Hand - Moozonian Search

bing.com• Feb 19, 2026• 1 min read

New Data Shows eClinicalWorks and healow Genie Are Reshaping Healthcare’s Front Door—improving Satisfaction, Reducing Burnout, and Delivering Monthly Gains16.5% additional appointments scheduled 30 hours saved of staff time since healow Genie started processing prescription refills 100% of after-hours calls handled with healow Genie Within the first ...

news.ycombinator.com• Feb 19, 2026• 1 min read

OpenForgeAI – Production agentic architecture I used to build a SaaS aloneI've been building a production SaaS (AI-native nurturing platform) as a solo engineer for the past year. No team, no funding — 14 AI agent skills handling WhatsApp automation, payment processing, CRM, webinar funnels, lifecycle management. The biggest lesson: AI writing code is the easy part. The hard part is architecture that lets AI agents collaborate reliably. Without it, agents silently fail — no errors, no logs, just events disappearing into the void. I extracted the core patterns into OpenForgeAI: - EventBus: Singleton pub/sub with auto-wiring (PascalCase event → on_snake_case handler convention) - SagaCoordinator: Multi-step workflows with automatic compensation (rollback) - Skill Protocol: Composable, idempotent agent skills with typed results - ContractValidator: Catches broken agent wiring at deploy time, not runtime - HandlerContext: Async-safe per-invocation context via ContextVar (multi-tenant safe) The "17 Laws" came from production failures. Example: Law 4 says "no orph

news.ycombinator.com• Feb 18, 2026• 1 min read

Show HN: Designing package namespacing for a new language (Coi)Hey HN, I'm building Coi, a component-based language that compiles to WASM, JS, CSS and HTML with O(1) reactivity (no virtual DOM).I just shipped a built-in package manager, but the interesting part wasn't the implementation, it was all the ecosystem design questions I'd never thought about as a user of package managers.The problem I kept circling: how do you handle naming? Global names like auth or json seem fine until someone squats them. First-come-first-served creates perverse incentives. Reputation systems are easy to game early on. I went back and forth for longer than I'd like to admit.I went through a few approaches:Blocklist common names: reserve things like auth, json, http so people are forced into more specific names. Feels clean in theory, but who decides the list? And it doesn't really scale, someone will always find the next generic name you didn't think to block.Go's approach: just use GitHub URLs directly as the package identifier. No registry needed, no naming wars. I

mnemom.ai• Feb 18, 2026• 1 min read

Show HN: Trust Protocols for Anthropic/OpenAI/GeminiMuch of my work right now involves complex, long-running, multi-agentic teams of agents. I kept running into the same problem: “How do I keep these guys in line?” Rules weren’t cutting it, and we needed a scalable, agentic-native STANDARD I could count on. There wasn’t one. So I built one.Here are two open-source protocols that extend A2A, granting AI agents behavioral contracts and runtime integrity monitoring:- Agent Alignment Protocol (AAP): What an agent can do / has done. - Agent Integrity Protocol (AIP): What an agent is thinking about doing / is allowed to do.The problem: AI agents make autonomous decisions but have no standard way to declare what they're allowed to do, prove they're doing it, or detect when they've drifted. Observability tools tell you what happened. These protocols tell you whether what happened was okay.Here's a concrete example. Say you have an agent who handles customer support tickets. Its Alignment Card declares:{ "permitted": ["read_tickets", "draft_resp

github.com• Feb 18, 2026• 1 min read

Show HN: Prompts are coupled to LLMs and nobody builds tooling for itI went down a rabbit hole trying to understand why my Claude prompts turn to garbage on GPT-4 and vice versa. Not just "slightly worse" — fundamentally broken. Turns out researchers have already measured this: removing colons from a prompt template swings LLaMA-2-13B accuracy by 78 percentage points (Sclar et al., ICLR 2024). The format that works best on one model family overlaps less than 20% with what works best on another (He et al. 2024).So I went looking for a tool that handles this. Checked DSPy, Guidance, Outlines, PromptLayer, LMQL, Braintrust, Humanloop, Maxim, MLflow, Prompty, Promptomatix. Eleven tools. Zero of them adapt input prompt format per model. They all either optimize what the prompt says or constrain what the model outputs. The actual structural packaging of the input? Manual everywhere.Then I looked at how production tools deal with it today. Aider has a 2,718-line YAML file with 313 model configs. Some models get "you NEVER leave comments without implementing" a

