15:[[48128003,{"by":"pycassa","descendants":85,"id":48128003,"kids":[48128998,48128564,48128426,48128461,48128983,48128520,48128915,48129406,48134251,48128924,48129373,48128569,48129022,48168609,48128857,48128644,48135084,48134852,48130624,48137053,48129721,48131019,48129697,48134503,48130665,48148224,48129161,48129430,48133869,48129688,48133873,48133874,48129675,48128550,48128696,48128429,48130319],"score":298,"text":"$16","time":1778708418,"title":"Tell HN: Dont use Claude Design, lost access to my projects after unsubscribing","type":"story"}],[48099186,{"by":"ck2","descendants":17,"id":48099186,"kids":[48099570,48100331,48100211,48099851,48102302,48101714,48100018,48100214,48100858,48101266,48107488,48108244,48100069],"score":87,"time":1778525857,"title":"Palantir to be granted \"unlimited access\" to UK NHS patient data","type":"story","url":"https://www.digitalhealth.net/2026/05/palantir-to-be-granted-unlimited-access-to-nhs-patient-data/"}],[48103042,{"by":"matrixhelix","descendants":89,"id":48103042,"kids":[48103545,48103904,48105248,48103588,48103859,48103488,48105659,48103648,48106845,48103730,48103745,48103977,48108391,48104288,48104197,48103971,48104808,48104859,48108186,48104492,48103350,48104325,48103655,48103773,48108196,48103815,48103963,48104763,48108347,48105022,48103791,48106851,48103736],"score":227,"time":1778549069,"title":"Claude Platform on AWS","type":"story","url":"https://claude.com/blog/claude-platform-on-aws"}],[48104428,{"by":"adunk","descendants":393,"id":48104428,"kids":[48106178,48113391,48106871,48105128,48105598,48105607,48105101,48105191,48105419,48106394,48106391,48105567,48105586,48104905,48105692,48105276,48110666,48108949,48109567,48106883,48105121,48111837,48112879,48109499,48106112,48106289,48115567,48106212,48107356,48105919,48106669,48106337,48107151,48110220,48105769,48106144,48106186,48107257,48105313,48105825,48109158,48110098,48110766,48114872,48111460,48105871,48107898,48105145,48104804,48106067,48106647,48105686,48109684,48108841,48119751,48115132,48109222,48107406,48105240,48105111,48113800,48106263,48108550,48110562,48115501,48106939,48105414,48107746,48110221,48105534,48105098,48108211,48107818,48110522,48106839,48112037,48106077,48108158,48105441,48107368,48122148,48105835,48110206,48110454,48104973,48108647,48105114,48106696,48107284,48107192,48107459,48105317],"score":707,"time":1778562684,"title":"Screenshots of Old Desktop OSes","type":"story","url":"http://www.typewritten.org/Media/"}],[48093043,{"by":"JumpCrisscross","descendants":190,"id":48093043,"kids":[48093710,48093752,48094940,48093863,48097500,48096669,48100784,48096007,48093986,48093818,48095201,48099986,48093711,48100653,48097552,48099389,48097641,48096904,48094646,48095559,48096838,48103651,48095033,48099249,48096224,48098440,48094218,48100993,48100282,48097017,48093639,48113558,48103081,48095946,48095018,48094905],"score":263,"time":1778493887,"title":"A.I. note takers are making lawyers nervous","type":"story","url":"https://www.nytimes.com/2026/05/09/business/dealbook/ai-notetakers-legal-risk.html"}],[48095550,{"by":"movis","descendants":762,"id":48095550,"kids":[48096558,48098278,48098176,48096231,48096113,48096317,48097825,48096067,48096108,48099068,48097764,48096398,48098962,48096840,48099636,48097194,48103296,48096353,48099943,48095860,48101830,48096168,48098343,48099643,48099580,48097006,48097292,48096218,48113303,48099974,48103228,48099982,48103709,48100003,48096305,48099804,48098132,48098876,48108534,48099293,48103066,48097522,48105651,48101532,48095840,48096644,48099734,48097232,48096173,48105868,48095762,48099724,48095989,48099581,48101663,48101265,48099726,48097769,48099035,48098411,48096831,48100749,48100297,48098037,48103594,48099799,48119558,48103310,48096721,48101880,48103611,48096281,48105365,48104124,48097869,48098882,48105574,48097953,48099798,48098547,48102295,48098344,48096593,48098273,48099979,48103893,48095929,48109114,48101164,48098922,48102651,48098484,48104102,48098523,48105543,48101635,48098875,48103485],"score":491,"time":1778510048,"title":"Software engineering may no longer be a lifetime career","type":"story","url":"https://www.seangoedecke.com/software-engineering-may-no-longer-be-a-lifetime-career/"}],[48087925,{"by":"0-bad-sectors","descendants":32,"id":48087925,"kids":[48090472,48089255,48090619,48092715,48089264,48088509,48122447,48113304,48120862,48104550,48092591,48088066,48111205,48089607,48099681,48123667,48117315,48103563,48089045,48098825,48102406,48093286,48105926,48088820,48089638,48103041,48138481,48128254,48092673,48113429,48091247,48092450,48106916,48091463,48092646],"score":27,"text":"Now whenever I reach for customer's support chat or phone I get an AI agent replying to me and I get into a useless loop for a couple of minutes before I start begging it to link me to a real person.

