news.ycombinator.com
Show HN: Attest – Test AI agents with 8-layer graduated assertionshttps://news.ycombinator.com/item?id=47122431I built Attest because every team I've seen building AI agents ends up writing the same ad-hoc pytest scaffolding — checking if the right tools were called, if cost stayed under budget, if the output made semantic sense. It works until the agent gets complex, then it collapses.60–70% of what makes an agent correct is fully deterministic: tool call schemas, execution order, cost budgets, content format. Routing all of this through an LLM judge is expensive, slow, and unnecessarily non-deterministic. Attest exhausts deterministic checks first and only escalates when necessary.The 8 layers: schema validation → cost/perf constraints → trace structure (tool ordering, loop detection) → content validation → semantic similarity via local ONNX embeddings (no API key) → LLM-as-judge → simulation with fault injection → multi-agent trace tree evaluation.Example: from attest import agent, expect from attest.trace import TraceBuilder @agent("support-agent") def support_agent(builder: TraceBuilder, uFeb 23, 2026 2:00 PM
news.ycombinator.com
Managed OpenClaw hosting, 60-second provisioninghttps://news.ycombinator.com/item?id=47121964I registered https://clawhosters.com 18 days ago. Wrote the first line of code that same day. The whole thing is live now with paying customers, so I figured it was time to share it here.The problem: deploying OpenClaw (the open-source AI agent framework) is annoying. Docker setup, Nginx config, SSL, firewall rules, Playwright browser installation for web tools, messenger bridge configuration. If you know what you're doing, it probably takes an hour. I got it down to 30-60 seconds.*The provisioning trick: prewarmed VPS*Snapshot-based provisioning alone gets you to about 3 minutes. Hetzner needs ~40s to create the VPS from a snapshot alone.The thing that actually makes it fast is prewarming. I keep 1-2 idle VPS per tier sitting ready. They're already booted from the snapshot, Docker is running, everything is warm. When a customer clicks "Create Instance," I claim a prewarmed server, SSH in, upload configs, start containers. 30-60 seconds, done.A background job monitors the pool and replFeb 23, 2026 1:19 PM
github.com
Show HN: Librarium – fan out research queries to many search/AI APIs in parallelhttps://github.com/jkudish/librariumI wanted deep research from inside my agents. Codex, Claude Code, and other AI coding agents don't have built-in deep research. They can search the web, but they can't fan out a query to Perplexity Deep, OpenAI Deep, Exa, Brave, etc.Librarium is a CLI that does exactly that. One command, many providers in parallel, structured markdown output. Because it's a CLI, any AI agent that can run shell commands can use it. I use it as a skill in Claude Code and it works out of the box.It also deduplicates citations across providers, so you can see which sources keep showing up independently, which is a useful trust signal when you're researching something you don't already know well.Feb 23, 2026 11:49 AM
news.ycombinator.com
Show HN: Coasty.ai – We just hit SOTA on OSWorld (computer-use agent benchmark)https://news.ycombinator.com/item?id=47117621Hey HN,We've been heads down building Coasty for the past few months and wanted to share a milestone, we just achieved state-of-the-art results on OSWorld, the most comprehensive benchmark for computer-use agents. For those unfamiliar, OSWorld tests whether AI agents can actually use a computer the way a human does navigating real desktop environments, clicking through UIs, filling out forms, managing files across apps.We're building an agent execution platform for computer-use AI. Think of it like Vercel but for AI agents that operate full desktop environments. You give Coasty a task, and it spins up a sandboxed environment where an agent can see the screen, move the mouse, type, and complete end-to-end workflows things like filing expenses, canceling subscriptions, filling out insurance forms, or navigating legacy enterprise software. Why this mattersMost "AI automation" today is really just API glue between apps that have APIs. But the vast majority of business workflows still live Feb 23, 2026 3:13 AM
news.ycombinator.com
Show HN: I built an async Telegram bot to snipe 80%+ retail price errorshttps://news.ycombinator.com/item?id=47116609Hi HN,I do some retail arbitrage and got incredibly frustrated missing out on massive 90% off glitches. By the time someone posts a deal on Slickdeals or DealNews, bots have already cleared the stock.I decided to fix my own problem and built Sniper Syndicate. It’s a custom Python Telegram bot that scrapes RSS feeds and Reddit in real-time. Instead of just forwarding links, I wrote some Regex logic to actually extract and calculate the discounts from raw HTML, dropping anything under a 20% true discount and filtering out sweepstakes or "back in stock" spam.It works beautifully for my own sourcing (it pings me the second a match is found), so I recently packaged the access via Whop: https://whop.com/snyper-sindicate/sniper-syndicate-membershi...I'm still improving how the bot handles rate limits at scale with aiohttp, so I'd love to hear your thoughts on the scraping logic or any ideas to make the price filters even better!Feb 23, 2026 12:49 AM
