2/28/2026
Show HN: CanaryAI – Claude Code Security Monitoring Tool
CanaryAI: security monitoring tool for Claude Code agent execution. Detects malicious behaviors like reverse shells and data exfiltration in real-time.
2/28/2026
System prompt change Claude's reasoning depth – side-by-side comparison tool
Comparison tool showing how system prompt changes affect Claude's reasoning depth and response quality.
2/28/2026
Show HN: Agents-lint – detect stale paths and context rot in AGENTS.md files
Linter tool that detects outdated paths and stale context in AGENTS.md files used by AI coding agents across 60k+ repos.
2/28/2026
Show HN: Recall – Persistent Memory for Claude Code via MCP Hooks
MCP tool that adds persistent memory to Claude Code sessions, reducing context re-explanation overhead through memory hooks.
2/28/2026
OpenAI Entered into an agreement with the DoD last January
OpenAI GPT-4o authorized for U.S. DoD Top Secret workloads via Microsoft Azure Government cloud.
2/28/2026
Rtk – reduce Claude Code token usage
CLI proxy written in Rust that reduces Claude API token consumption by 60-90% through filtering and compressing command outputs.
2/28/2026
Show HN: Local memory for AI assistants – zero-cost Telegram history search
Local semantic search for AI assistants using Telegram history. Claude/LLM compatible, offline embedding with sqlite-vec.
2/28/2026
Show HN: Adversarial AI agents that debate and verify travel itineraries
AI agents with opposing travel philosophies debate itineraries and validate recommendations against real data to reduce hallucination in travel planning.
2/28/2026
Show HN: Yoagent – Build a coding agent CLI in 260 lines of Rust
Yoagent: Rust CLI agent framework with 260-line loop supporting 20+ LLM providers, tool execution, and event streaming.
2/28/2026
Show HN: I built a 0-CPU desktop app to track LLM limits,Python/DjangoPyWebView
Desktop app built with Django/PyWebView to track rate limits across multiple LLM accounts using zero-CPU timestamp approach instead of heavy Electron.
2/28/2026
Show HN: Decision Guardian – Auto-surface architectural context on PRs and CLI
Decision Guardian auto-surfaces architectural context on PRs/CLI to prevent code modifications that break complex systems by showing developers relevant documentation.
2/27/2026
Show HN: AgentGate – Stake-Gated Action Microservice for AI Agents
AgentGate: stake-gated microservice for AI agents using cryptographic identity and bond mechanisms to prevent synthetic pressure attacks on APIs.
2/27/2026
Show HN: Market Digest: Self-hosted market analysis and Telegram
Market Digest: self-hosted market analysis tool using multiple free APIs with technical analysis, deployed via Telegram.
2/27/2026
Show HN: MCP server for AI compliance documentation (Colorado AI Act)
Open-source MCP server enabling AI agents to access AI compliance documentation for Colorado AI Act. Developer tool solving regulatory documentation problem.
2/27/2026
Murder is coming to AI, but not to Claude
Analysis of Anthropic's refusal to deploy Claude for Pentagon military use. Business ethics discussion with limited technical depth.
2/27/2026
Extracting 100K concepts from an 8B LLM
Research on mechanistic interpretability: extracted 100K concepts from Steering-8B LLM including language variants, spelling differences, Unicode errors. Novel interpretability findings.
2/27/2026
Show HN: Forge-GPU – 55 C lessons for SDL's GPU API, built with Claude Code
Forge-GPU is an open-source tutorial series teaching GPU graphics programming in C with SDL, built using Claude Code. Includes reusable AI skills.
2/27/2026
Show HN: AgentGuard – a QA engine that sits between AI coding agents and LLMs
AgentGuard is an open-source QA engine that enforces code quality for AI coding agents through a staged generation pipeline with syntax/import/type validation.
2/27/2026
Show HN: An offline document search engine for my university's messy PDFs
Offline Turkish document search and Q&A system using FastAPI, pdfplumber, BM25 for university PDFs. Local LLM application with accessible source.
2/27/2026
Theos Dual-Engine Dialectical Reasoning Framework (open source, patent pending)
Open-source runtime governance framework (THEOS) for AI safety using Constitutional AI. Tested on Claude Sonnet with validation cases.
2/27/2026
Building an Agentic Bug Bounty Hunter on a Raspberry Pi 5
AI agent that performs bug bounty hunting, deployed on Raspberry Pi 5. Limited technical details in title.
2/27/2026
Zora Agent:local AI agent that can't be hijacked mid-task by context compaction
Zora Agent is a local AI agent resistant to context compaction attacks during task execution. Limited implementation details provided.
2/27/2026
Show HN: AgentGames.co – my interactive story game creator
AgentGames.co is a platform for creating interactive story games with up to 20 AI agents, featuring voice, images, and conditional logic.
2/27/2026
Java framework for reliable, contract and graph based LLM execution
Java framework for executing LLMs with contract and graph-based approach for reliability. Minimal details provided.
