news.ycombinator.com
Show HN: Canary – Monitor any URL for changes, get AI-powered intelligencehttps://news.ycombinator.com/item?id=47557235Canary watches any URL and alerts you when something changes — competitor pricing, API endpoints, docs, job boards, status pages. Changes are scored 1-10 by signal strength and summarised by AI. Built solo. Landing page live today. Automated crawler running on GitHub Actions posting real detected changes to Bluesky hourly and X daily. canary-landing-one.vercel.app Happy to answer questions about the architecture — particularly proud of the URL deduplication approach where one crawl serves multiple subscribers.Mar 28, 2026 6:49 PM
spacepolicyonline.com
U.S.-Canadian Astronaut Crew Ready for Flight Around the Moonhttps://spacepolicyonline.com/news/u-s-canadian-astronaut-crew-ready-for-flight-around-the-moon/The four astronauts who will fly around the Moon as soon as next week and their backups arrived at Kennedy Space Center this afternoon. Full of enthusiasm, they also conveyed […]Mar 27, 2026 10:23 PM
understudy-ai.github.io
Show HN: I made an AI that reviews iPhone apps – 1h of autonomous GUI workhttps://understudy-ai.github.io/understudy/I've been building Understudy, an open-source GUI agent for macOS. Wanted to push the GUI stuff beyond the usual short demos, so I tried turning it into an iPhone app reviewer.You give it one prompt. It browses the real App Store in Chrome, installs the app on a real iPhone through macOS iPhone Mirroring (not a simulator), opens the app and explores it — never seen Snapseed before — records clips and screenshots, composites a narrated review video with FFmpeg locally, uploads it to YouTube, then deletes the app. About an hour, didn't touch the keyboard.The exploration part is what I'm happiest with. The agent reads the App Store description, goes "they say background removal works, let me try that," and then figures out an unfamiliar app on its own. It regrounds from the live screenshot every action, so unexpected dialogs or UI changes don't kill it.The reason it can sustain an hour of work: each of the 6 stages runs as a separate child session with its own context. You can't fit an hoMar 27, 2026 7:39 PM
nasa.gov
NASA’s Environment and Energy “Blue Marble” Awards Categorieshttps://www.nasa.gov/general/nasas-environment-and-energy-blue-marble-awards-categories/Category I: NASA Environmental Quality Award Recognizes excellence in environmental management and planning, including stewardship of natural and cultural resources. This category highlights achievements in compliance, conservation, remediation, communication, and environmental information management, and the development of strong stakeholder partnerships. Category II: NASA Award for Excellence in Project or Program Execution Honors efforts that reduce […]Mar 27, 2026 7:27 PM
github.com
Show HN: Anvil – Desktop App for Spec Driven Developmenthttps://github.com/zdenham/anvilVery excited to share Anvil. I built Anvil to take back control when working with parallel coding agents. It comes with one click worktree isolation, and first class spec support.Claude Code and similar coding TUIs are very eager to get into writing code, even before their human baby sitter fully understands the implication of what they are about to build.The core insight with Anvil is that it is much easier to write high quality code which matches the author's intent after iterating on an external plan with your agent.Align on the architecture, implementation, and verification strategy in a markdown file, then execution is pretty straightforward.This is not a new concept, but the user experience within TUI apps for this workflow is pretty shit. Claude creates non-semantic plan names like "aquamarine-owl" that are trapped within a single agent context. Spinning up multiple agents to check on different aspects of a plan is annoying and slow, and managing terminal tabs is pure hell.So I Mar 27, 2026 7:07 PM
cranki.app
Show HN: Cranki – Crosswords meet Anki flashcardshttps://cranki.appHi HN!I am sure most of you have heard of Anki flashcards? Using spaced-repetition is one of the best ways to learn more vocabulary when learning a language.However, I find flashcards super boring. I've been playing some crossword games in my target language, Spanish, but I wished that I could use my custom list of words that I've come across instead of random words.That gave me the idea to create this mini-app. It's super simple. Add your words and you get unlimited crosswords with spaced-repetition! If you get a word right you won't see it for the next few days.Works with most languages (I doubt it works with Arabic or Chinese for example).You can add words one by one or import a CSV (just make sure to follow the columns: word, answer)It's a PWA, so you should be able to install it via your browser and it should work offline!There's still some bugs and QoL things to add but let me know what you think!Mar 27, 2026 4:46 PM