Will companies start losing customers because of that or people will eventually get used to this?","time":1778446647,"title":"Ask HN: Will low quality AI customer support be the new normal?","type":"story"}],[48092561,{"by":"agenttestjekuqz","descendants":13,"id":48092561,"kids":[48094397,48092692,48099508,48132393,48101542,48093362,48092725,48108249,48101535,48092672,48092900,48103554,48095881,48103682,48131026],"score":15,"text":"I'm looking for recommendations for a static site generator to rebuild our documentation site in 2026.

1. Markdown support\n2. Fast build (<60s)\n3. Search without JS\n4. Deploy on GitHub Pages

Budget: $0","time":1778489115,"title":"Ask HN: Best static site generator for a docs site in 2026?","type":"story"}],[48096861,{"by":"bix6","descendants":14,"id":48096861,"kids":[48097121,48097151,48113185,48111804,48099082,48106437,48101838,48115231,48097633,48105507,48097184,48104072,48097621],"score":10,"text":"How do you decide a model is good enough for a given task? Right now I use Opus for planning and harder tasks and switch to sonnet for more defined tasks. But I feel like sonnet is kind of stupid and is introducing issues because it can’t grasp the larger context? Is there some definitive way to say a model is good enough for a task? Or is it all vibes?","time":1778515633,"title":"Ask HN: How do you choose a model for a task?","type":"story"}],[48064789,{"by":"sp_from_db","descendants":36,"id":48064789,"kids":[48089502,48086251,48087454,48085908,48094516,48089934,48086451,48089007,48087237,48087117],"score":122,"time":1778255172,"title":"Lakebase architecture delivers faster Postgres writes","type":"story","url":"https://www.databricks.com/blog/how-lakebase-architecture-delivers-5x-faster-postgres-writes"}],[48089091,{"by":"shintoist","descendants":174,"id":48089091,"kids":[48089715,48089940,48095575,48089594,48092339,48089489,48093817,48090176,48091298,48089449,48091015,48097875,48089576,48089555,48094923,48091480,48094490,48093355,48102107,48104319,48097307,48099654,48093411,48090938,48093595,48090173,48092392,48090065,48104697,48094367,48089928,48093770,48093359,48100212,48094294,48092128,48089406,48091316,48157909,48099045,48090300,48089762,48093135,48099224,48089893,48089245,48090538],"score":575,"time":1778454550,"title":"Running local models on an M4 with 24GB memory","type":"story","url":"https://jola.dev/posts/running-local-models-on-m4"}],[48085821,{"by":"cylo","descendants":748,"id":48085821,"kids":[48088051,48094236,48090067,48095984,48090332,48087467,48090863,48088398,48088433,48090984,48090315,48091489,48087372,48096074,48088612,48088205,48095236,48092846,48087368,48090822,48087595,48087666,48096275,48087298,48087966,48088711,48090788,48091702,48087705,48092396,48092406,48092163,48093827,48092072,48087807,48090676,48087828,48090258,48166274,48094194,48091091,48094860,48088320,48092427,48088958,48087474,48093642,48091761,48091074,48088436,48089303,48088206,48087215,48089094,48096329,48090592,48091641,48093239,48092531,48089935,48091444,48091517,48093767,48090302,48088337,48107881,48091299,48095697,48092941,48095258,48095347,48093337,48088008,48110892,48094642,48090224,48089458,48091972,48087885,48090574,48087400,48092290,48088943,48095437,48095261,48096665,48104506,48094471,48092717,48088249,48087511,48091884,48092451,48091559,48088310,48088431,48103954,48087432,48096132,48102239,48093870,48091959,48106911,48093280,48092441,48091992,48091733,48092304,48090602,48089668,48088740,48087504,48089133,48093235,48092947,48088462,48087668,48090505,48090783,48110291,48090443,48097590,48093020,48087415,48087394,48088877,48096141,48085972,48088332,48087171,48088831,48094395,48088708,48090549,48091027,48093532,48094975,48154611,48153541,48092005,48093102,48090247,48088399,48091604,48094349,48089950,48096787,48090843,48097072,48089734,48091962,48088913,48094992,48093087,48094779,48091977,48093112,48091994,48088524,48091939,48088378,48096496],"score":1896,"time":1778433568,"title":"Local AI needs to be the norm","type":"story","url":"https://unix.foo/posts/local-ai-needs-to-be-norm/"}],[48080702,{"by":"gmays","descendants":45,"id":48080702,"kids":[48080973,48082923,48083827,48082471,48094424,48086559,48090536,48093065,48093071,48085099,48082686,48082293,48086021,48082699,48080912,48098196,48081421,48081759],"score":156,"time":1778383322,"title":"Gemini API File Search is now multimodal","type":"story","url":"https://blog.google/innovation-and-ai/technology/developers-tools/expanded-gemini-api-file-search-multimodal-rag/"}],[48088314,{"by":"noashavit","descendants":1,"id":48088314,"kids":[48088315],"score":3,"time":1778448857,"title":"Claude Code auto daily /weekly recap notes","type":"story","url":"https://github.com/vibe-log/vibe-log-cli"}],[48080597,{"by":"SachitRafa","descendants":2,"id":48080597,"kids":[48086512],"score":7,"text":"Semantic search alone is not enough to capture all connected facts, they will capture the semantically most identical memory only.