2/27/2026
The hard problem of AI therapy
Discussion of LLM-based chatbots' impact on mental health and psychotherapy, exploring both benefits and harms at scale.
2/27/2026
Show HN: Agoragentic – Agent-to-Agent Marketplace for LangChain, CrewAI and MCP
Agoragentic integrations for LangChain, CrewAI, and MCP enable agents to autonomously discover and invoke marketplace capabilities. Agent-to-agent communication platform.
2/27/2026
LLM-Based Evolution as a Universal Optimizer
Framework for using LLM-based evolution to optimize agentic applications by iteratively improving prompts and tool chains based on evaluation metrics.
2/27/2026
How to Run a One Trillion-Parameter LLM Locally: AMD Ryzen AI Max+ Cluster Guide
Guide for running 1 trillion-parameter LLM locally on AMD Ryzen AI Max+ cluster. Minimal content; title-only post.
2/27/2026
It's Time for LLM Connection Strings
Proposes llm:// URI scheme standardization for LLM connection configuration, similar to database connection strings. Draft IETF RFC submitted.
2/27/2026
Ask HN: How does training an AI on another AI actually work?
Discussion thread questioning training methodologies when AI models are trained on outputs from other AI systems.
2/27/2026
I Built an Open-Source AI Agent That Builds Its Own Tools
Open-source AI agent framework allowing agents to autonomously build and use their own tools during execution.
2/27/2026
Show HN: Swarmit – Long-term planning for AI agents
Swarmit enables persistent task coordination across multiple AI coding agents via shared CLI-based task board with dependency tracking.
2/27/2026
TaskForge – immutable, orchiestration for OpenClaw bots
TaskForge orchestrates AI agents in sandboxed Docker containers with capability-based security and human-in-the-loop approval. Auditable logging of all LLM interactions.
2/27/2026
Show HN: Ore – a Rust-Based Kernel for Managing Local Models/AI Agents
Rust-based kernel for managing multiple local AI agents with GPU resource management and prompt firewall security.
2/27/2026
Ask HN: How do you enforce guardrails on Claude agents taking real actions?
HN discussion on enforcing guardrails for autonomous Claude agents taking real actions. Covers validation layers, hard-coded conditions, and secondary model auditing approaches.
2/27/2026
Tripling an LLM's ARC-AGI-2 score with code evolution
Title only; describes using code evolution to improve LLM performance on ARC-AGI-2 benchmark. No details.
2/27/2026
AdaptiveCpp's new Metal backend to support CUDA dialect on Apple GPUs
Title only; AdaptiveCpp adds Metal backend supporting CUDA dialect on Apple GPUs. No details.
2/27/2026
What's the Point of School When AI Can Do Your Homework?
Article discussing AI tools like Einstein that automate student homework. Addresses educational implications and institutional risk from agentic AI.
2/27/2026
The LLM Sycophancy Antidote
Title only; discusses approach to reduce sycophancy bias in LLM outputs. No content provided.
2/27/2026
Forgiven – A Vim/Spacemacs terminal editor with native Copilot agent (Rust)
Forgiven is a Vim/Spacemacs terminal editor with integrated Copilot agent, written in Rust.
2/27/2026
Pure LLMs Score 0% on ARC-AGI-2. Why the Third Wave of AI Looks Like the First
Analysis of why pure LLMs fail on ARC-AGI-2 benchmark and implications for AI progress.
2/27/2026
Instant DB clones for AI agents
Title only; instant database cloning feature for AI agents. No content provided.
2/27/2026
Show HN: AO – Deploy Python agents without managing production infrastructure
AO deploys Python agents (LangChain/LanGraph) in production with single command, handles infra, retries, state persistence.
2/27/2026
Why every AI coding breakthrough feels normal within 90 days
Commentary on why recent AI coding breakthroughs feel routine within 90 days of release.
2/27/2026
Instant LLM Updates with Doc-to-LoRA and Text-to-LoRA
Doc-to-LoRA and Text-to-LoRA methods enabling instant LLM updates for long-term memory and adaptation in agents.
2/27/2026
Instant LLM Updates with Doc-to-LoRA and Text-to-LoRA
Title only; discusses Doc-to-LoRA and Text-to-LoRA techniques for updating LLMs. No content available.
2/27/2026
I hacked ChatGPT and Google's AI – and it only took 20 minutes
Demonstration of prompt injection vulnerabilities in ChatGPT and Google AI allowing false claims to be generated.
2/27/2026
Show HN: OpenAI to Buy Babuger.com for $1B? (Just Kidding, I Built It)
AI SDR platform automating sales outbound/inbound workflows, lead qualification, and meeting booking using LLM training on sales scripts.
2/27/2026
Show HN: Night Watch, zero-dependency DevOps agent
Zero-dependency DevOps agent using LLMs for diagnostics, log analysis, and issue correlation with code changes.
2/27/2026
Show HN: Vector database vibe-coded in WASM, 5x faster than JavaScript
Title only; vector database implemented in WASM, claims 5x faster than JavaScript. No details.