141,47509678,47509259,47509175,47510732,47509081,47509756,47509240,47513050,47510698,47509070,47517841,47514769,47509137,47508858,47508994,47509679,47509014,47512949,47509065,47514770,47508998,47510242,47511364,47549742,47514417,47510681,47512600,47512281,47511888,47542405,47508939,47540699,47509149,47526956,47509949,47515307,47511933,47511667,47509356,47513022,47510453,47509309,47508948,47510033,47508864],"score":746,"time":1774384268,"title":"Is anybody else bored of talking about AI?","type":"story","url":"https://blog.jakesaunders.dev/is-anybody-else-bored-of-talking-about-ai/"}],[47475930,{"by":"tosh","descendants":46,"id":47475930,"kids":[47507458,47511088,47510710,47507423,47512766,47508541,47507637,47509539,47516713,47509241],"score":140,"time":1774173323,"title":"Data Manipulation in Clojure Compared to R and Python","type":"story","url":"https://codewithkira.com/2024-07-18-tablecloth-dplyr-pandas-polars.html"}],[47439413,{"by":"usrbinenv","descendants":147,"id":47439413,"kids":[47500867,47500561,47502157,47500593,47500765,47500588,47501012,47509012,47502665,47500526,47501360,47502874,47508530,47508731,47501611,47501839,47501650,47509293,47502091,47553091,47501343,47505550,47509602,47503117,47502583,47506168,47500789,47502877,47504703,47509203,47501098],"score":127,"time":1773928152,"title":"Qite.js – Frontend framework for people who hate React and love HTML","type":"story","url":"https://qitejs.qount25.dev"}],[47499672,{"by":"jberthom","descendants":106,"id":47499672,"kids":[47500311,47499984,47500594,47507728,47507003,47500072,47508231,47501289,47500124,47500306,47504657,47500259,47500795,47598318,47500065,47500553,47501072,47521650,47499937,47501515,47500018,47501986,47578308,47507203,47500879,47512934,47499968,47507229,47502729,47514234,47500293,47510958,47525072,47501069,47501177,47508601,47499959,47500786,47514918,47500426,47588572,47598767,47528571,47507562,47522033,47500788,47507327,47563282,47501768,47553224,47512267,47524746,47519493,47510731,47510460,47509170,47499977],"score":161,"text":"$19","time":1774338406,"title":"Show HN: ProofShot – Give AI coding agents eyes to verify the UI they build","type":"story","url":"https://github.com/AmElmo/proofshot"}],[47484267,{"by":"adinhitlore","descendants":5,"id":47484267,"kids":[47494163,47485392,47485017,47496902],"score":3,"text":"Assuming your software meets some of these:

* Better than potential successors of SOTA transformers: Mamba, Hyenna, RWKV, xLSTM;\n* Gemini/claude estimate potential ip values in the millions;\n* All implemented in C (ok this may be a minus) but also ported to C# and F# - so no python nor rust;\n* humble code size: 10-15k lines of code but i mean come on gpt-1 was under 1000 lines...in python.

In short:

the problem: transformers are slow and could be smarter;\nsolution: you have fast and smart alternative. \n10x!","time":1774228313,"title":"What would you do if you have AI software that may be transformers alternative?","type":"story"}],[47466513,{"by":"pangon","descendants":12,"id":47466513,"kids":[47468879,47477819,47476181,47473166,47475191,47469570,47582764,47526531,47470756,47478459,47469015,47496881,47478290],"score":27,"text":"$1a","time":1774096989,"title":"Show HN: AI SDLC Scaffold, repo template for AI-assisted software development","type":"story","url":"https://github.com/pangon/ai-sdlc-scaffold/"}],[47495510,{"by":"latand6","descendants":29,"id":47495510,"kids":[47496557,47501877,47497552,47530661,47509300,47496900,47525767,47509319,47536748,47497473,47500238,47552610,47497379,47496752,47498046,47555066,47506635,47516520,47496711,47497373],"score":10,"text":"$1b","time":1774302192,"title":"Ask HN: Is using AI tooling for a PhD literature review dishonest?","type":"story"}],[47494999,{"by":"poopcat","descendants":5,"id":47494999,"kids":[47495943,47495864,47495623,47496684],"score":30,"text":"https://github.com/elder-plinius/V3SP3R","time":1774299408,"title":"Flipper Zero gets an AI upgrade","type":"story","url":"https://gizmodo.com/flipper-zero-everyones-favorite-legally-dubious-hacker-tool-gets-an-ai-upgrade-2000736967"}],[47487536,{"by":"mooreds","descendants":319,"id":47487536,"kids":[47495187,47492239,47491752,47487719,47487781,47487646,47497273,47500261,47496109,47488992,47494129,47487856,47502728,47499173,47494281,47492383,47496834,47496491,47499559,47487770,47487611,47493117,47491943,47498404,47492557,47494045,47493651,47493776,47501345,47507719,47487740,47499791,47495457,47496904,47487626,47493409,47496170,47496692,47493012,47499172,47487712,47487764,47497333,47501590,47493443,47487737,47496118,47501133,47493183,47493046,47492205,47497069,47500901,47501657,47498384,47492867,47492487,47487718,47497943,47501512,47500606,47487765,47499060,47500943,47495256,47501710,47497894,47492193,47487608,47488094,47492049,47488009,47487912,47506160,47500101,47502201,47526636,47496250,47496270,47496490,47497191,47497727,47499393,47493923,47488038,47487935,47494010,47487617,47487662,47494365,47496661,47497375],"score":320,"time":1774261953,"title":"I built an AI receptionist for a mechanic shop","type":"story","url":"https://www.itsthatlady.dev/blog/building-an-ai-receptionist-for-my-brother/"}],[47476422,{"by":"mft_","descendants":120,"id":47476422,"kids":[47477552,47477630,47481034,47476717,47477005,47478292,47476848,47479799,47484838,47478657,47476802,47495878,47477260,47485701,47479779,47477659,47478173,47484353,47477741,47478146,47477139,47477975,47482341,47479203,47476883,47480145,47487594,47485739,47487344,47478883,47491377,47482280,47486920,47480254,47484308,47477658,47477908,47480264,47482678,47478893,47483164,47479916,47484041,47481315,47485171,47478402,47484548,47494762,47486589,47486732,47478577,47483317,47487716,47477049,47479922,47476718,47476917,47476725],"score":398,"time":1774179003,"title":"Flash-MoE: Running a 397B Parameter Model on a Laptop","type":"story","url":"https://github.com/danveloper/flash-moe"}],[47480447,{"by":"wallflower","descendants":402,"id":47480447,"kids":[47482405,47485520,47485439,47482107,47487783,47482822,47482768,47483134,47486042,47499421,47482060,47484772,47485837,47489510,47485958,47485237,47490941,47486436,47491542,47490242,47485742,47490537,47490272,47493123,47486807,47482953,47483140,47486162,47485740,47482188,47487307,47488828,47482478,47560373,47491210,47485355,47490617,47487703],"score":227,"time":1774203481,"title":"What young workers are doing to AI-proof themselves","type":"story","url":"https://www.wsj.com/economy/jobs/ai-jobs-young-people-careers-14282284"}],[47482760,{"by":"raybb","descendants":74,"id":47482760,"kids":[47483423,47486570,47484812,47484251,47484930,47484914,47484053,47483939,47483602,47486515,47485334,47488228,47486435,47486165,47483421,47495741,47487717,47494582,47484655,47483159,47484897,47486865,47488303,47484035,47483228,47485679,47490890,47483641,47488208,47484003,47487658,47524077,47505334,47486947,47484590,47498087,47487925,47485711,47487273,47498301,47487692,47484426],"score":132,"time":1774217514,"title":"They’re vibe-coding spam now","type":"story","url":"https://tedium.co/2026/02/25/vibe-coded-email-spam/"}]]

user:	QubridAI
created:	December 28, 2025
karma:	-19
about:	Qubrid AI - The Full AI Stack is designed to give developers, researchers, and enterprises the GPU performance, AI-ready software, and cost-efficiency needed to unlock the full potential of AI.

1.	1 points by QubridAI 54 days ago \| discuss
2.	1 points by QubridAI 55 days ago \| discuss
3.	1 points by QubridAI 56 days ago \| discuss