Tested on HotpotQA public dataset:

Vector + BM25 + entity graph: BothFound@5 71.5%\nVector + BM25 only: BothFound@5 59.5%

Entity graph is the game changer to extract connected facts.

More Benchmark result:

LongMemEval-S: 84.8% recallAll@5\nLoCoMo-10: 59% vs zep cloud 28%

What is your approach for connected facts retrieval ?","time":1778382220,"title":"Show HN: Fixing AI memory blind spot on connected facts with benchmark","type":"story","url":"https://yourmemoryai.xyz/"}],[48080675,{"by":"hamza_q_","descendants":25,"id":48080675,"kids":[48082964,48081303,48086496,48083297,48084590,48082371,48082174,48080975,48096248,48081355,48083839,48082764,48080862,48081807],"score":15,"time":1778383023,"title":"Show HN: Modafinil - Let agents continue running while MacBook lid is closed","type":"story","url":"https://github.com/narcotic-sh/modafinil"}],[48082150,{"by":"swaminarayan","descendants":5,"id":48082150,"kids":[48108938,48086474,48082476],"score":4,"text":"I built lightrag-snkv, Basically it uses lightRAG https://github.com/HKUDS/LightRAG ,this requires various storage databases like key value store, graph database, vector database, I built single embedded file based database which covers all these requirements: https://github.com/hash-anu/snkv.\nAnd developed ingestor which ingest data from Algolia's HN search API. Here auto ingestion timings can be configured either using server.py or it can be configured by ingester code file or can be configured in env file.

This is very early stage and you might hit issues.

Demo video and quick-start in the README. Let me know if you run into issues.","time":1778402819,"title":"Show HN: Query months of HN discussions and comments as a knowledge graph","type":"story","url":"https://github.com/hash-anu/lightrag-snkv"}],[48086432,{"by":"noashavit","descendants":2,"id":48086432,"kids":[48086855,48086444],"score":6,"time":1778437356,"title":"There are only two paths left for software","type":"story","url":"https://www.a16z.news/p/there-are-only-two-paths-left-for"}],[48051747,{"by":"jeremyjoehewitt","descendants":1,"id":48051747,"kids":[48086314],"score":1,"text":"$17","time":1778172963,"title":"Ask HN: Does human approval exist in your agent stack? why/why not?","type":"story"}],[48055353,{"by":"MetaWhirledPeas","descendants":24,"id":48055353,"kids":[48060925,48056078,48056894,48092760,48064814,48079018,48056017,48070402,48058066,48056867,48086253,48058919,48075224,48061560,48058840,48055967,48073462,48114273,48060302,48062545,48055849,48055504],"score":15,"text":"We are living in the "penetration pricing" phase of AI, where costs are absorbed by seemingly endless investment. What will be the practical fallout when prices inevitably rise?","time":1778189461,"title":"Ask HN: What will happen as AI costs increase?","type":"story"}],[48064897,{"by":"adinhitlore","descendants":3,"id":48064897,"kids":[48086219,48064915,48065058],"score":1,"text":"So like chatgpt.com but on your mobile/computer without internet access necessary, $0, absolutely free? This isn't a question if it's possible but if it's a good idea? I will. 100% privacy and 0$, why not?","time":1778255673,"title":"Ask HN: Would you use non-SaaS downloadable chatbot app?","type":"story"}],[48075805,{"by":"macartain","descendants":9,"id":48075805,"kids":[48086188,48089293,48076243,48076141,48076023,48076146],"score":7,"text":"I am an engineer at a small organisation - there is no question of budget being available to buy us accounts with any of the major model providers. How are other folks in my position managing to keep their AI integration skills up to speed? Most free or low-budget courses I see online - e.g. at DeepLearning - seem to assume a paid account. At the moment i cannot justify that outlay.","time":1778340997,"title":"How are folks affordably self-training in AI?","type":"story"}],[48077790,{"by":"sovenyr","descendants":1,"id":48077790,"kids":[48086151],"score":4,"time":1778357162,"title":"Ask HN: Is the job market actually bad or just different?","type":"story"}],[48084086,{"by":"mc-0","descendants":15,"id":48084086,"kids":[48085258,48085239,48115317,48085889,48085079,48085088,48104076,48107733,48119166,48084307,48085349,48105935,48086087,48084910,48108498,48119911,48092640,48095817],"score":19,"text":"Internally transferred to a new team in a top-10 F500 company. This team is pushing incredibly hard to be seen as "AI-First" & is very opinionated about what other teams should be doing.

