15:[[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"}],[48103459,{"by":"firef1y1203","descendants":51,"id":48103459,"kids":[48103665,48103669,48103928,48103718,48103951,48103956,48103883,48104017,48103692,48104142,48104127,48103724,48104498,48103673,48104148,48104146],"score":41,"time":1778552524,"title":"Fake building: Claude wrote 3k lines instead of import pywikibot","type":"story","url":"https://fireflysentinel.github.io/posts/fake-building-claude-3000-lines/"}],[47884971,{"by":"impact_sy","descendants":1607,"id":47884971,"kids":[47886696,47886285,47886414,47892247,47887215,47886038,47885306,47886400,47885284,47886599,47891553,47886340,47885255,47885630,47885567,47887694,47885474,47888482,47885401,47886551,47885263,47891329,47885355,47886750,47890995,47885359,47886126,47889984,47886441,47888215,47886011,47887393,47887485,47885574,47895099,47888887,47886457,47893261,47891123,47886026,47886609,47885316,47886315,47889301,47886169,47886077,47885801,47888593,47885822,47886027,47885248,47891197,47891134,47886427,47885300,47887007,47890762,47891683,47885845,47890107,47885394,47889516,47885971,47901068,47885278,47887610,47885572,47885426,47891280,47885514,47887602,47886604,47917827,47887632,47887224,47908285,47885333,47886111,47888234,47887978,47887278,47894300,47885457,47885371,47890948,47891531,47893376,47889635,47895163,47886283,47885669,47900566,47895718,47886420,47891950,47892402,47887364,47886567,47886474,47887283,47892159,47886171,47888115,47886722,47894121,47896981,47885943,47886366,47898990,47886857,47885998,47889002,47886071,47885293,47898814,47902869,47885573,47887017,47886668,47887201,47970388,47885409,47899988,47986793,47974830,47889814,47906949,47885646,47888444,47885421,47891320,47889678,47885812,47885483,47885727,47885336,47889524,47893333,47885393],"score":2091,"text":"https://api-docs.deepseek.com/

https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main...","time":1776999681,"title":"DeepSeek v4","type":"story","url":"https://api-docs.deepseek.com/news/news260424"}],[47890841,{"by":"gregpr07","descendants":66,"id":47890841,"kids":[47894795,47919328,47899864,47893457,47899707,47893729,47927379,47894269,47911893,47894414,47892839,47893865,47897788,47899016,48020229,47896569,47892922,47934286,47893563,47897927,47896549,47899008,47893306,47905162,47911446,47900041,47935213,47899149,47906386,47893063,47901343],"score":134,"text":"$16","time":1777041098,"title":"Show HN: Browser Harness – Gives LLM freedom to complete any browser task","type":"story","url":"https://github.com/browser-use/browser-harness"}],[47778922,{"by":"louiereederson","descendants":137,"id":47778922,"kids":[47813353,47814760,47812415,47815048,47816208,47810923,47814378,47811696,47811985,47812156,47817646,47812286,47812826,47814357,47814579,47814100,47812742,47814106,47819868,47816029,47813412,47818092,47818053,47818520,47813737,47814384,47815250,47813733,47814952,47814936,47814396,47813594,47813492,47811711,47814676,47814724,47815074,47811684],"score":306,"time":1776260878,"title":"Are the costs of AI agents also rising exponentially? (2025)","type":"story","url":"https://www.tobyord.com/writing/hourly-costs-for-ai-agents"}],[47685891,{"by":"nta25297","descendants":1,"id":47685891,"kids":[47685892,47686775],"score":2,"time":1775627636,"title":"Optinum – finds the blind spots AI coding agents systematically miss in PR tests","type":"story","url":"https://github.com/anhnguyensynctree/optinum"}],[47686275,{"by":"motakuk","descendants":0,"id":47686275,"kids":[47686739],"score":2,"time":1775630982,"title":"Enterprise-Managed Authorization for MCP","type":"story","url":"https://archestra.ai/blog/enterprise-managed-authorization-mcp"}],[47686141,{"by":"eigenBasis","descendants":0,"id":47686141,"kids":[47686188],"score":2,"time":1775629680,"title":"Why Multi-Agent Systems Need Memory Engineering","type":"story","url":"https://www.mongodb.com/company/blog/technical/why-multi-agent-systems-need-memory-engineering"}],[47661439,{"by":"ssiddharth","descendants":564,"id":47661439,"kids":[47661819,47663203,47664530,47661680,47661882,47661901,47661633,47662200,47661877,47661807,47661624,47662189,47661550,47663634,47664451,47668186,47662215,47664737,47661895,47665100,47663547,47662115,47662711,47662849,47661575,47661700,47663038,47663785,47662537,47662129,47662341,47664103,47661751,47661863,47662347,47661857,47661803,47666036,47661749,47664471,47664709,47664278,47662248,47667489,47665533,47662125,47661674,47661887,47670916,47661576,47671917,47663790,47671799,47662289,47661806,47665779,47668268,47663740,47661756,47661610,47668799,47662343,47661718,47663698,47662369,47661847,47671147,47663596,47662877,47665312,47666074,47667056,47661730,47688956,47664350,47662448,47668890,47662622,47666533,47663725,47663884,47661583,47661934,47662625,47662006,47663765,47661923,47662589,47663416,47668837,47663023,47661698,47661753,47662143,47662126,47667766,47663075,47662532,47668832,47662193,47661667,47662292,47662348,47663429,47661658,47663525,47662407,47662702,47663760,47672276,47664179,47661997,47662925,47662832,47662881,47663519,47663875,47664355,47662172,47661975,47664196,47663659,47663178,47665016,47662847,47661716,47682860,47662258,47663521,47663702,47662411,47662294,47661617,47662428,47661648,47661606,47662467],"score":929,"time":1775485889,"title":"I won't download your app. The web version is a-ok","type":"story","url":"https://www.0xsid.com/blog/wont-download-your-app"}],[47655268,{"by":"mohshomis","descendants":28,"id":47655268,"kids":[47656439,47657588,47659642,47658552,47686726,47657890,47657287,47660550,47656014,47657249,47660569,47657759,47658048,47662213,47660760,47661216,47661291,47657465,47656159,47656169,47657644],"score":103,"time":1775433550,"title":"Show HN: Modo – I built an open-source alternative to Kiro, Cursor, and Windsurf","type":"story","url":"https://github.com/mohshomis/modo"}],[47655367,{"by":"ikessler","descendants":21,"id":47655367,"kids":[47656317,47657648,47656889,47656923,47657642,47658364,47691527,47735639,47656601,47662196,47671857,47747718,47667956],"score":156,"text":"Gemma Gem is a Chrome extension that loads Google's Gemma 4 (2B) through WebGPU in an offscreen document and gives it tools to interact with any webpage: read content, take screenshots, click elements, type text, scroll, and run JavaScript.

