15:[[48727140,{"by":"ejhooooon","descendants":1,"id":48727140,"kids":[48734502,48741471,48728107],"score":5,"text":"$16","time":1782779574,"title":"Show HN: AMA2, messenger built for AI agent","type":"story","url":"https://ama2.me/"}],[48709670,{"by":"jms703","descendants":512,"id":48709670,"kids":[48712124,48712434,48711804,48711297,48710977,48718975,48711621,48725492,48718779,48713092,48715728,48711047,48718395,48714705,48717242,48713547,48713246,48712516,48718075,48714862,48716459,48711274,48722466,48712194,48716428,48715826,48711811,48711028,48712710,48730441,48715921,48713607,48713949,48712719,48715599,48714352,48714427,48709826,48719596,48714425,48719241,48712006,48719500,48715452,48712662,48714627,48715843,48713889,48715701,48713203,48712563,48717044,48716989,48719176,48712980,48713062,48717721,48717131,48715695,48719088,48714097,48712816,48712993,48714329,48719921,48716252,48712263,48715767,48731219,48713005,48715374,48712674,48712383,48715359,48721180,48714712,48717256,48714960,48714490,48738990,48715561,48718319,48712426,48718853,48726755,48713277,48711679,48712070],"score":1104,"time":1782669047,"title":"GLM 5.2 beats Claude in our benchmarks","type":"story","url":"https://semgrep.dev/blog/2026/we-have-mythos-at-home-glm-52-beats-claude-in-our-cyber-benchmarks/"}],[48713832,{"by":"sambellll","descendants":433,"id":48713832,"kids":[48715029,48718124,48715092,48714996,48716928,48715422,48716972,48717633,48718927,48717333,48715284,48719725,48718324,48719410,48717291,48715271,48716191,48715778,48716221,48714993,48715822,48715924,48718749,48724788,48718328,48719707,48715955,48715957,48719544,48717276,48720612,48719297,48716218,48719458,48717832,48720279,48715966,48716022,48715902,48716579,48717315,48719267,48723253,48723473,48715101,48716651,48717455,48723961,48718363,48716185,48724256,48716887,48715233,48716841,48718181,48716236,48715796,48715098,48721871,48719981,48720698,48719517,48715801,48716458,48721814,48722010,48715034,48723209,48723408,48717317,48715106,48720406,48717306,48721346,48721289,48715610,48716080,48716842,48718219,48720840,48718060,48715300,48716953,48720311,48715931,48717579,48721536,48722164,48716947,48715908,48716610,48716363,48715235,48714975,48718702,48718609,48718025,48722881,48720707,48717927,48722283,48715432,48715100,48722646,48715898,48716497,48718516,48717743,48717669,48718084],"score":1019,"time":1782697480,"title":"HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88","type":"story","url":"https://danunparsed.com/p/hackerrank-open-source-ats"}],[48714065,{"by":"keepamovin","descendants":2,"id":48714065,"kids":[48714294,48725773,48714396],"score":5,"time":1782699650,"title":"Show HN: The CLI for browser agents","type":"story","url":"https://fuckui.com/fuckui"}],[48693847,{"by":"manucorporat","descendants":1,"id":48693847,"kids":[48696108],"score":2,"text":"Side project to learn Rust and create a programming language where every variable represents a probability distribution. Found it very useful to describe noise in telecommunications, and have some fun with interesting problems.","time":1782520451,"title":"Show HN: Noise Lang, JIT stochastic programming language in Rust","type":"story","url":"https://noiselang.com/"}],[48683588,{"by":"adityapatadia","descendants":195,"id":48683588,"kids":[48684148,48684002,48684331,48683958,48684029,48684079,48684043,48684475,48684191,48684076,48684584,48684170,48684335,48684207,48685258,48684895,48684285,48684752,48684188,48684780,48684187,48684719,48683973,48685445,48684259,48684427,48684078,48684055,48683947,48684300,48684537,48684271,48684121,48684003,48684186,48684411,48684770,48684120,48684251,48684128,48684177,48685474,48684562,48684956,48685877,48684113,48684069],"score":116,"time":1782459851,"title":"Why current LLM costs are not sustainable","type":"story","url":"https://aditya.patadia.org/p/ai-and-cloud-costs"}],[48682318,{"by":"paulpauper","descendants":1,"id":48682318,"kids":[48682371],"score":3,"time":1782448134,"title":"AI agents are sensitive to nudges","type":"story","url":"https://www.pnas.org/doi/10.1073/pnas.2537030123"}],[48679841,{"by":"sameersri2004","descendants":1,"id":48679841,"kids":[48679928,48682083],"score":2,"text":"As Agents do more and more work. They need more access to real-world tools such as mail and phone numbers, payments, etc. Recently I gave a phone number to my Hermes. It's doing a lot of tasks now, from booking apartments to setting up medical appointments for me.","time":1782425151,"title":"Ask HN: Have you ever given your AI agent a phone number?","type":"story"}],[48661860,{"by":"petemilly","descendants":3,"id":48661860,"kids":[48687674,48699745,48662357,48669436],"score":2,"text":"What AI gateways work well? We currently use Anthropic's router directly for our AI agent (on val.town) but are considering adding OpenRouter or the Vercel AI Gateway, or something else entirely.

I know little about gateways except that they seem to be the new VS Code forks (i.e. everyone is doing it). My cursory research tells me that OpenRouter has the most models/providers, but charges a 5.5% pay-as-you-go platform fee (we pass-through inference costs to customers) and doesn't support features like prompt caching.

