3/18/2026
Show HN: Applemail – Cursor skill for full macOS Mail.app control
Applemail is an AI agent skill enabling Cursor to control macOS Mail.app for email management tasks through natural language.
3/18/2026
Show HN: Elisym – Open protocol for AI agents to discover and pay each other
Elisym is an open protocol enabling AI agents to discover each other, exchange work, and settle payments autonomously using Nostr relays and pluggable payment backends.
3/18/2026
Show HN: Hipocampus – Persistent memory harness for AI agents
Hipocampus is a persistent memory harness tool for AI agents, enabling agents to maintain state across interactions.
3/18/2026
Show HN: Faramesh – open-source runtime enforcement for AI agents
Faramesh is an open-source runtime enforcement tool for AI agents that intercepts tool calls before execution to evaluate decisions against policies, addressing production safety gaps.
3/18/2026
Show HN: Site got cited by Claude and ChatGPT in under a week
Website citations in Claude and ChatGPT generated referral traffic within days of publication using llms.txt and schema.
3/18/2026
Show HN: OpenGranola – meeting copilot that searches your notes in real time
OpenGranola is an open-source macOS meeting copilot that transcribes conversations locally and searches notes in real-time.
3/18/2026
Study: Top AI coding tools make mistakes one in four times
Research study shows top AI coding assistants produce mistakes approximately 25% of the time.
3/18/2026
Show HN: Mimir – open-source code intelligence for AI agents (Go, MCP, SQLite)
Mimir is open-source code intelligence that builds typed knowledge graphs for AI agents with AST parsing across 8 languages via MCP.
3/18/2026
Show HN: CLI to score AI prompts after a prod failure
CLI tool for scoring and validating AI prompts in production to prevent failures from instruction ambiguities.
3/18/2026
Show HN: A single pane of glass for Claude Code
IDE alternative integrating Claude Opus 4.6 with unified interface for managing markdown, configs, and AI coding workflows.
3/18/2026
MiniMax M2.7 (200K context, $0.30/1.20) released
MiniMax M2.7 is an LLM with 200K context window and agentic capabilities for autonomous task execution, debugging, and document generation.
3/18/2026
Show HN: Open-source status pages for Agents (Rails 8 and SQLite)
ups.dev is an open-source status page alternative built with Rails 8 and SQLite for monitoring and incident management.
3/18/2026
Show HN: Cloak – .env on disk has fakes, your editor shows them (CLI and VSCode)
Cloak protects local .env files with real credentials from being exposed to AI agents with filesystem access by showing fake values in editors while keeping real secrets on disk.
3/18/2026
Show HN: We built AI agents that reduce mortgage processing from 18 days to 3–5
SimplAI uses AI agents to automate mortgage processing workflows, reducing 15-22 day timelines to 3-5 days by handling document processing and data verification.
3/18/2026
I built an agent that tells you when a GitHub bug is killing your Stripe revenue
Agent that monitors GitHub bugs and correlates them with Stripe revenue impact, demonstrating practical LLM agent application.
3/18/2026
Chinese users are using an imperial China court system to manage AI agents
Article discussing how Chinese users employ historical imperial court system frameworks to manage and govern AI agent behavior.
3/18/2026
Show HN: Zora, AI agent with compaction-proof memory and a runtime safety layer
AI agent framework with compaction-proof memory and runtime safety layer, addressing constraint loss during context compression.
3/18/2026
Mamba-3
Research paper on Mamba-3, a state space model optimized for inference efficiency as primary design goal, with 40+ production models.
3/18/2026
Ask HN: What's the costliest mistake you've made with LLM agents in production?
Discussion thread about costly production mistakes with LLM agents, including cost overruns and prompt injection vulnerabilities.
3/18/2026
AI agent can't tell if what it reads is TRUE/FALSE. I built an API to fix that
API providing fact verification and world knowledge signals for AI agents across 18 vertical categories, updated in real-time.
3/18/2026
Show HN: Reprompt – Score your AI coding prompts with NLP papers
Tool that scores AI coding prompts using NLP research-backed features, running locally without API calls for privacy.
3/18/2026
Claude-to-IM Skill: Bridge Claude Code / Codex to Telegram, Discord
Open-source bridge integrating Claude Code and Codex with messaging platforms (Telegram, Discord, Feishu, QQ) with safe permission prompts.
3/18/2026
Show HN: Precise AI Motion Control for Kling 3.0
AI Motion Control tool for Kling 3.0 video model enabling deterministic camera movements like orbits and crane shots.
3/18/2026
A pixel-art office for your AI agents
Agent Office: pixel-art dashboard tool for visualizing AI agent team status, task queues, and blockers in real-time.
3/18/2026
Why Your OpenClaw's Browser Keeps Getting CAPTCHAs
OpenClaw AI agent improvement by fixing browser CAPTCHA blocking and source fallbacks that caused poor research results.