news.ycombinator.com
Show HN: A Vaadin Algebra and Calculus Solver Built with AI Assistancehttps://news.ycombinator.com/item?id=47116251Hi HN,I’d like to share a side project that has gradually become my main creative outlet: The Algebrator, a web-based algebra and calculus solver I built using Java, Spring Boot, Vaadin 24, and a symbolic math engine under the hood.It started as a personal attempt to revisit the math I loved in middle/high school, but it evolved into a multi-year design experiment in AI-augmented software engineering. I built and iterated on this app using a paired-programming workflow with LLMs (ChatGPT + GitHub Copilot), and the project ended up teaching me more about agentic AI development than anything else I’ve worked on.What it doesLets users type equations, inequalities, and expressions in a calculator-like UISolves algebraic equations, systems, trigonometry, calculus operations, and iterative “problem templates”Supports fraction/decimal modes, radians/degrees, comparison operators, user-defined functions, and multi-character variablesIncludes utility “extras” like prime generation, Fibonacci, rFeb 23, 2026 12:04 AM
news.ycombinator.com
Show HN: OctoFlow v1.0.0 – GPU VM where the GPU runs autonomously, CPU is BIOShttps://news.ycombinator.com/item?id=47113604Three days ago I posted OctoFlow 0.83 here (GPU-native programming language, 2.2 MB binary). The feedback was great. Since then I've pushed v1.0.0 with the thing I've actually been building toward: a GPU Virtual Machine.The idea: the GPU is the computer, the CPU is the BIOS.You boot a VM, program a dispatch chain of kernel instances, submit once with vkQueueSubmit, and everything — layer execution, inter-layer communication, self-regulation, compression, database queries — happens on the GPU without CPU round-trips. The CPU just provides I/O. let vm = vm_boot() let prog = vm_program(vm, kernels, 4) vm_write_register(vm, 0, 0, input) vm_execute(prog) let result = vm_read_register(vm, 3, 30) 4 VM instances, one submit, no CPU involvement between stages.The memory model is 5 SSBOs: Registers (per-VM working memory), Metrics (regulator signals), Globals (shared mutable — KV cache, DB tables), Control (indirect dispatch params), Heap (immutable bulk data — quantized weights).What makes it iFeb 22, 2026 6:55 PM
github.com
Show HN: Aethene – Open-source AI memory layerhttps://github.com/akhilponnada/aetheneHey HN, I'm shipping my first open-source project and I'm pretty nervous about it. Aethene is an AI memory API – it gives your AI apps persistent memory. Store conversations, extract facts automatically, search semantically, handle contradictions gracefully. It works well thank most of the memory projects available on the market currently.Why I built this: I was building AI agents and kept running into the same problem – they forget everything. Every conversation starts from zero. I wanted something that could: - Auto-extract facts from conversations (not just store raw text) - Handle "user moved from SF to NYC" without keeping both as true - Search by meaning, not just keywords - Version everything (who said what, when) Tech stack: - TypeScript + Hono (fast, edge-ready) - Convex (real-time DB + vector search) - Gemini (embeddings + extraction) What it does: # Store memory curl -X POST /v1/content -d '{"content": "User loves hiking, lives in SF"}' # Recall naturally curl -X POST /v1/reFeb 22, 2026 1:57 PM
github.com
Show HN: Pq – Simple, durable background tasks in Python using Postgreshttps://github.com/ricwo/pqAt work we were using python-rq for background tasks. It does the job for simple things, but we kept bumping into limitations. We needed to schedule tasks hours / days out and trust they'd survive a restart. We wanted periodic tasks with proper overlap control. So we built a scheduling / enqueuing system around Postgres to bring these durability capabilities to python-rq. This worked fine for a while but was trickier to reason about due to its more complicated architecture (we'd run two separate services just for getting jobs from Postgres into the rq Redis queue, plus N actual task workers).pq is a simpler approach: It's a Postgres-backed background-task library for Python, using SELECT ... FOR UPDATE SKIP LOCKED for concurrency safety. You just run your N task workers, that's it. The stuff I think is worth knowing about: - Transactional enqueueing -- you can enqueue a task inside the same DB transaction as your writes. If the transaction rolls back, the task never exists. This is theFeb 22, 2026 1:03 PM
cuto.video