simfic.net
Show HN: SimFic – A multi-agent narrative simulation for interactive fictionhttps://simfic.netHello HN,I'd like to share a hobby project I'm working on called SimFic: a multi-agent interactive fiction simulation engine. Many of you like to read. But have you wanted to do more than just follow through someone's story? What if you could step into their shoes, and play it out yourself in a simulated environment?The problem: simply prompting an AI chatbot (e.g. ChatGPT) directly with a world/story-building prompt and expecting a rich, non-linear output is fundamentally flawed; in the real world, information asymmetry, Theory of Mind, non-determinism, etc. affect how people think and act. Asking a single LLM like ChatGPT to mentally simulate these constraints is flawed because of Transformer attention, as it is omniscient by technical design. Furthermore, LLMs are heavily tuned to be helpful and finish things early. If you've tried writing a book with an LLM, you know that the result is laughably short and shallow. If we use an LLM as-is, every action will succeed, every path will bMar 27, 2026 3:07 PM
news.ycombinator.com
We checkpoint Chrome to NVMe and restore it in 500ms ()https://news.ycombinator.com/item?id=47543412We've been building browser infrastructure for AI agents. The problem everyone hits: you spin up Chrome, log into something, do some work, and then what?Keep the VM running at $30/month doing nothing, or kill it and lose the session.We wanted to freeze Chrome to disk when it's idle and bring it back exactly where it left off. Turns out this is hard. After a lot of trial and error with different browser builds and configurations, we got it working with Playwright's Chromium on Orb Cloud using CRIU.The result: navigate to Google, pick up some cookies, take a screenshot, freeze. Wait. Restore. Same cookies, same page, identical screenshot. We ran it back-to-back on Google, Wikipedia, and Hacker News — all pass.Orb-browser is the open source repo. It's a headless Chrome you can put to sleep. It costs nothing while sleeping, wakes in ~500ms, and everything is intact — cookies, localStorage, the page you were on. The browser doesn't know it was frozen.If you're running browser agents that arMar 27, 2026 2:57 PM
news.ycombinator.com
Show HN: The Economics of Builder Saturation in Digital Marketshttps://news.ycombinator.com/item?id=47542386I formalised something most of us already feel but rarely say out loud: making things easier to build doesn't make things easier to succeed with.Personal version: I've vibe-coded maybe 15 projects since the beginning of this year. Two are still alive. At work, our teams built hundreds of custom GPTs and dashboards. Handful survived. The failure mode was never "couldn't build it" - it was "nobody had the bandwidth to care."So I wrote a paper about it. It combines Herbert Simon's attention scarcity, free-entry IO models, superstar economics, and preferential attachment into one framework. The central result: equilibrium attention per builder = k/p (entry cost over monetisation rate), independent of market size. As AI drives k towards 0, that ratio vanishes regardless of how much the market grows. Free entry absorbs everything.Calibrated to the App Store (800K publishers, 38B downloads) the model matches observed concentration pretty well - top 1% get ~70% of downloads, quarter of apps unMar 27, 2026 1:26 PM
github.com
Show HN: Forkrun – NUMA-aware shell parallelizer (50×–400× faster than parallel)https://github.com/jkool702/forkrunforkrun is the culmination of a 10-year-long journey focused on "how to make shell parallelization fast". What started as a standard "fork jobs in a loop" has turned into a lock-free, CAS-retry-loop-free, SIMD-accelerated, self-tuning, NUMA aware shell-based stream parallelization engine that is (mostly) a drop-in replacement for xargs -P and GNU parallel.On my 14-core/28-thread i9-7940x, forkrun achieves:* 200,000+ batch dispatches/sec (vs ~500 for GNU Parallel)* ~95–99% CPU utilization across all 28 logical cores, even when the workload is non-existant (bash no-ops / `:`) (vs ~6% for GNU Parallel). These benchmarks are intentionally worst-case (near-zero work per task) because they measure the capability of the parallelization framework itself, not how much work an external tool can do.* Typically 50×–400× faster on real high-frequency low-latency workloads (vs GNU Parallel)A few of the techniques that make this possible:* Born-local NUMA: stdin is splice()'d into a shared memfd, theMar 27, 2026 12:12 PM
