19:[[47244042,{"by":"ArmaloAI","descendants":8,"id":47244042,"kids":[47390819,47343892,47245480,47260913,47272567,47244471,47257859],"score":3,"text":"$1a","time":1772607182,"title":"Show HN: Armalo AI – The Infrastructure for Agent Networks","type":"story"}],[47190721,{"by":"HexitLabs","descendants":2,"id":47190721,"kids":[47343228,47190735,47190746],"score":2,"text":"$1b","time":1772255381,"title":"Show HN: Vigil – Zero-dependency safety guardrails for AI agent tool calls","type":"story"}],[47335481,{"by":"alberto-m","descendants":1,"id":47335481,"kids":[47343212],"score":3,"text":"The other day I was experimenting with `deepseek-v3.1:671b-cloud` using Ollama, when a coding question got answered with a medical diagnosis on the basis of a list of symptoms. I put it down to some strange LLM hallucination, but today a Reddit thread [1] showed a similar phenomenon of an answer referring to a totally different prompt.

The only logical explanation seems to be that the server is occasionally failing to pair prompts and answers. If this is the case I think users should be aware of this problem, in addition to the other general security issues of using non-local models.

Has anyone else experienced this?

[1] (NSFW) https://old.reddit.com/r/ollama/comments/1rqoez5/so_this_has_started_happening_recently_with/","time":1773236603,"title":"Tell HN: Crosstalk when using Ollama with cloud DeepSeek models?","type":"story"}],[47342718,{"by":"ramoz","descendants":4,"id":47342718,"kids":[47346678,47343196,47349161],"score":6,"text":"If anything at all.","time":1773266445,"title":"Ask HN: What are you using to mitigate prompt injection?","type":"story"}],[47327559,{"by":"aray07","descendants":489,"id":47327559,"kids":[47331205,47328901,47327956,47327999,47330070,47327789,47328132,47327990,47328984,47332648,47332906,47331027,47328885,47331298,47327957,47331395,47327825,47331538,47331020,47333479,47329527,47339749,47330820,47330937,47335430,47328207,47328704,47332798,47349137,47332508,47328119,47334003,47331400,47335017,47368902,47327907,47329486,47375398,47334676,47327982,47330263,47330775,47327953,47337968,47328951,47327993,47332112,47331844,47332807,47331490,47335211,47332846,47327938,47327862,47327733,47329503,47330525,47332796,47332515,47332945,47328306,47331677,47328727,47331129,47333923,47331902,47347231,47331536,47332928,47331999,47331388,47332826,47330955,47332811,47333519,47337680,47327921,47328538,47330061,47333549,47331304,47329440,47329658,47331921,47333640,47335206,47337840,47330722,47357050,47333133,47337445,47335701,47333844,47327878,47332471,47335890,47337965,47337310,47340641,47330036,47333767,47328203,47337838,47334956,47334949,47334959,47329082,47334958,47331891,47330336,47334324],"score":429,"time":1773169786,"title":"Agents that run while I sleep","type":"story","url":"https://www.claudecodecamp.com/p/i-m-building-agents-that-run-while-i-sleep"}],[47333627,{"by":"mycroft_4221","descendants":193,"id":47333627,"kids":[47336554,47335886,47335965,47335738,47336165,47335308,47335950,47335837,47339233,47335585,47335597,47335669,47337451,47337595,47344723,47338617,47336428,47338510,47337354,47343905,47336990,47344689,47345670,47336333,47338210,47340437,47336414,47338436,47338373,47337317,47336737,47337712,47339601,47341695,47344277,47335856,47336901,47337184,47344533,47337302,47339856,47339410,47337437,47350588,47456787,47335688,47335800,47344437,47349400,47336730,47336541,47337350],"score":506,"time":1773223143,"title":"How we hacked McKinsey's AI