Show HN: Cuto – Turn raw footage into viral shorts in 30s with an AI prompthttps://cuto.videoI built Cuto because I was tired of spending hours on the repetitive parts of video editing: splitting long footage, syncing audio and captions, filtering out bad takes, and adjusting the pacing. I just wanted to focus on the storytelling.Cuto is an AI video editing workspace that automates the grunt work. You drag and drop up to 12 video clips, write a quick prompt detailing what you want (e.g., "Find the best 30s product highlight and add a strong CTA"), and the engine does the rest.Within 30 seconds, it hands back an editable visual timeline with: - Auto-split shots and filtered out silences/bad takes - Synchronized captions with rhythm alignment - Selected highlight effects and custom watermark branding - Auto-generated social-ready copy for TikTok, Shorts, and ReelsUnlike traditional NLEs (Non-Linear Editors), we wanted an interface where the visual parameters are controlled intuitively by natural language first, giving you a solid 80% draft instantly. From there, you can still twFeb 22, 2026 10:11 AM
bing.com
Beat creative burnout: Get this AI content and marketing bundle for just $25http://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d9345414c45598c02b4c266ba713b&url=https%3A%2F%2Fwww.cultofmac.com%2Fdeals%2F2026-ai-powered-digital-marketing-content-creation-bundle&c=9962778480892777227&mkt=en-usThe 2026 AI-Powered Digital Marketing & Content Creation Bundle teaches creators, entrepreneurs and marketers how to use AI tools like ChatGPT and Canva to build scalable content workflows without ...Feb 21, 2026 10:40 PM
news.ycombinator.com
Show HN: A Product Commercialization Framework for Solo Developershttps://news.ycombinator.com/item?id=47101212Most solo developers spend 90% of their time writing code and 10% on everything else. After a few years of building indie products — some failures, a couple that found traction — I've tried to document the framework I wish I'd had earlier.The framework has four phases: clarify, build, grow, monetize. None of it is novel. The value is in applying it in order, because most failures happen when you skip the first phase and go straight to the second.Clarify first. Before writing code: who exactly are you solving a problem for, and what problem are you solving? The answer needs to be specific enough to function as a filter for future decisions. "Productivity software for remote workers" is not an answer. "A tool that helps freelance developers track billable time without switching contexts" is closer.Once you have that, validate the problem using a simple heuristic: Frequency × Intensity × Scale. A problem that's frequent, painful, and affects many people is real. A problem that scores highFeb 21, 2026 2:35 PM
wordy.info
Show HN: Wordy – Learn languages from real movie and TV clips with quizzeshttps://wordy.info/Hey HN, I'm Sándor. I built Wordy because I kept pausing Netflix to look up words in a dictionary. So I made a subtitle tool for myself, and it kind of snowballed from there.What it does: You watch short clips from real movies and TV shows, then get quizzed on the vocabulary you just heard. The app tracks every word automatically and adapts to your level. 20+ languages.How it evolved: Started as an iOS app for interactive streaming subtitles. Then I added curated video clips with quizzes. Then vocabulary tracking, spaced repetition, and personalized learning paths. Now it's iOS + Android + a Chrome extension (dual subtitles for Netflix and YouTube).The whole thing is a solo project. Along the way it won a $30K prize at Hungary's biggest startup competition, which gave me the push to go full-time on it.Stack: Swift (iOS), Kotlin (Android), Chrome extension.Would love to hear what you think, especially from anyone who's tried learning a language from media before. What worked? What didn'Feb 21, 2026 12:53 PM
github.com
Show HN: Elecxzy – A lightweight, Lisp-free Emacs-like editor in Electronhttps://github.com/kurouna/elecxzyHi HN. I am a programmer from Japan who loves Emacs. I am building elecxzy. It is a free (zero-cost), lightweight, Emacs-like text editor for Windows.I designed it to be comfortable and ready to use immediately, without a custom init.el. Here is a quick overview:- Provides mouse-free operation and classic Emacs keybindings for essential tasks (file I/O, search, split windows, syntax highlighting).- Drops the Lisp execution engine entirely. This keeps startup and operation lightweight.- Solves CJK (Chinese, Japanese, Korean) IME control issues natively on Windows.I never managed to learn Lisp. I just copy-pasted snippets to maintain my init.el. However, I loved the Emacs keybindings. I loved operating an editor entirely without a mouse. I wanted an editor I could just open and use immediately. Also, standard Emacs binaries for Windows often have subtle usability issues for CJK users.So, I thought about whether I could build an Emacs-like text editor using Electron, the same framework asFeb 21, 2026 12:49 PM
news.ycombinator.com
Facebook Is Doing A-Okayhttps://news.ycombinator.com/item?id=47096447Note: I don't own any equity in Meta, or work there, or have a soft corner for them or their competition in any business category.There's a way to view just your friends' and groups' updates without seeing all the AI doom scrolling stuff.They are the largest peer to peer marketplace.They are the most popular platform for organizing groups.Apparently Facebook dating is popular (maybe someone can comment on that).If anything, despite all the dark patterns on the Wall, they are only growing and finding new ways to grow more.Feb 21, 2026 1:25 AM
news.ycombinator.com