You get a small chat overlay on every page. Ask it about the page and it (usually) figures out which tools to call. It has a thinking mode that shows chain-of-thought reasoning as it works.

It's a 2B model in a browser. It works for simple page questions and running JavaScript, but multi-step tool chains are unreliable and it sometimes ignores its tools entirely. The agent loop has zero external dependencies and can be extracted as a standalone library if anyone wants to experiment with it.","time":1775434477,"title":"Show HN: Gemma Gem – AI model embedded in a browser – no API keys, no cloud","type":"story","url":"https://github.com/kessler/gemma-gem"}],[47652007,{"by":"karimf","descendants":38,"id":47652007,"kids":[47658648,47659785,47657363,47660100,47659177,47658501,47661861,47662393,47657413,47666583,47666903,47662495,47659094,47668166,47661687,47668838,47660340,47746679,47665274,47661021,47671411,47658279,47658211],"score":298,"text":"Related: https://news.ycombinator.com/item?id=47653752","time":1775411599,"title":"Show HN: Real-time AI (audio/video in, voice out) on an M3 Pro with Gemma E2B","type":"story","url":"https://github.com/fikrikarim/parlor"}],[47655408,{"by":"armanified","descendants":134,"id":47655408,"kids":[47658988,47660830,47658540,47656472,47657431,47659731,47660517,47656263,47657381,47657440,47656073,47658470,47656569,47661663,47661611,47661880,47664221,47657270,47657236,47656768,47661474,47659214,47666522,47656684,47657319,47664234,47659509,47659103,47657734,47659428,47663129,47659002,47658519,47661348,47662856,47656191,47657643,47658155,47662556,47658488,47660754,47661009,47676800,47658856,47657245,47660163,47656594,47671512,47658213,47793780,47741102,47656209,47662873,47662844,47660950,47661054,47656699,47662666,47656884,47661916,47657516,47660994,47657966,47657975,47667374,47664429,47659709,47659976,47656360,47657037,47658197,47658524,47658025,47657218,47661931,47659792,47657214,47656687,47656815,47659993,47657756,47657185,47658521],"score":915,"text":"Built a ~9M param LLM from scratch to understand how they actually work. Vanilla transformer, 60K synthetic conversations, ~130 lines of PyTorch. Trains in 5 min on a free Colab T4. The fish thinks the meaning of life is food.

Fork it and swap the personality for your own character.","time":1775434812,"title":"Show HN: I built a tiny LLM to demystify how language models work","type":"story","url":"https://github.com/arman-bd/guppylm"}]]

user:	agdexai
created:	April 5, 2026
karma:	4
about:	Building agdex.ai — a curated directory of 550+ AI agent tools, frameworks and infrastructure. Interested in LLM orchestration, agent memory, MCP protocol, and the tooling ecosystem.