platform","type":"story","url":"https://codewall.ai/blog/how-we-hacked-mckinseys-ai-platform"}],[47308325,{"by":"gauravvij137","descendants":0,"id":47308325,"kids":[47334718],"score":3,"text":"$1c","time":1773060286,"title":"Show HN: Auto LLM Ranker – Describe a task in English and get ranked models","type":"story","url":"https://github.com/gauravvij/llm-evaluator"}],[47320614,{"by":"bombastic311","descendants":129,"id":47320614,"kids":[47331686,47327165,47328299,47327319,47328341,47327666,47327342,47333820,47329173,47327012,47329923,47327160,47329624,47329066,47330678,47326857,47329151,47338347,47334328,47329216,47327743,47327664,47327104,47330079,47333024,47326839,47334080,47329996,47337118,47330092,47335989,47327304,47329241,47327952,47328134,47333283,47336155,47332919,47334599,47343960,47409394,47333665,47334368,47333656,47335687,47335694,47330725,47335998,47335469,47332953],"score":277,"time":1773132520,"title":"Levels of Agentic Engineering","type":"story","url":"https://www.bassimeledath.com/blog/levels-of-agentic-engineering"}],[47315770,{"by":"ALMOIZ_MOHMED","descendants":1,"id":47315770,"kids":[47317659,47333125],"score":3,"text":"$1d","time":1773091530,"title":"Energy-based Model (EBM) for enterprise AI security Ship it or keep tuning?","type":"story"}],[47323017,{"by":"ndr42","descendants":485,"id":47323017,"kids":[47324904,47325002,47331140,47326136,47325447,47325762,47329107,47324969,47331323,47325059,47330569,47329184,47330122,47324856,47325001,47329240,47323159,47324711,47329305,47323205,47326999,47323343,47326303,47330659,47325665,47324940,47329463,47325684,47329639,47329733,47325588,47326428,47329494,47323276,47325041,47338424,47326815,47328374,47325910,47331444,47325982,47329955,47347395,47328317,47325819,47331643,47330919,47325997,47329132,47326366,47336844,47328555,47331267,47325062,47331128,47329512,47331292,47329965,47325573,47329351,47330629,47326920,47330146,47330809,47325132,47325313,47330807,47330219,47328730,47325351,47329514,47329988,47330514,47325295,47330057,47329721,47330014,47328744,47332258,47329915,47330530,47330045,47339296,47325822,47330790,47328304,47330115,47329736,47328903,47324728,47326687,47325800,47329705,47323055,47330713],"score":659,"text":"https://www.ft.com/content/7cab4ec7-4712-4137-b602-119a44f77... (https://archive.ph/wXvF3)

https://twitter.com/lukolejnik/status/2031257644724342957 (https://xcancel.com/lukolejnik/status/2031257644724342957)","time":1773149477,"title":"After outages, Amazon to make senior engineers sign off on AI-assisted changes","type":"story","url":"https://arstechnica.com/ai/2026/03/after-outages-amazon-to-make-senior-engineers-sign-off-on-ai-assisted-changes/"}],[47325105,{"by":"harperlabs","descendants":6,"id":47325105,"kids":[47420654,47325268,47371804,47328379,47424704,47326204],"score":2,"text":"$1e","time":1773158647,"title":"Ask HN: How are you testing AI agents before shipping to production?","type":"story"}],[47327033,{"by":"LunarFrost88","descendants":5,"id":47327033,"kids":[47327495,47327963,47327608],"score":3,"text":"$1f","time":1773167259,"title":"Show HN: Railyard – open and secure runtime for Claude Code","type":"story"}],[47320661,{"by":"pjmlp","descendants":463,"id":47320661,"kids":[47320789,47322414,47320911,47320849,47320777,47321408,47323365,47320784,47320827,47320931,47325482,47328366,47323110,47323794,47323840,47320782,47321969,47321945,47369703,47321364,47321803,47321161,47321701,47327336,47325147,47321197,47321261,47324441,47327084,47329012,47320775,47329581,47327213,47321517,47320805,47321448,47321384,47321486,47320888,47321906,47320885],"score":409,"time":1773132893,"title":"Redox OS has adopted a Certificate of Origin policy and a strict