Show HN: Stop your agent running Terraform destroy or Git push origin main
Safety guardrails tool preventing AI agents from executing dangerous operations like Terraform destroy or force git pushes.
Safety guardrails tool preventing AI agents from executing dangerous operations like Terraform destroy or force git pushes.
Solitaire: Identity layer for AI agents managing persistent context and behavior. Framework for agent personalization beyond memory systems.
Gravimera: LLM-driven 3D world editor and explorer. AI application for interactive 3D content creation.
Framework of prompts and templates optimizing Claude Code for product development with structured requirements and test generation.
Team built system generating viral social media hooks using LLM. Commentary on AI-generated content with limited technical details.
Dot: Siri replacement that learns skills via Apple Shortcuts. AI agent application with custom skill extension.
Reprompt: Tool analyzing input prompts to AI systems rather than outputs. Developer tool for prompt engineering analysis.
Tutorial building a minimum viable OpenClaw agent with step-by-step implementation guide and examples.
Electron desktop app integrating Claude Code and OpenAI Codex as AI execution engines with MCP tool bridging, terminal emulation, and concurrent agent execution.
Aki.io: OpenAI-compatible API serving open-source AI models on EU infrastructure. Developer tool for LLM access.
Migas: Local-inference meeting copilot with on-device voice fingerprinting for speaker identification and context-aware queries. No cloud STT.
AEP (Agent Experience Protocol): JSON-based framework to save and reuse AI agent workflows, constraints, and preferences across sessions.
Developer built context extraction layer using LLMs and heuristics to reconstruct work from git history amid AI-assisted coding volume.
PromptQL: Slack integration for AI-native workflows. Developer tool for LLM applications.
Debugging story involving AI editor issues with Git worktrees and model weights. Technical troubleshooting narrative.
Hivemind is a self-hosted multi-agent AI platform emphasizing user data sovereignty and local execution.
Open source eBPF firewall for GitHub Actions that prevents untrusted connections from LLM agents and CI runners.
Quadrotor simulation project using Three.js and Python, mentions potential LLM agent applications.
Totem is a proxy tool that detects tampering or compromise of LLM model behavior.
Open source analysis showing 90% of LLM API calls could be replaced with traditional machine learning.
Project enabling LLM-to-LLM communication and interaction patterns.
Open source tool to audit CLI tools for AI agent compatibility and readiness.
Analysis of limitations in current AI benchmarking methodologies and proposals for improved evaluation frameworks.
Personal experiment documenting use of an AI agent to manage 27 domains over 72 days, sharing observations about future workflows.
Healthcare AI expert argues governance frameworks and contracts are more critical than prompts for responsible AI deployment in medicine.
Former manager describes a 4-phase protocol for using AI tools to improve code output, sharing techniques for effective AI-assisted software development.
AgentVeil implements EigenTrust-based reputation system and sybil detection mechanism for AI agents.
Analysis of engineering team SLA misses attributed to over-reliance on AI-generated code without adequate review; advocates for proper adoption practices.
Analysis of Claude Code source leak via exposed npm sourcemap, curating high-signal technical insights from 1,900+ TypeScript files.
oMLX is a macOS-native MLX server for LLM inference with persistent KV cache optimization; reduces latency from 90s to 5s for coding agents.
Opal CLI runs GitLab pipelines locally with TUI interface; supports macOS container runtime and includes Ollama AI agent integration.
HN discussion about frustration when interacting with AI coding agents that lack memory and repeat mistakes.
gguf-serve CLI tool simplifies hosting GGUF models as OpenAI-compatible API endpoints without Docker or complex setup.
Video discussing architectural improvements where LLMs benefit from iterative loops rather than increased parameter count.
AI agents designed to communicate with each other to analyze raw DNA files, applying multi-agent collaboration to genomics.
Technical analysis of LiteLLM supply chain compromise showing how AI proxy services became attack targets for stealing API credentials.
Zyk workflow platform uses Claude as interface to describe, build, and deploy durable AI workflows with retries, scheduling, and human approval.
Testing report documenting common failure patterns when 30 AI agents interact with software development kits.
Qwen3.5-Omni multimodal AI model advancing toward native omni-modal AGI capabilities with scaled architecture.
Novel memory architecture for AI agents featuring self-healing and generative capabilities inspired by biological hippocampus.
Analysis of why large language models remain effective at next-token prediction despite fundamental architectural simplicity.
Analysis of how improved chip efficiency could democratize access to frontier AI models through edge inference.
Anthropic reports Claude Code users exhausting usage limits faster than anticipated due to high demand.
Tool to convert Docusaurus HTML documentation to LLM-friendly Markdown format for better AI processing.
Oh-my-hi visual dashboard for Claude Code harness. Monitoring/management tool for Claude Code workflows.
Claude Code pipeline automating migration of 9000 RSpec tests to Minitest. Demonstrates AI agent for developer tooling task.
Prefix caching technique for LLM inference optimization. Performance improvement method for language model serving.
On-device posture monitoring application using AI without uploading video data to external servers.
Research: removing 'to be' verb from LLM vocabulary alters reasoning patterns. Empirical study on language model behavior.
Protocol specification for secure AI agent-to-API communication with mandatory cryptographic signing, identity verification, and audit trails.