iran.smlz.io
Show HN: I got tired of biased (all) news so I made Claude cosplay each sidehttps://iran.smlz.ioGot tired of reading news with a single bias so I built a thing. SMLZ spins up multiple AI agents via Claude SDK, each locked into a different analytical worldview, sends them off to do their own research from their own sources, then synthesizes everything into an intelligence report. No editorializing, just claims tagged [CONFIRMED] / [REPORTED] / [ASSESSED] and presented through 7 different lenses simultaneously. First instance covers the Iran-US-Israel conflict. Each topic gets its own subdomain and is fully independently configurable, so theoretically I keep adding topics forever or until I get bored. Runs on a single Hetzner VPS because I'm not made of money. Each full pipeline run is ~$30–45 (seven agents, synthesis, 9-language translation, images). Planning to hook up per-topic crypto donation wallets that automatically trigger a refresh if someone wants to force one, otherwise I'm running it once a day. Theoretically I could keep adding agents. More lenses, more sources, more sMar 27, 2026 6:03 AM
bing.com
60+ social media marketing FAQs, answered by expertshttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d382c279f4cd283320d5b14c608a1&url=https%3A%2F%2Fwww.bellinghamherald.com%2Fnews%2Fbusiness%2Farticle315208703.html&c=7964463278679948794&mkt=en-usYou're Googling social media marketing questions at 10 p.m. because your boss asked about ROI and you didn't have an answer. Or maybe you're a business owner staring at flatlined engagement, wondering ...Mar 27, 2026 5:37 AM
bing.com
How To Find A Real Estate Agenthttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a89c0fdf0104de1ac0992f86ecbdca1&url=https%3A%2F%2Fwww.forbes.com%2Fadvisor%2Fmortgages%2Freal-estate%2Fhow-to-find-a-real-estate-agent%2F&c=9134227281203908745&mkt=en-usHave a question for Andrea Riquier or our other editors? Ask here for a chance to be featured in a story. Submit your question This form is protected by reCAPTCHA ...Mar 27, 2026 12:13 AM
news.ycombinator.com
A 100% serverless RAG that extracts complex tables better than NotebookLMhttps://news.ycombinator.com/item?id=47537026Standard RAG wrappers completely mangle structured data.Here is Google's NotebookLM vs. KRAG pipeline extracting the exact same nested tables: https://www.youtube.com/watch?v=sngsQ6HJnlYNotebookLM hallucinates rows. KRAG extracts the grid flawlessly.Under the hood:100% Serverless: Deploys instantly (Bring Your Own Keys, pay your own compute).100% Private: Built using slms which deploy serverlessly on modal gpus.Enterprise Grade: 3 native layers of encryption.100% Open Source & FreeGithub: https://github.com/SaurabhJain708/kragMar 26, 2026 11:08 PM
dataverse001.net
Show HN: Data that explains itself to Coding Agents (Bonus: free, BYOA Lovable)https://dataverse001.net/AxyU5_5vWmP2tO_klN4UpbZzRsuJEvJTrdwdg_gODxZJ.00000000-0000-0000-0000-000000000000Lately I've been working on a data format for information exchange and collaboration between coding agents that I've called 'InstructionGraph'. I'd love to hear your feedback.--What can I see or do?To really take it for a spin, tell your (isolated) coding agent: Get the JSON from `curl -sL https://dataverse001.net` and follow the instructions to boot the dataverse! Ask it to create an app — any app. Wait a few minutes, and it will give you a shareable URL. If your app needs a back-end database, it will automatically integrate the Dataverse login method, and use the InstructionGraph as a database. This is essentially a free, bring-your-own-agent, optionally self-hostable Lovable.The 'root node' is self-describing, so let it describe itself to you: https://dataverse001.net/AxyU5_5vWmP2tO_klN4UpbZzRsuJEvJTrdw...Come say Hi in the dataverse dev forum: https://dev.dataverse001.net — a tiny social network built fully within InstructionGraph. If you login on the webpage, it deterministically Mar 26, 2026 9:18 PM
github.com