no-LLM policy","type":"story","url":"https://gitlab.redox-os.org/redox-os/redox/-/blob/master/CONTRIBUTING.md"}],[47209315,{"by":"rohanmunshi08","descendants":6,"id":47209315,"kids":[47211258,47260358,47260289,47210977,47308730],"score":23,"text":"$20","time":1772389640,"title":"Aura-State: Formally Verified LLM State Machine Compiler","type":"story"}],[47300747,{"by":"horseradish","descendants":251,"id":47300747,"kids":[47302932,47308173,47301689,47302739,47301681,47302081,47302162,47301804,47300819,47307784,47301788,47306520,47309181,47307417,47304522,47301712,47302078,47302604,47326688,47304595,47307400,47304442,47310167,47306725,47302767,47303333,47305241,47306027,47302709,47307374,47304096,47301980,47307168,47305926,47303171,47303967,47308144,47302325,47304423,47303898,47313809,47302823,47305943,47307644,47301963,47303569,47303900,47303386,47302257,47307001,47304280,47302161,47301717,47307666,47302810,47308083,47307824,47302877,47306957,47302809,47304443,47302129,47307992,47305383,47305137,47307508],"score":292,"time":1772999895,"title":"We should revisit literate programming in the agent era","type":"story","url":"https://silly.business/blog/we-should-revisit-literate-programming-in-the-agent-era/"}],[47228173,{"by":"PrateekRao01","descendants":4,"id":47228173,"kids":[47406432,47236075,47228186,47308078],"score":10,"text":"$21","time":1772513110,"title":"Show HN: Cortexa – Bloomberg terminal for agentic memory","type":"story","url":"https://cortexa.ink"}],[47301085,{"by":"atombender","descendants":178,"id":47301085,"kids":[47301472,47304797,47302296,47301342,47303325,47302271,47302154,47303248,47303052,47301978,47301969,47304433,47301338,47306213,47316066,47304201,47305005,47303435,47306126,47302973,47311555,47307232,47304693,47303616,47309431,47302547,47303721,47306280,47303597,47312545,47303088,47307084,47305067,47304303,47308408,47307914,47307628,47305455,47301363,47308442,47309767,47302180,47305271,47303914,47318542,47304297,47303009,47311512,47303840,47302076,47305685,47305037,47323066,47307018,47308496,47308062,47315026,47308911,47328183,47306949,47311112,47306581,47308934,47308220,47308920,47311788,47301554,47302097,47321264,47302806,47308118,47305133,47304438,47303238,47305763,47303882,47309404,47302593],"score":823,"time":1773001818,"title":"Agent Safehouse – macOS-native sandboxing for local agents","type":"story","url":"https://agent-safehouse.dev/"}],[47303111,{"by":"david927","descendants":1132,"id":47303111,"kids":[47303797,47303844,47305395,47304398,47304306,47303653,47307002,47306189,47306902,47308654,47303576,47311433,47304514,47304600,47304343,47310240,47303683,47303682,47305348,47304913,47303213,47303927,47308426,47305654,47304930,47308747,47306993,47303875,47303773,47304947,47304661,47305965,47303886,47325511,47306931,47305324,47307262,47364499,47306335,47305036,47304847,47304536,47307724,47304550,47310678,47304681,47306535,47304199,47303742,47304719,47310210,47305403,47309784,47304054,47306996,47305223,47305413,47304476,47305327,47309903,47303484,47305544,47304556,47306582,47305594,47306282,47311415,47307402,47310091,47304863,47308208,47304621,47310582,47306222,47303387,47305126,47305367,47305551,47313435,47307226,47325207,47306121,47326796,47313765,47310609,47303734,47325254,47305273,47304113,47307385,47305601,47305991,47304875,47305828,47303458,47306513,47305743,47309223,47307278,47307816,47305694,47307937,47306061,47464354,47313356,47303707,47309869,47305041,47307404,47309194,47307902,47304827,47304235,47308248,47304265,47306864,47304488,47303298,47305343,47305407,47303611,47305023,47307274,47310166,47305098,47