Show HN: Natural language search across Kalshi and Polymarket (API and MCP)https://news.ycombinator.com/item?id=47095815Hi HN, we built a better search for Kalshi and Polymarket. You can use it on the web, via API, or through MCP so AI agents can query it too.There are about 80k active contracts across both platforms right now. Sports, weather, crypto, politics. Tonight's NBA games, BTC up or down in 5 mins. You name it. We trade prediction markets everyday and needed a smarter way to search across all of them.We started by ingesting everything into Postgres and writing SQL to get insights. Then we plugged in Claude to query the DB for us. Type a question, get structured results. That actually worked and felt like what search should be. But the answers were only as good as the data, and the data was a mess.The core problem: both platforms structure their data completely differently. One NBA game on Kalshi is dozens of separate contracts, each with its own ticker. Polymarket has the same game as a handful of contracts with different naming. One says "Cleveland," the other says "Cavaliers." You can't searFeb 21, 2026 12:02 AM
streakout.app
Show HN: Minimalistic workout tracker for iPhone – Stats, Trends, Streakshttps://streakout.app/Streakout is a minimal iOS workout tracker designed to extend Apple Health & Fitness. It reads your existing workout data and turns it into beautiful tiles, charts, and timelines – helping you answer one of the most important questions for your health: “How often did I move this week, this month, this year…?”Because what really matters is showing up.There’s no signup, no clutter, no subscription, no data sharing, and no noise – only the most relevant stats and insights, neatly packaged into two simple screens: overview and calendar feed.Would deeply appreciate some feedback.LET'S MOVEFeb 20, 2026 3:06 PM
attena.xyz
Show HN: Natural language search across Kalshi and Polymarkethttps://www.attena.xyz/Hi HN, we built a better search for Kalshi and Polymarket.There are about 80k active contracts across both platforms right now. Real-time probability estimates for everything from tonight's NBA games to whether Bitcoin will be up in 5 minutes. But the data is fragmented with no unified way to search it.We trade prediction markets everyday and needed a smarter way to search across all of them. We started by ingesting everything into Postgres and writing SQL. Then we built a Claude wrapper that queries the DB in plain English. Type a question, get structured results. That actually worked and felt like what search should be. But the answers were only as good as the data, and the data was a mess.The core problem: Both platforms structure their data completely differently. One NBA game on Kalshi is dozens of separate contracts, each with its own ticker. Polymarket has the same game as a handful of contracts with different naming. One says "Cleveland," the other says "Cavaliers." You can't sFeb 20, 2026 2:48 PM
news.ycombinator.com
Ask HN: How did you get your drive back?https://news.ycombinator.com/item?id=47088256Feeling lost at the moment, and don't feel like repeating myself to an LLM for advice again so I thought I would maybe ask here.So I'm in my early 30's, I've built the career and got all the high position roles one strives for in our profession, travelled everywhere, lived and worked overseas and done everything I've wanted to except starting my own business.2020 came which I feel wiped me out (mentally and economically). I've just now I would say recovered a bit, set myself up with a year worth of living expenses and moved abroad.This should be the time where someone should go head on and try make some money on their own but instead I'm just tired. I feel like when I was younger all that effort came so easy, I was curious, I wanted things. Now I'm so lost, anxious and filled with doubt thanks to global events and AI.I went on a shroom trip last week that taught me that I am capable, but the after effects wore off and I'm back to being me. I feel like a 20 year old. How do I not waste Feb 20, 2026 2:10 PM
github.com
Show HN: A native macOS client for Hacker News, built with SwiftUIhttps://github.com/IronsideXXVI/Hacker-NewsHey HN! I built a native macOS desktop client for Hacker News and I'm open-sourcing it under the MIT license.GitHub: https://github.com/IronsideXXVI/Hacker-NewsDownload (signed & notarized DMG, macOS 14.0+): https://github.com/IronsideXXVI/Hacker-News/releasesScreenshots: https://github.com/IronsideXXVI/Hacker-News#screenshotsI spend a lot of time reading HN — I wanted something that felt like a proper Mac app: a sidebar for browsing stories, an integrated reader for articles, and comment threading — all in one window. Essentially, I wanted HN to feel like a first-class citizen on macOS, not a website I visit.What it does:- Split-view layout — stories in a sidebar on the left, articles and comments on the right, using the standard macOS NavigationSplitView pattern.- Built-in ad blocking — a precompiled WKContentRuleList blocks 14 major ad networks (DoubleClick, Google Syndication, Criteo, Taboola, Outbrain, Amazon ads, etc.) right in the WebKit layer. No extensions needed. Toggleable iFeb 20, 2026 2:02 PM