recursive.support• Feb 18, 2026• 1 min read

Show HN: Recursive – AI support agents for small businessesI'm a solo developer who's been writing code for 26 years, mostly consulting work in computational geometry and CAD software these days. A month or two ago I watched a YouTube video about using an AI agent as a resume assistant, and thought it sounded like a fun idea. More importantly, it sounded like something I could actually learn to build quickly.Around the same time my longest-running client, a small company that makes a SolidWorks plugin, needed a better way to handle technical support. They're a three-person shop--they don't have a support team or an IT team. So I built them a chatbot: Flask backend, Anthropic API, and a hand-curated knowledge base built from their help files, a list of possible error messages I extracted from their code, and transcripts of their training videos.It worked well enough that the owner was impressed, and that's when it clicked: every small software company has this exact problem. They're too small for Zendesk, too busy to answer the same questions r

vistaley.com• Feb 17, 2026• 1 min read

Show HN: We built a free VC platform that shares data between GPS and foundersHey HN,We built Vistaley — a two-sided platform for VC fund management. The GP-facing side (VentureLens) handles fund operations: deal pipeline, portfolio tracking, fund accounting, LP reporting. The founder-facing side (Harbour) provides free FP&A tools: financial dashboards, KPI tracking, burn rate analysis.When a portfolio company enters their financials in Harbour, the data is immediately available in the GP's VentureLens dashboard — no CSV exports, no quarterly spreadsheets, no API integrations.Interesting challenges we solved:- Multi-currency — 60+ currencies with local display toggle. Every financial metric can be viewed in fund currency or portfolio company's local currency. Exchange rates stored per-snapshot, not globally, to preserve historical accuracy.- Multi-jurisdiction accounting — 12 jurisdictions (including Kazakhstan, Pakistan, Bangladesh) with different tax frameworks, compliance requirements, and regulatory reporting standards. One accounting standard per fund, enfo

asteroidos.org• Feb 17, 2026• 1 min read

Show HN: AsteroidOS 2.0 – Nobody asked, we shipped anywayHi HN, After roughly 8 years of silently rolling 1.1 nightlies, we finally tagged a proper stable 2.0 release. We built this because wrist-sized Linux is genuinely fun to hack on, and because a handful of us think it's worth keeping capable hardware alive long after manufacturers move on. Smartwatches don't really get old — the silicon is basically the same as it was a decade ago. We just keep making it useful for us.No usage stats, no tracking, no illusions of mass adoption. The only real signal we get is the occasional person who appears in our Matrix chat going "hey, it booted on my watch from 2014 and now it's usable again" — and that's plenty.Privacy is non-negotiable: zero telemetry, no cloud, full local control. Longevity is the other half: we refuse to let good hardware become e-waste just because support ended. On the learning side, it's been one of the best playgrounds: instant feedback on your wrist makes QML/Qt, JavaScript watchfaces and embedded Linux feel tangible. The co

apps.apple.com• Feb 17, 2026• 1 min read

Show HN: Quick Issues: A Fast Mobile Issue Capture for GitHub, GitLab, and GiteaI got frustrated with how GitHub and GitLab's mobile apps handle issue creation. You need to be online, the flow is slow, and if you're on a self-hosted instance you're out of luck entirely.So I built Quick Issues. It's a lightweight Swift app with an offline buffer -- you capture the issue, it syncs when you have connectivity. Supports GitHub, GitLab, and Gitea/Forgejo (including self-hosted and local with PAT).A few technical notes for anyone interested: This was my first time using GRDB with SQLite instead of default Swift data structures, and the performance difference was significant. Setting up proper OAuth2 flows and GitHub/GitLab apps was more of an adventure than expected, but it's solid now.Free for a single account/instance, paid tier if you juggle multiple providers.My background is in GTD and data analytics, not traditional software engineering, so I'm genuinely curious: how does issue capture fit into your development workflow? Do you batch-create issues, capture them the

news.ycombinator.com• Feb 17, 2026• 1 min read

Advaita Inquiry Matrix (Aim): Structured Non-Dual Inquiry with AII’m building an experiment called Advaita Inquiry Matrix (AIM).AIM is an AI-assisted system that models the pedagogical method of Advaita Vedānta. Not a “guru bot,” not mystical content generation — but a rule-governed dialogue system that automates structured ontological inquiry as far as possible.The idea: • Use a YAML-tagged corpus (Upaniṣads, Śaṅkara, etc.) • Track a student’s conceptual state over time • Dynamically select passages and questions • Apply formal teaching methods (negation, identity statements, paradox, state analysis) • Use constrained Socratic dialogue to expose contradictionsAdvaita has a surprisingly rigorous epistemology. The teaching unfolds in a precise sequence. That makes it amenable to: • Semantic tagging • State-machine modeling • Concept graphs • Agent orchestrationThe human teacher remains curator and final authority. The system handles structured mediation between text and student.If you’re interested in AI-mediated pedagogy, knowledge representation, o