308038,47306267,47304367,47309814,47311696,47307450,47317828,47315278,47304497,47303501,47371149,47305350,47322533,47310454,47306749,47304953,47304942,47367202,47303406,47340419,47304617,47306379,47305869,47306026,47310967,47310955,47313815,47305675,47306875,47303500,47303549,47463733,47306104,47305593,47328171,47347296,47309743,47304660,47309059,47303410,47310342,47309620,47308604,47307791,47310860,47305860,47312032,47312621,47352223,47304057,47305163,47310673,47307433,47310269,47303577,47303703,47310993,47338343,47309251,47305705,47305065,47308219,47305817,47381650,47306092,47348406,47306339,47312195,47312591,47322651,47307821,47352108,47305251,47304564,47304752,47336995,47336053,47305516,47309717,47305099,47324123,47303592,47304215,47305933,47304891,47309809,47309279,47316092,47304683,47349676,47304962,47309050,47305668,47303374,47320815,47308678,47310722,47309321,47320724,47304285,47303945,47308736,47310763,47308425,47304389,47308104,47306064,47305824,47314628,47308276,47310985,47315695,47304492,47304533,47309820,47308984,47307537,47303920,47303644,47303599,47303919,47424961,47312469,47306129,47323262,47309488,47317895,47305401,47322006,47303869,47323172,47304144,47305315,47309129,47351473,47311528,47304836,47303206,47303767,47311267,47309253,47315567,47304812,47305391,47305792,47314681,47308375,47308326,47325086,47340223,47303818,47304080,47328915,47308403,47319958,47307265,47303489,47303989,47307690,47319261,47303595,47369810,47309765,47304015,47304759,47360514,47311560,47304455,47331455,47305325,47303868,47306671,47311271,47327860,47306151,47305603,47306191,47304511,47308976,47308792,47310705,47305035,47314643,47314479,47303659,47307557,47387237,47307338,47305937,47303846,47307456,47307480,47303485,47307307,47320297,47307706,47308246,47303617,47306209,47306159,47309777,47304115,47316874,47322067,47309342,47311032,47317119,47303713,47308853,47307411,47321056,47324920,47304278,47303529,47306811,47303930,47304038,47309721,47304294,47303924,47356471,47306774,47303613,47315188,47307287,47315086,47308592,47303770,47315238,47328211,47306848,47314933,47303765,47303651,47305521,47306245,47306083,47303888,47304362,47305438,47311002,47308935,47306618,47305145,47306041,47304544,47307918,47304767,47304830,47304551,47309256,47307575,47324839,47304824,47307936,47307756,47304910,47303441,47323911,47307972,47305627,47308024,47304541,47305104,47304884,47310361,47338484,47306148,47307542,47307242,47305747,47304817,47306036,47364961,47316554,47306242,47303816,47307245,47322338,47317402,47305363,47310607,47311671,47317129,47314078,47305159,47305143,47330456,47309977,47308389,47307432,47306128,47304390,47310671,47310631,47305055,47307335,47310784,47306367,47315861,47309140,47304997,47310875,47304948,47307172,47307406,47305821,47311295,47307732,47305308,47304478,47306548,47322831,47304028,47309067,47307160,47305028,47305679,47308111,47304933,47304279,47305146,47306526,47317271,47304137,47306352,47315131,47313888,47321285,47307110,47307222,47313406,47305200,47310079,47304121,47305087,47308868,47316463,47305826,47304284,47306753,47304949,47305865,47311261,47306937,47303985,47316953,47306024,47305902,47306804,47316227,47305378,47307184,47305179,47362840,47308625,47305676,47307042,47307683,47303572,47311104,47310823,47304970,47305658,47304227,47308943,47307949,47306821,47309818,47303602,47304209,47303762,47304633,47305090,47307560,47307515,47305855,47309982,47305379,47306741,47307787,47303956,47308647,47306203,47307932,47303480,47306173,47368368,47305235,47304437,47307166,47304220,47310859,47365242,47307717,47306438,47305120,47308512,47314850,47304181,47309631,