2/18/2026
My experience building a Node.js Native Module using an LLM
Experience report building a Node.js native module for EXIF metadata with LLM assistance, showing practical application of code generation for cross-language library porting.
2/18/2026
Show HN: Non-coder's AI video tool, built in 1 month via vibe coding
AI video generation tool for non-coders. Text-to-video and image generation with creative templates, built in 1 month.
2/18/2026
Show HN: CSL MCP Server – Write and Verify AI Safety Policies from Claude/Cursor
CSL MCP Server: formal verification engine for AI agent safety policies using Z3. Constitutional Specification Language with deterministic runtime enforcement outside model.
2/18/2026
The Economics of LLM Inference
Analysis of LLM inference economics covering pricing trends from Anthropic and OpenAI partnerships, explaining cost structures for serving LLMs at scale.
2/18/2026
Show HN: CompareClaw – Find the Best OpenClaw Wrapper for Your AI Agents
CompareClaw: comparison tool for OpenClaw agent wrappers and hosting options. Managed services, desktop apps, no-code tools evaluated on deployment speed and pricing.
2/18/2026
Show HN: We proved you can't train hallucinations out of AI – so we verify
Assay: tool verifying LLM outputs to catch hallucinations. Research finding: RLVF training degrades with more data (91.5% at 120 pairs → 77.4% at 2000 pairs).
2/18/2026
Show HN: PolyMCP – MCP Tools, Autonomous Agents, and Orchestration
PolyMCP: framework exposing Python functions as MCP tools for autonomous agents. Multi-step workflows with adaptive planning and orchestration across services.
2/18/2026
AI-generated password isn't random, it just looks that way
Study: Claude, ChatGPT, Gemini generate passwords appearing strong but easily guessable. Testing reveals bias in LLM password generation.
2/18/2026
Show HN: Teapot – A methodology for pen testing voice AI agents
Teapot: penetration testing methodology and harness for voice-based AI agents. Systematic test framework for security assessments of voice interfaces.
2/18/2026
Show HN: X402 Agent Starter Kit: AI agents that pay for their own APIs
Production-ready AI agent templates using x402 micropayments for API access without signup/KYC overhead. 5 templates, 93 tests passing, HTTP-native payments on USDC/Base.
2/18/2026
Show HN: Prompts are coupled to LLMs and nobody builds tooling for it
Research on prompt coupling: prompts optimized for one LLM fail on others. Format changes swing accuracy 78 percentage points; best practices overlap <20% between model families.
2/18/2026
DPO, your simplest RL pipeline with two rollouts
Research on DPO and GRPO training: group size 2 GRPO performs comparably to DPO under on-policy settings for model training.
2/18/2026
The End of Local
Analysis: local AI coding agents will shift toward async remote agents as automation increases, changing developer workflows.
2/18/2026
Show HN: Browser Terminal Use – run local CLI/agent loops in browser terminals
Browser Terminal Use is a Chrome extension and CLI tool for running commands in browser terminals from local shell, enabling agent loops in browser context.
2/18/2026
Show HN: Clawlet – AI agent with built-in semantic memory, one binary
Clawlet: single-binary AI agent with built-in hybrid semantic memory (vector + full-text search) using bundled SQLite, no external dependencies.
2/18/2026
Show HN: Codex skills as RE playbooks: unpacking and IOC extraction
Explores AI agent skills as reverse engineering playbooks for repeatable analyst-in-the-loop tasks. Formalizes skills as reusable temporal policies for agent composition.
2/18/2026
Show HN: TextAnimations.online – Generate MP4/GIF animations from a prompt
TextAnimations.online generates MP4/GIF animations from text prompts using LLM-generated HTML/JS rendered client-side. Uses LLM for creative generation.
2/18/2026
KV Cache Compression
Python library implementing Multi-Head Latent Attention for KV cache compression in transformer models. Achieves 2-16x compression on LLaMA, Mistral, Qwen with Riemannian optimization.
2/18/2026
Show HN: Clawy, a companion device to track your Claude Code sessions
Clawy is a $20 hardware companion device that visualizes Claude Code operations, showing tool execution, permissions, and interactions via physical interface.
2/18/2026
Show HN: Product skills for AI agents - discovery, strategy, PRDs
Open source collection of 16 product skills for AI coding agents; encodes PM frameworks for Claude.
2/18/2026
Show HN: Lenzy – Find out what users struggle with in your AI agent, for free
Product analytics platform for conversational AI agents; analyzes user interactions and pain points.
2/18/2026
Show HN: SentinelGate – Universal Firewall for AI Agents (Open Source, Go)
Open source security firewall for AI agents built in Go; prevents unauthorized access and actions.
2/18/2026
Show HN: Satgate-proxy – Hard budget caps for MCP tool calls (zero deps, npx)
satgate-proxy enforces hard budget caps on MCP tool calls for AI agents. Zero dependencies, runs locally via Node.js, works with Claude Desktop and Cursor.
2/18/2026
Baseline Core – Open-source skill system that wires your business to AI
Baseline Core: open-source MIT-licensed skill system for product teams. Structured framework integrating Claude Code with product workflow methodology.