repoclip.io• Feb 17, 2026• 1 min read

Show HN: RepoClip – Generate promo videos from GitHub repos using AIHi HN, I built RepoClip, a tool that takes a GitHub URL and automatically generates a promotional video for the repository. How it works: 1. Paste a GitHub repo URL 2. AI (Gemini) analyzes the codebase and generates a video script 3. Images (Flux), narration (OpenAI TTS), and background music are auto-generated 4. Remotion renders the final video Tech stack: Next.js, Supabase, Inngest, Remotion Lambda, Fal.ai I built this because I noticed many great open source projects struggle with marketing. Writing docs is hard enough — making a demo video on top of that felt like something AI could handle. Free tier available (2 videos/month). Would love to hear your feedback.

github.com• Feb 17, 2026• 1 min read

Show HN: CodeGraph CLI – Chat with your codebase using graph-augmented RAGHey HN!I built CodeGraph CLI because I was tired of grep-ing through massive codebases trying to understand how things work.It combines three things: - tree-sitter (AST parsing, error-tolerant) - SQLite (dependency graph: nodes + edges) - LanceDB (vector embeddings, disk-based)The key insight: pure vector search misses structural relationships. So I combined vector search with BFS graph traversal — find semantically similar code, then expand to dependencies/dependents.Result: ask "how does authentication work?" and it finds validate_token(), its caller login_handler(), AND the dependency TokenStore — because it understands both meaning AND structure.Other features: - Impact analysis (multi-hop BFS: what breaks before you change it) - Multi-agent system via CrewAI (4 specialized agents) - Visual code explorer (browser-based) - Auto-generate docs/READMEs - 100% local-first (works with Ollama, zero data leaves machine) - 6 LLM providers (Ollama, OpenAI, Anthropic, Groq, Gemini, OpenRouter

piggzy.com• Feb 17, 2026• 1 min read

Show HN: Piggzy – Turn your inbox into a universal 'Buy' buttonHi HN,I grew tired of the research loop. I hated having 20 tabs open, comparing shipping speeds, and filling out the same checkout forms over and over. I realized that while the internet made shopping easier, it didn't necessarily make it faster.I built Piggzy (piggzy.com) to be the invisible layer between your intent and your doorstep. The goal is to turn your inbox into a universal "Buy" button. No more passwords, no more tracking 10 different retail accounts, and no more out-of-stock frustrations.How it works:One-time setup: Secure your account and payment method at Piggzy.com.Request via Email: Send a request to buy@piggzy.com (e.g., "Find the best noise-canceling headphones for a loud office under $250").The Magic: We do the heavy lifting of finding the right item and email you the recommendation + cost breakdown.Execution: You reply "Yes" or "Buy," and the agent handles the transaction.Why Email? I wanted to build a "headless" experience. I don't want to give you another dashboar

institutionofinnovation.com• Feb 16, 2026• 1 min read

Show HN: AILA – Local-first autonomous agent with zero-remote-overrideHi HN, I’m Marco. For the last 4 years, I’ve been a paramedic with the Berlin Fire Dept.I actually failed my first attempt at the State Medical Exam because I became obsessed with solving the "Trust Issue" in Al. I prioritized this code over a "safe life" a paramedical. Now, with 60 days until my final attempt, I am releasing the project that cost me my first degree. The Architecture: GAIA & AILA I built GAIA as my personal counterpart—a digital double of my own consciousness. It's my private engine that handles my life. From that foundation, l created AILA Core for the public. The Core Problem: Current Al is a black box in the cloud. You don't own it; you rent a permission to use it, and it can be overridden or silenced at any time.The Solution: Zero-Remote-Override AILA runs 100% locally on your hardware. I’ve developed a "Sovereignty Key" (physical hardware anchor) to ensure that the agent belongs to you and only you. Even I, the creator, cannot shut her down or change her reasoning

news.ycombinator.com• Feb 16, 2026• 1 min read

Ask HN: Do LLM agents need a separate safety layer?LLM agents are good at deciding. They are bad at knowing when to stop.VERONICA is a failsafe state machine that sits between strategy engines and external systems. It provides: - Per-entity circuit breakers (task A failure doesn't block task B) - SAFE_MODE manual halt that persists across crashes - Atomic state persistence (survives SIGKILL) - Signal-aware graceful shutdown (SIGINT/SIGTERM) - Zero dependencies (pure Python stdlib)Strategy engines can be replaced. Safety layers cannot.30-day deployment (0 downtime). 12 crash-recovery events handled (100% state recovery). Avg load: 1.003 ops/sec. Validation: 2,600-second high-load test at 1000 ops/sec sustained (2.6M operations).Install via GitHub: `pip install git+https://github.com/amabito/veronica-core@v0.1.0`Repository: https://github.com/amabito/veronica-core