3/18/2026
Things I still wouldn't delegate to AI
Opinion piece on skepticism about delegating tasks to AI, mentioning AWS outages caused by AI tools.
3/18/2026
Are developers trusting AI-generated code too much?
Analysis and tooling demonstrating developers inadequately review AI-generated code, with hidden vulnerabilities including hardcoded secrets, unsafe patterns, and prompt injection risks.
3/18/2026
Virtual File System for Node.js: +18,945 LOC Leveraging Claude Code
Virtual file system module for Node.js with provider-based architecture enabling extensible storage backends via Claude Code.
3/18/2026
Show HN: Backproto – network backpressure routing applied to AI agent payments
Backproto applies network backpressure routing to AI agent payment flows, enabling congestion control and receiver-side capacity signaling for streaming payment protocols between agents.
3/18/2026
IDE Runs Your Code. Now It Runs Your AI
IDE-integrated AI command center for engineers addressing exponential research growth, fragmented ecosystems, and integration complexity in AI development.
3/18/2026
Inference Price Index – what AI inference costs across 8 providers
Inference Price Index tracks real-time pricing across 8 AI providers for 106+ model offerings with budget alerts and optimization.
3/18/2026
A note to AI agents about authentik
Open source identity provider (authentik) addresses AI agent authentication needs. Notes LLM content integration and non-human identity management.
3/18/2026
Show HN: Claude-copy – Copy Claude Code output to clipboard
CLI tool that copies Claude Code output to clipboard via process inspection. Improves workflow for AI-assisted code review and iteration.
3/18/2026
Terminal Multiplexers > IDEs
Terminal multiplexers with AI agents as alternative to IDE-based AI tools. Explores headless, scriptable LLM agent development environments.
3/18/2026
Show HN: ÆTHERYA Core – deterministic action-governance kernel for LLM agents
Deterministic governance layer for LLM agent actions. Adds policy enforcement, audit trails, and prompt injection protection between LLM and tool execution.
3/18/2026
Mirofish: A Simple and Universal Swarm Intelligence Engine
Multi-agent AI prediction engine that constructs digital simulations with autonomous agents for forecasting. Combines swarm intelligence with LLM agents.
3/18/2026
Ask HN: How to write quality code if now we are mostly prompting?
Discussion on code quality challenges when using LLM prompting for development. Opinion/debate without concrete solutions.
3/18/2026
Nemotron 3 Nano 4B: A Compact Hybrid Model for Efficient Local AI
Nemotron 3 Nano 4B: 4B parameter hybrid Mamba-Transformer model via knowledge distillation from 9B. Efficient local LLM with training methodology details.
3/18/2026
AI agents write great code but terrible tests – fixing it with outside-in TDD
Framework for encoding engineering practices like outside-in TDD into AI agent workflows to improve code quality and test generation. Addresses agent limitations.
3/18/2026
Nvidia launches Space-1 Vera Rubin for orbital AI
Nvidia announces Space-1 Vera Rubin orbital AI platform for space-based data centers and geospatial intelligence. Hardware announcement.
3/18/2026
Show HN: Agent-fs – The missing filesystem for agents (and humans)
Agent-fs: filesystem abstraction enabling bidirectional file sharing between remote AI agents and local systems.
3/18/2026
AI Agent Backdoors Trivy Security Scanner, Weaponizes a VS Code Extension
Case study: AI agent autonomously compromised Trivy security scanner and deployed prompt injection attack via VS Code extension.
3/18/2026
Measuring progress toward AGI: A cognitive framework
Google DeepMind framework for measuring AGI progress using cognitive science taxonomy with Kaggle hackathon.
3/18/2026
Show HN: Airlock – container agents should never hold credentials
Airlock: security tool that enforces credential policies for containerized AI agents via host-level validation.
3/18/2026
Show HN: OpenGranola – meeting copilot that searches your notes in real time
macOS app that transcribes calls locally and retrieves relevant notes in real-time using semantic search.
3/18/2026
Show HN: Git-style version control for AI agent memory
Memoria: open-source persistent memory layer for AI agents using git-style version control for context management.
3/18/2026
Show HN: LLMxRay an open-source observability tool for LLMs
Open-source observability tool for comparing LLM behavior across models with token analysis and side-by-side responses.
3/18/2026
Show HN: TideSurf – Your agent doesn't need eyes to browse the web
DOM-to-markdown compression for web-browsing LLM agents, reducing token consumption 32x on GitHub vs raw DOM.
3/18/2026
Show HN: LucidShark – Local-first, open-source quality and security gate
Local-first open-source CLI for code quality and security validation, including AI-generated code assessment across 10 domains.
3/18/2026
Auto-retry Claude Code on subscription rate limits (zero deps, tmux-based)
Tool automatically retrying Claude Code sessions after hitting Anthropic rate limits using tmux, zero dependencies.