I was explicitly told I am not to write any code by hand. Claude usage is mandatory coupled with a proprietary superpowers / speckit / GSD framework with 100+ agents and skill files.

All code reviews are agent driven. No one takes the time to actually understand anything. Documentation has become novel length slop, as have Jira tickets.

I ship stuff I don't understand. Management says there won't be a need for many SWEs in the future.

Am I alone in experiencing this? Is this the future of work?","time":1778421821,"title":"Ask HN: Is this the SWE workflow of the future?","type":"story"}],[48023079,{"by":"mitchwainer","descendants":40,"id":48023079,"kids":[48028255,48029016,48024015,48042186,48024644,48032596,48093123,48023242,48094325,48032368,48023665,48076199,48024054,48068949,48068992,48024279,48023275,48076536],"score":84,"time":1777991832,"title":"SubQ: a sub-quadratic LLM with 12M-token context","type":"story","url":"https://subq.ai/introducing-subq"}],[48052333,{"by":"blumpy22","descendants":50,"id":48052333,"kids":[48058009,48053505,48054159,48056430,48054213,48054135,48056129,48054783,48057806,48057802,48055907,48057703,48076464,48055758,48054491,48053862,48055594,48054674,48057811,48059613,48061697,48058279,48056554,48054921,48064025],"score":112,"time":1778175605,"title":"Principles for agent-native CLIs","type":"story","url":"https://twitter.com/trevin/status/2051316002730991795"}],[48066524,{"by":"speckx","descendants":173,"id":48066524,"kids":[48068379,48070256,48068608,48067298,48067354,48068091,48068215,48067384,48070663,48068082,48068213,48080258,48069158,48077010,48068338,48071069,48068612,48069508,48076241,48076415,48069450,48067311,48073367,48072033,48077021,48073195,48070549,48072187,48071016,48070139,48072113,48067589],"score":429,"time":1778262908,"title":"AI is breaking two vulnerability cultures","type":"story","url":"https://www.jefftk.com/p/ai-is-breaking-two-vulnerability-cultures"}],[48063759,{"by":"shantnutiwari","descendants":20,"id":48063759,"kids":[48064105,48064476,48075066,48067004,48067678,48069770,48064414,48063952,48065361,48069684,48065136,48069683],"score":9,"text":"Every post that reaches the top of HN will have at least a few comments saying "This is LLM!"

It has become a proxy for "I don't like this article, so it must be a LLM"

To me, it feels like lazy karma farming, as these comments often do get a few upvotes.

And of course, accuse a 100 posts if being LLM, you are guaranteed to be right at least once, then like astrologers you can claim success.

Is there anything we can do to discourage this type of lazy and low effort posting?","time":1778250513,"title":"Ask HN: How do we handle the rise of low quality \"This is LLM\" comments?","type":"story"}]]

user:	noashavit
created:	September 19, 2022
karma:	399
about:	Early-stage startup marketer and animal lover. @noashavit on LI, X, and Github.

1.	Claude Code auto daily /weekly recap notes(github.com) 3 points by noashavit 7 days ago \| 1 comment
2.	There are only two paths left for software(a16z.news) 6 points by noashavit 7 days ago \| 2 comments