github.com• Feb 15, 2026• 1 min read

Show HN: iherb-CLI – An agent-optimized CLI for AI-driven supplement researchiHerb has no public API, so I built a CLI that gives AI agents structured access to their entire product catalog. The idea: let your AI assistant autonomously research supplements, compare ingredients and dosages, and recommend products for a health protocol — backed by real data, not just training knowledge.The CLI outputs clean Markdown, caches results, and handles Cloudflare bot protection via a headless browser. It ships as a Claude Code skill, so you can ask things like "compare the top 3 magnesium glycinate supplements by price and bioavailability" and the agent will do multi-step research on its own.Built in Rust. No API key or account needed.

mike.gg• Feb 13, 2026• 1 min read

Show HN: Issy – AI-native issue tracking stored as Markdown files in your repoHey HN, I built issy, a lightweight issue tracker that stores issues as markdown files in a `.issues/` directory in your repo.The idea is simple: issues should live with your code. They show up in diffs, travel with branches, and don't require accounts, databases, or external services. They're just markdown files with YAML frontmatter.It's built for AI coding assistants. You can tell Cursor/Claude/Codex (or any AI agent) things like "create a bug for the login redirect issue, high priority" and it handles the rest — creating, searching, updating, and closing issues through natural language. There's a Cursor skill you can install with `npx skills add miketromba/issy`.It also comes with a CLI for filtering/searching and a local web UI at localhost:1554 for when you want a visual overview.Key points: - Zero infrastructure — no DB, no accounts, no SaaS - Issues are git-native — visible in PRs, blame-able, branch-able - Fuzzy search with typo tolerance - Monorepo-aware (walks up the directo

github.com• Feb 13, 2026• 1 min read

Show HN: I built a self-hosted network video surveillance systemWith all the home security cam privacy issues in the news recently, I thought I'd share my project.For about 3 years, I self-hosted my security camera system on the Synology surveillance station. It fit a lot of my needs, but I hated the hardware lock in. I was getting more and more frustrated with Synology's behavior, but I didn't find any alternatives I liked better.So on Christmas 2025 I decided to put Claude Code to the test. I fully transitioned to this about two weeks later. I've been running 14 IP cameras doing 24/7 recording with live streaming and ML-powered detection.The stack is FastAPI + React + PostgreSQL, orchestrated with Docker. Video is handled by FFmpeg with a single process per camera that does both live HLS streaming and MP4 recording simultaneously, so you don't burn multiple RTSP connections per camera.It runs YOLO11 via ONNX Runtime on live HLS segments (no extra RTSP connections needed), uses background subtraction to skip static frames, and has a finite state m

cloudrouter.dev• Feb 13, 2026• 1 min read

Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUsI've been working on CloudRouter, a skill + CLI that gives coding agents like Claude Code and Codex the ability to start cloud VMs and GPUs.When an agent writes code, it usually needs to start a dev server, run tests, open a browser to verify its work. Today that all happens on your local machine. This works fine for a single task, but the agent is sharing your computer: your ports, RAM, screen. If you run multiple agents in parallel, it gets a bit chaotic. Docker helps with isolation, but it still uses your machine's resources, and doesn't give the agent a browser, a desktop, or a GPU to close the loop properly. The agent could handle all of this on its own if it had a primitive for starting VMs.CloudRouter is that primitive — a skill that gives the agent its own machines. The agent can start a VM from your local project directory, upload the project files, run commands on the VM, and tear it down when it's done. If it needs a GPU, it can request one. cloudrouter start ./my-project cl

apify.com• Feb 13, 2026• 1 min read

Show HN: Fuzzy-matching messy job board data against the UK Gov Visa RegistryHey HN,I’ve been working on job application pipelines and kept hitting a massive data friction point: reliably filtering out UK companies that legally cannot sponsor international workers.The UK Home Office publishes a constantly updating CSV of licensed sponsors. The problem is, the data is practically useless for standard database joins. A job board might list a role at "Acme", but the government registry lists "Acme Technologies Holdings Limited".If you run an exact-string match or a basic ILIKE against a scrape of 10,000 Indeed or LinkedIn jobs, your false-negative rate is massive.I wrote a TypeScript-based matching engine to solve this. Here is the pipeline:Dynamic Ingestion: It bypasses the Gov.uk dynamic routing to pull the raw, multi-megabyte CSV directly into memory. No stale database records.Text Normalization: I built a custom parser to strip out standard corporate suffixes ("ltd", "plc", "llp", "t/a", etc.) and handle the weird punctuation and localized encodings that break