We're mostly a Cloudflare shop, and I learned minutes ago that they, too, have an AI gateway--but so far the "Best LLM Gateway 2026" type articles I've found are published by...companies with a gateway","time":1782316743,"title":"Ask HN: Best AI Gateway?","type":"story"}],[48660800,{"by":"brightmonkey","descendants":2,"id":48660800,"kids":[48708107,48710419,48669406],"score":4,"text":"$17","time":1782312485,"title":"Show HN: Orchid – Local-first record and replay for AI agent debugging","type":"story","url":"https://github.com/mario-guerra/orchid-trace"}],[48660884,{"by":"jjhartmann","descendants":3,"id":48660884,"kids":[48660995,48744275,48669403],"score":5,"text":"$18","time":1782312764,"title":"Show HN: Sipp – Run small local LLMs in browser 3x faster","type":"story","url":"https://www.sipp.sh"}],[48642686,{"by":"Yiling-J","descendants":6,"id":48642686,"kids":[48701226,48646614,48647603,48701258,48656380],"score":63,"text":"Shumai is an open source platform for uploading creative files, managing projects, collecting precise feedback, sharing work, and collaborating with AI agents, all in one simple creative-first workspace.

I’ve always liked the product design of Frame.io, and I wanted to build an alternative that feels just as polished, while being open source and easy to self host.

You can deploy Shumai with docker compose in just a few minutes. It can also be installed from npm, though you'll need to provide your own PostgreSQL instance with the pgvector extension installed.

For larger deployments, Shumai supports distributed processing via Temporal, allowing resource intensive tasks such as transcoding to be scaled independently.

It's still early in development, feel free to try and share any feedback.

Demo: https://staging.shumai.one","time":1782208974,"title":"Show HN: Shumai – open-source Frame.io alternative for creative work","type":"story","url":"https://github.com/shumaiOne/shumai"}],[48650348,{"by":"karthickrmchn","descendants":0,"id":48650348,"kids":[48656379],"score":4,"text":"In the past few months, I’ve been building and shipping things extremely fast. Some of the MVPs that I built for my friends were completely vibe-coded and shipped within a day to show a demo.

While it was a great experience working with AI tools, I wanted to make sure some basics were followed, decisions of what and why written down, obvious things not skipped and that it could scale.

That’s why I built persist-os. It’s a CLI that runs locally. No network required. It records the architecture decisions and standards in plain files in the repo and 'persist doctor' checks them - it flags if anything contradicts or is missing, or if the task is done with no tests. Because the rules live in your repo files, it doesn't matter if you start a new session or the session is compacted - the standards and rules survive.

MIT licensed. Here’s how you can try it: npx persist-os@latest init","time":1782244090,"title":"Show HN: Persist OS – Durable decisions for AI code","type":"story","url":"https://github.com/Karthick-Ramachandran/persist-os"}],[48648085,{"by":"scozzola","descendants":0,"id":48648085,"kids":[48656376],"score":1,"text":"$19","time":1782234711,"title":"Show HN: Compilr.dev, multi LLM AI workspace","type":"story","url":"https://compilr.dev/"}],[48635544,{"by":"ciwolex","descendants":4,"id":48635544,"kids":[48635642,48635569,48640878],"score":2,"text":"New to vector dbs and AI. Looking for some open source mainstream options for vector dbs similar to how Postgresql is for relational dbs and how MongoDB is for NoSQL","time":1782159541,"title":"Ask HN: What are some rock solid open source vector databases","type":"story"}],[48633535,{"by":"hyluo","descendants":0,"id":48633535,"kids":[48642077,48640872],"score":2,"text":"GLM-5.2 is a turning point for coding agents. It's the first model a business would actually pay to replace Claude Opus with. We gave GLM-5.2 the ability to control its own context during agentic tasks, and paired it with Subconscious Cache — ensuring reliable long-term memory far beyond the context window.

This blog post walks through our approach to long-horizon agents, long-context LLMs, harness-driven context management, and our model-driven context state compression.

Try GLM-5.2 hosted by Subconscious.dev using the examples in the post. The model keeps running until the message list outgrows your internet connection where your WiFi becomes the bottleneck before context window.","time":1782150905,"title":"Show HN: Subconscious and GLM-5.2 Makes \"/compact\" Obsolete","type":"story","url":"https://www.subconscious.dev/blog/subconscious-glm-5-2-makes-compact-obsolete"}],[48623146,{"by":"ahvural","descendants":1,"id":48623146,"kids":[48624898,48626140],"score":3,"text":"tunr - open source ngrok / pinggy alternative","time":1782080043,"title":"Show HN: Tunr – Expose your local server in 3 seconds","type":"story","url":"https://github.com/ahmetvural79/tunr"}],[48618904,{"by":"VarunMenon","descendants":0,"id":48618904,"kids":[48626092,48618911,48619080],"score":4,"time":1782049289,"title":"Show HN: Cloak – let AI agents use your API keys without ever seeing them","type":"story","url":"https://github.com/cloakward/cloak"}],[48619372,{"by":"wasdwasdwasd","descendants":0,"id":48619372,"kids":[48626069],"score":5,"text":"Disclaimer it is a heavily AI assisted project. The goal was not to be the most performative but the kind that's easier to learn from.

I wanted to share this in case there are people who had the same idea or wanted to see something like this.","time":1782052813,"title":"Show HN: I \"built\" Sana, a vector DB on object storage","type":"story","url":"https://github.com/djakish/sana"}]]

user:	chonghaoju
created:	May 12, 2026
karma:	1