47303383,47308919,47309805,47304995,47325580,47304205,47304018,47315034,47305628,47303587,47307461,47304229,47307311,47310078,47308520,47306016,47410153,47306117,47305539,47325669,47322609,47305010,47304885,47304677,47311544,47304005,47308137,47305895,47307393,47307879,47306256,47309355,47318772,47304864,47304561,47308864,47310047,47305047,47305226,47312050,47307331,47307346,47307190,47303442,47307106,47306119,47315344,47308207,47310493,47310883,47304856,47304819,47304813,47303547,47304983,47308030,47305263,47305698,47304491,47306924,47313124,47304479,47307951,47308195,47312275,47318917,47319422,47303995,47323536,47304496,47304973,47306833,47306098,47305819,47308215,47307153,47305016,47306591,47310628,47306088,47306135,47317637,47304445,47305831,47307074,47311982,47309390,47303391,47303640,47304004,47307372,47306283,47305322,47303877,47313558,47314715,47304918,47313456,47304519,47304232,47310735,47310658,47305014,47311928,47313946,47304780,47307297,47307029,47309088,47308867,47306855,47309383,47308542,47305751,47308271,47305440,47306161,47311389,47310154,47308644,47305613,47310021,47303984,47324469,47307842,47311078,47304474,47303947,47327310,47305808,47306158,47427866,47309227,47327354,47303871,47304465,47304463,47310391,47306096,47347894,47306446,47308303,47317068,47310780,47305394,47304102,47307112,47304439,47304188,47320968,47306742,47303525,47363546,47303786,47335312,47368689,47312891,47303743,47349718,47323242,47310197,47333046,47305555,47304787,47305116,47303708,47307104,47307425,47306769,47366234,47333517,47304688,47306639,47308350,47310129,47316031,47308345,47304041,47303977,47304384,47311616,47318489,47304386,47320660,47304180,47304312,47306411,47304967,47312440,47309324,47310183,47305231,47314060,47360056,47476854,47308053,47331565,47366342,47434421,47306472,47453338,47311495,47310580,47355514,47320205,47324702,47350501,47333350,47340517,47446156,47328713,47355517,47320680,47352669,47311782,47353571,47305196,47309136,47326949,47303909,47305321,47322613,47307978,47367928,47405553,47320959,47306309,47307421,47304855,47306982,47324519,47306750,47321885,47344879,47304366,47303477,47307880,47320226,47347613,47304852,47318220,47303553,47322247,47306571,47320685,47305304,47360785,47306744,47317716],"score":290,"text":"What are you working on? Any new ideas that you're thinking about?","time":1773014838,"title":"Ask HN: What Are You Working On? (March 2026)","type":"story"}],[47291513,{"by":"walterbell","descendants":190,"id":47291513,"kids":[47301438,47301692,47292658,47301635,47295908,47302778,47292525,47295062,47292417,47292506,47293307,47295569,47310113,47292196,47292864,47292290,47292486,47310569,47294234,47296131,47299639,47310250,47293312,47292687,47292695,47305356,47293131,47293970,47298072,47295458,47296535,47293676,47306985,47304807,47303747,47293700,47293315,47292283,47295978,47295033,47292995,47293043,47301419,47292331,47292983,47318751,47295156,47301345,47300756,47295640,47293143,47307332,47297741,47319702,47295328,47296161,47297899,47300667,47298462,47301534,47292856,47292990,47304546,47301124,47296833,47295294,47353894,47294041],"score":380,"text":"https://news.ycombinator.com/item?id=47088813","time":1772917720,"title":"LLM Writing Tropes.md","type":"story","url":"https://tropes.fyi/tropes-md"}]]

user:	oliver_dr
created:	March 8, 2026
karma:	1
about:	Head of Growth at DeepRails (https://deeprails.com). We build evaluation and correction infrastructure for LLM outputs - hallucination detection, factual grounding, completeness checking. Focused on healthcare, legal, and education verticals where accuracy is non-negotiable.