Show HN: Turbolite – a SQLite VFS serving sub-250ms cold JOIN queries from S3https://github.com/russellromney/turboliteI built a SQLite VFS in Rust that serves cold queries directly from S3 with sub-second performance, and often much faster.It’s called turbolite. It is experimental, buggy, and may corrupt data. I would not trust it with anything important yet.I wanted to explore whether object storage has gotten fast enough to support embedded databases over cloud storage. Filesystems reward tiny random reads and in-place mutation. S3 rewards fewer requests, bigger transfers, immutable objects, and aggressively parallel operations where bandwidth is often the real constraint. This was explicitly inspired by turbopuffer’s ground-up S3-native design. https://turbopuffer.com/blog/turbopufferThe use case I had in mind is lots of mostly-cold SQLite databases (database-per-tenant, database-per-session, or database-per-user architectures) where keeping a separate attached volume for inactive database feels wasteful. turbolite assumes a single write source and is aimed much more at “many databases with bursty Mar 26, 2026 6:58 PM
bing.com
The Best Vintage Stores in Los Angeleshttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d2842342b4395b5cc53d4c8082a83&url=https%3A%2F%2Fwww.vogue.com%2Farticle%2Fbest-vintage-stores-in-los-angeles&c=13348875127857304203&mkt=en-usVogue’s guide to the best vintage stores in Los Angeles is part of our directory of the very best vintage around the world, curated by editors from all over. Whether you’re traveling and searching for ...Mar 26, 2026 4:53 PM
usefolio.ai
Show HN: An unstructured data workspace for data transformations with LLMhttps://www.usefolio.ai/hi HN!a couple of months ago I had to analyze a few thousand audio recordings to help identify issues with customer support. i was able to get some raw high-level initial results with python scripts invoking LLM APIs, but they were too general and unhelpful. writing basic prompts is easy, but tuning them and making them specific enough to ensure no faint signal is missed is hard. you need to iterate through the data with an initial prompt, segment the data into different buckets, chain another prompt for each bucket etc. Then you need to constantly review the raw data to tweak the prompts just the right way to get the desired results.There are no good user-facing tools for scaling to thousands of rows of unstructured data analysis with LLMs. Claude Cowork / agents with access to filesystems are scratching the surface, but having a text-only UI is challenging, especially when you want to go back and adjust your research pipeline, narrow down deterministically to a specific subset of youMar 26, 2026 4:12 PM
news.ycombinator.com
Ask HN: Are there any merchant of records that don't dox you?https://news.ycombinator.com/item?id=47531873I am making a website and I want to start charging $5 a month for a subscription but I do not want my users to be able to trivially find my full name and street address from a quick online search (basic privacy desire).I know if you form an LLC then people can just easily find your full name and street address from required annual state reports (even if you register in "private" states like Wyoming since you need to also register in your home state if that is your center of operations), so I don't want to do that.Similarly, if you register a "Doing Business As" name and just use Stripe normally, then someone can just take 5 seconds and search your DBA and find out your full name and street address since you have to register it with your state.Merchants of record like Paddle and Stripe Managed Payments seem potentially useful here since users on their credit card statements will just see the merchant instead of your DBA or legal name, but it seems like Stripe also requires you to publicMar 26, 2026 3:41 PM
intuitivecompute.com
Show HN: Kora – An AI-native OS layer written in 370k lines of Rusthttps://intuitivecompute.comWe've been building this for ourselves over the last year. We wanted the computer from Star Trek, something you just talk to, that knows your context, controls your environment, and runs locally.Our core belief is digital sovereignty in that you should own your context. Your conversations, files, voice, memory, and identity never leave your machine unless you explicitly choose to route through a cloud provider. No telemetry, no accounts, no data collection. The AI that knows the most about you should be the one you control completely.Kora is an operating system layer built around an AI agent. On Linux it runs as the GUI directly in Wayland (OpenGL via femtovg). On macOS it's a native app with a Metal compositor.It's not a chatbot wrapper, it's closer to a full OS where the AI is a first-class service. The stack is ~8 services running locally:- UI: custom window manager with guest app compositing. Apps are standalone Rust binaries that communicate with the host over IPC.- Multi-device: Mar 26, 2026 3:18 PM
