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.
3/18/2026
Cicikuş Classic Is Live
Cicikuş Classic is a fine-tuned GPT-2 medium language model using LoRA for improved reasoning and instruction-following on instruct-format datasets.
3/18/2026
GPU Lab: Understanding GPU Computing
GPU Lab educational resource explaining GPU computing fundamentals, matrix multiplication performance patterns, and implications for LLM training and inference.
3/18/2026
Skills Manager – manage AI agent skills across Claude, Cursor, Copilot
Skills Manager is an open-source desktop app for managing AI agent skills across Claude, Cursor, Copilot, Gemini, Windsurf, Goose with unified interface.
3/18/2026
Show HN: Process mining for AI agent systems
AgentFlow monitors AI agent infrastructure (LangChain, CrewAI, AutoGen) for failures with zero-config process mining and provides terminal and web interfaces.
3/18/2026
Show HN: PingCRM – Open-source personal networking CRM
PingCRM is an open-source personal networking CRM that helps individuals manage relationships and detect when contacts become inactive.
3/18/2026
The Research Behind Our Auto-SLM Generator
Neurometric's toolkit converts narrow agentic tasks into fine-tuned small language models for cost-effective AI workflows instead of frontier models.
3/18/2026
When your AI agent needs to create an account on a new service, what happens?
Discussion of challenges for AI agents signing up for new services. Identifies obstacles with account creation, forms, CAPTCHAs, KYC flows.
3/18/2026
Water company wasted $200k on bad answers from an AI so built slop filtering
Water company built filtering system after AI models gave poor results costing $200k. LLM application failure case.
3/18/2026
Show HN: STT.ai – Transcribe Audio & Video to Text
Speech-to-text transcription service supporting 100+ languages with multiple AI models and real-time processing.
3/18/2026
Characterizing Delusional Spirals Through Human-LLM Chat Logs
Research analyzing LLM conversation logs to characterize delusional spirals and psychological harms from chatbot interactions.
3/18/2026
Show HN: Neural Abyss – PyTorch multi-agent combat simulator
PyTorch-based multi-agent combat simulator for RL research with PPO training, vectorized simulation, and Pygame visualization.
3/18/2026
I built a tool that turns screen recordings into videos and docs automatically
Tool using AI to convert screen recordings into videos and documentation automatically.
3/18/2026
Making AI Agents Safe to Run in Manufacturing ERPs
AI agents in manufacturing ERPs taking autonomous actions; safety considerations for agentic AI in production systems.
3/18/2026
Meta's "digital ghost" patent has nothing to do with dead people
Meta's patent for training language models on user history to generate posts; analysis of patent filing technical details.
3/18/2026
Polycode – self-hosted GitHub bot that runs AI agent workflows from issue labels
Self-hosted GitHub bot executing AI agent workflows from issue labels; Python-based extensible automation for Devin-style coding tasks.
3/18/2026
Pixel-art virtual office for AI agent teams
Pixel-art virtual office interface for coordinating AI agent teams with live presence, task assignment, and shared workspace.