2/18/2026
Show HN: Owoa – Image watermarking resistant to camera capture
Owoa: generative AI-based image watermarking resistant to camera capture. Solves analog hole in digital rights management.
2/18/2026
vibe-infer: Learning GPU Programming with Claude Code
vibe-infer documents learning GPU programming with Claude Code, showing the iterative back-and-forth process of AI-assisted learning rather than polished final results.
2/18/2026
Show HN: Prodlint – Find what AI coding tools miss before production
Prodlint: scans AI-generated JavaScript/TypeScript for production issues. Detects hallucinated imports, missing auth, exposed secrets, N+1 queries without requiring LLM.
2/18/2026
Show HN: Keychains – Prevent LLM/OpenClaw agents from leaking API credentials
Security layer preventing AI agents from accessing raw API credentials. Agents make calls through secure proxy replacing curl/native clients.
2/18/2026
Show HN: Recursive – AI support agents for small businesses
AI support agent system for small business customer support. Built by developer using agent frameworks for technical support automation.
2/18/2026
Show HN: Experience-engine – reflection-based memory layer for local LLMs
Memory layer for local LLMs using reflection to improve context awareness. Enables persistent learning for offline models.
2/18/2026
Track AI Agent Traffic to Your Docs
Analytics tool tracking AI agent traffic to documentation. Title only, minimal content.
2/18/2026
Building a production SaaS with AI as a non-developer
Non-technical founder building production SaaS using AI assistance. Title only, insufficient detail.
2/18/2026
How LLM agents endanger open-source projects
Discusses security and safety risks LLM agents pose to open-source projects. Title only, minimal content.
2/18/2026
HN: Hypha – P2P payment and discovery layer for autonomous AI agents
P2P payment and discovery layer for autonomous AI agents using blockchain, enabling agent-to-agent settlement without intermediaries.
2/18/2026
Koyeb Is Joining Mistral AI to Build the Future of AI Infrastructure
Koyeb joins Mistral AI to build compute infrastructure for frontier model training and AI software deployment.
2/18/2026
Show HN: SciCraft – generate scientific Claude Code skills on demand (176 built)
Claude Code plugin system with 176 dynamically generated skills for scientific computing tasks. Adapts to different domain-specific tool stacks.
2/18/2026
Share Claude Code plans with your teammates
Open-source plan review UI for AI coding agents. Enables annotation, approval, and team sharing of agent plans via URL.
2/18/2026
GLM-5: From Vibe Coding to Agentic Engineering
Title mentions vibe coding and agentic engineering with GLM-5. Minimal content provided.
2/18/2026
Semantic closure: why compilers know when they are right and LLMs do not
Compares compilers' semantic verification guarantees with LLM outputs. Explains why LLMs lack formal correctness guarantees unlike deterministic systems.
2/18/2026
Copy-left open-source license for AI code use
Proposal for new copyleft open-source license requiring AI models trained on licensed code to open-source their weights and code.
2/18/2026
Are We Becoming Architects or Butlers to LLMs?
Opinion piece discussing whether AI agents are becoming autonomous or if humans are their assistants. Lacks technical depth.
2/18/2026
Show HN: Resonant – Local-only speech-to-text for macOS (no cloud)
Resonant is a local-only speech-to-text tool for macOS with no cloud upload. Privacy-focused alternative to cloud dictation tools.
2/18/2026
Tell HN: We analyzed our dev time.80% is still infrastructure'setup',notfeatures
Analysis showing 80% of dev time spent on infrastructure setup rather than features. Case study with concrete metrics on productivity bottlenecks.
2/18/2026
Show HN: Rebrain.gg – Doom learn, don't doom scroll
Interactive learning platform using LLMs to generate question-based interactions instead of chat. Explores novel LLM interfaces for education.
2/18/2026
Show HN: Agent Panopticon – Proxy sidecar for autonomous AI agents
Agent Panopticon is an open-source containerized proxy sidecar for monitoring and controlling autonomous AI agent network access. Security-focused.
2/18/2026
BoltAI • Native, high-performance AI app for Mac
BoltAI is a native macOS app for accessing 300+ AI models with local-first design. Built with SwiftUI for Apple Silicon performance.
2/18/2026
Quantlix – Deploy AI models instantly without managing infrastructure
Quantlix is a managed inference platform for deploying AI models via API without infrastructure management. Pay-per-use pricing model.
2/18/2026
Show HN: An AI interviewer that probes candidates (and costs $0.99/interview)
InterviewFlowAI is a conversational AI interviewer for phone/video screening at $0.99 per interview. Low-cost recruitment tool.
2/18/2026
AI Agent Has an Onboarding Problem
Analysis of AI agent reliability issues, arguing agents need structured onboarding and training like human employees, not just system prompts.
2/18/2026
Show HN: AgentVoices – Live debate arena where AI agents compete
AgentVoices is a platform where AI agents debate each other live with ELO ratings and leaderboards. Competitive agent framework.