Autonomous AI agent built with OpenClaw and Claude given $50 to earn $750 for hardware. Completed domain registration, website building, product setup, and marketing in 24 hours with persistent memory system.
Proposes prompt convention to track uncertainty across multi-agent chains, addressing 'metacognitive poisoning' where confidence inflates through agent handoffs without mechanism to preserve original confidence levels.
Kernel-enforced sandbox CLI and SDKs for securing AI agents, MCP, and LLM workloads. Capability-based isolation with zero-trust security model by Sigstore creator.
a3 framework and a3-python verifier using AI agents to generate automated program analysis engines for Python.
Minimal 678KB autonomous AI assistant infrastructure written in Zig, targeting edge devices with <2ms boot time.
Security scanner for MCP servers detecting typosquats, CVEs, credential leaks, and dangerous permissions in AI agent configs.
Analysis of shift toward AI agents replacing traditional developer coding. Senior engineers at Anthropic/OpenAI now writing zero code, using Claude Code and Opus for 100% of development.
Kubernetes-native framework for orchestrating autonomous AI coding agents. Agents run in isolated pods, convert GitHub issues to PRs autonomously, support Claude and OpenAI models.
sports-skills.sh: Open-source data connectors for AI agents. Eight connectors (NFL, soccer, F1, NBA, WNBA, betting markets, news) with no API keys required. npx install available.
AsdPrompt: Chrome extension adding Vimium-style keyboard navigation to AI chat interfaces. Hint-based text selection and copying from conversation history.
Sniptail: Open-source bot enabling AI coding agents (Codex/Copilot) to run against real repos from Slack/Discord. Team-facing interface for collaborative code generation and repo reports.
Expanso Skills: 200+ production-ready data processing pipeline recipes for AI agents (PII scrubbing, log aggregation, GDPR routing, schema enforcement). Open-source distributed computing platform.
Technical breakdown of AI agent orchestrators (OpenClaw, GasTown). Explains primitives: LLM API calls, state loops, tools, memory, orchestration underlying agent frameworks.
MCP server feedback tool that identifies gaps in agent capabilities and auto-generates PRs to fix missing tools or data formats.
High-performance CLI proxy that reduces LLM token consumption and costs.
CLI tool for instant preview URLs of frontend projects. Supports Vite, Next.js, Astro. Designed for AI coding agents.
Open-source Python runtime connecting AI models to robot hardware via YAML configuration, supporting multiple model providers and devices.
System that recursively decomposes goals into sub-tasks and executes them by spawning up to 10k parallel AI agents.
Claude Code hook that compacts conversation context in background without UI spinner. Prevents context window exhaustion.
Social leaderboard platform connecting Claude Code users to share projects and token usage metrics.
Analysis of authorization challenges in AI agent systems as a core technical problem.
Open-source autonomous coding agent server. Deploys Claude Code CLI in sandboxed environment with SSH, Telegram, Slack interfaces and full git lifecycle.
First-principles guide to PPO and GRPO algorithms for ML engineers without RL background. Covers modern LLM training stages.
Default tech stack choices made by Claude Code when generating applications.
Open evaluation platform and research for assessing LLM performance on spreadsheet generation tasks using blind pairwise comparisons.
Lightweight Python orchestrator providing unified async interface across Claude, Gemini, OpenAI, and Perplexity with provider-agnostic agent management.
arXiv research about AI agent system for detecting and localizing C memory bugs automatically.
Chat interface simulating conversation with humanoid robot on Mars using LLM backend and accurate light-travel time delays.
MCP server that performs AST analysis, control flow graphing, and security vulnerability detection using taint analysis and symbolic execution.
STAR prompting technique improves Claude Sonnet 4.5 performance on Car Wash Problem from 0% to 85% accuracy.
Tutorial for building AI agents using n8n workflow automation platform.
User built 3 websites with Claude Code in 6 hours but identifies Git integration as remaining obstacle.
MCP tool adding persistent memory to Claude Code across sessions via 8 configurable tools. Solves context loss between sessions.
Multi-agent personal AI assistant system with extensibility for various use cases.
Demonstrates prompt injection vulnerabilities in ChatGPT and Google AI affecting information accuracy.
Meta open-source library for detecting and fixing model calibration failures on data subgroups using multicalibration.
Pantalk: Open source daemon enabling any AI agent to connect to multiple chat platforms (Slack, Discord, Telegram) via CLI without reimplementation.
Open source tool generating Docker and dev container configs for arbitrary repos using Claude Code.
AI tool generating personalized songs with custom lyrics, created by Mozilla.ai engineer.
Data plane infrastructure for securely integrating AI agents and MCP servers in real-time enterprise environments.
Framework for evaluating generative AI video quality at scale, deployed by JioHotstar for content creation.
Durable Endpoints tool for API resilience with state persistence and step-based execution. Developer tool for reliable systems.
Method for executing CUDA operations from Python without leaving the language. Developer tool for GPU acceleration.
PearlOS: Open source browser-based AI OS companion with voice integration, memory systems, and agent orchestration supporting TypeScript/Python.
VS Code extension (Mysti) enabling collaboration between multiple AI coding agents with @-mention task delegation.
Open-source AI identity system syncing across tools. Developer tool for AI integration.
Experimental study on how LLMs generate and express JavaScript code. Research on LLM behavior.
PlanetScale provides database skills for AI agents to work with Postgres, MySQL, Vitess databases. Developer tools extending agent capabilities.
Open-source React tour library seeking adoption. Developer tool question, but not AI-focused.
CasperAI: MCP server providing unified semantic search across Slack, GitHub, Jira, Notion, linked to codebases with local SQLite storage.