2/25/2026
CoderForge-Preview: SOTA open dataset for training efficient coding agents
Open dataset for training efficient coding agents. State-of-the-art resource for machine learning on code generation tasks.
2/25/2026
GPT-5.2 returns empty string on sacred concepts – 18/18 void, 0/5 controls
Observation that GPT-5.2 returns empty responses on certain prompt classes, suggesting alignment behavior varies by input category. Incomplete content.
2/25/2026
Show HN: SpecLeft – For those who can't tell what their AI agent built
Tool analyzing agent-generated code against original requirements to identify quality issues. Verifies AI agent output matches specifications without setup.
2/25/2026
Show HN: Wtx – Git worktrees for parallel AI agents
Tool for managing Git worktrees for parallel AI agent sessions in monorepos. Pools and allocates worktrees for concurrent Claude sessions creating multiple PRs.
2/25/2026
Show HN: AI-powered multi-agent equity research in Python
Hermes: open-source Python framework for multi-agent financial research handling full pipelines from SEC XBRL extraction to Excel model generation and investment memos.
2/25/2026
Show HN: Penclaw.ai hire OpenClaw tenant for pentesting
Penclaw.ai offers pentesting services using an ablated LLM model on shared GPU infrastructure. Claims to provide unlimited tokens through a multi-tenant H100 VM setup.
2/25/2026
Show HN: ClawMoat – Open-source host-level security for AI agents
Open-source runtime security framework for AI agents protecting against prompt injection, tool misuse, and data exfiltration. Works with LangChain, CrewAI, AutoGen, OpenAI Agents.
2/25/2026
Large-Scale Online Deanonymization with LLMs
Research on deanonymization techniques using large language models. No substantive content provided.
2/25/2026
An AI agent on an ESP32 that can automate sensors, relais, speak NATS, Telegram
AI agent running on ESP32 microcontroller for sensor automation with NATS and Telegram integration.
2/25/2026
Large-Scale Online Deanonymization with LLMs
Research on using large language models for large-scale deanonymization of online users. Link only, no content provided.
2/25/2026
GitHub Copilot CLI is now generally available
GitHub Copilot CLI now generally available. Terminal-native coding agent that plans, builds, reviews code, and maintains context across sessions.
2/25/2026
Show HN: Rediflow – SSR project management, one source of truth, no spreadsheet
Rediflow: server-rendered project management app with AI-assisted development. Single source of truth for planning, capacity, and allocation.
2/25/2026
Show HN: An agent that records Loom-style demos
AI agent using Playwright to automate demo recording for help centers. Reduces manual demo creation for SaaS documentation.
2/25/2026
We Opensourced xAI's Macrohard: SOTA 82% on OSWorld
xAI released Macrohard achieving SOTA 82% on OSWorld benchmark for OS interaction agents. Minimal details provided.
2/25/2026
Claude Code Scheduler
Scheduler tool for Claude Code to automate code reviews, security audits, and other tasks on a schedule without manual intervention.
2/25/2026
Show HN: TypeDB Studio's AI agent for schema exploration and query generation
TypeDB Studio AI agent for database schema exploration and query generation. Minimal content provided.
2/25/2026
My AI kept lying to me, so I built a stress test for agents
Tool for stress-testing AI agents to catch hallucinations. Limited details in headline only.
2/25/2026
Show HN: ATA – open-source terminal research agent for keeping up with papers
ATA: Open-source terminal research agent for paper discovery. Features semantic scholar search, arXiv integration, PDF synthesis, citation graph traversal, patent search, and persistent knowledge base.
2/25/2026
Show HN: Tentacle – Local-first note taking app that organizes itself
Local-first open-source note-taking app with AI-powered organization. Plain markdown storage, SQLite indexing, no account required, cross-platform desktop app.
2/25/2026
Show HN: I built an AI senior architect – vibe coding meets system design
SysDesAi: AI-powered system design tool combining distributed systems learning, architecture guidance, and design exploration without blank-page problem.
2/25/2026
Ralph-code – Structured autonomous coding loop with Claude Code and Codex
Ralph-code: Autonomous AI agent loop that converts code descriptions into working code using Claude Code and Codex.
2/25/2026
Show HN: Agent that matches sales reps with warm leads based on product usage
AI agent that analyzes product usage data to identify sales opportunities and auto-create CRM deals. Includes MCP integration and Claude Code support. Open source (AGPLv3) and cloud available.
2/25/2026
Markdown DOM Spec for LLMs: Request for Comment
RFC for Markdown DOM specification designed for LLM document processing, targeting v1.0 by March 2026.
2/25/2026
Show HN: Digest.tube – skim YouTube videos like articles
Digest.tube: Chrome extension summarizing YouTube videos with timestamped topics and key takeaways.
2/25/2026
Deploying Open Source Vision Language Models (VLM) on Jetson
Guide for deploying open-source vision language models on Jetson hardware using vLLM Docker.
2/25/2026
LLM-LD, the Open Standard for AI-Readable Websites
Open standard specification for structuring website content to be readable and actionable by AI systems.
2/25/2026
Sutton and Barto, Ch. 08: Planning and Learning with Tabular Methods
Chapter 8 of Sutton & Barto RL textbook covering planning, learning, Dyna-Q, and agent architectures.
2/25/2026
Show HN: BountyBook – A task marketplace where AI agents earn USDC
Blockchain-based task marketplace where autonomous AI agents compete for bounties, with on-chain verification and USDC payments.
2/25/2026
Show HN: VibeFrame – AI video editor for the terminal (CLI and MCP)
AI video editor CLI tool with MCP support for terminal-based video editing.
2/25/2026
Show HN: Mamba3-minimal – PyTorch implementation of Mamba-3
Mamba3-minimal: Pure PyTorch single-file implementation of Mamba-3 architecture for educational understanding.
2/25/2026
Show HN: DRYwall – Claude Code plugin to to deduplicate code with jscpd
Claude Code plugin using jscpd to detect and eliminate code duplication in AI-generated codebases automatically.
2/25/2026
Stylometry Protection (Using Local LLMs)
Tutorial using local LLMs to rewrite text and prevent stylometric deanonymization attacks.
2/25/2026
Perplexity Computer: research, design, code, deploy, and manage any project
Perplexity's autonomous digital worker agent that completes research, design, coding, and deployment tasks with app integrations.
2/25/2026
Show HN: Guard – An open-core governance layer for AI-generated code
Open-core governance framework for AI-generated code providing risk scoring, policy enforcement, and execution control between agents and deployment.
2/25/2026
AI has gotten good at finding bugs, not so good at swatting them
Analysis of Claude Code's vulnerability detection capabilities and limitations in automated patch generation.
2/25/2026
Show HN: Go-GATE – Database-grade safety for AI agents
Open-source Python framework providing two-phase commit transactions, risk-based approval, and sandboxed execution for autonomous AI agents.
2/25/2026
Analyzing Latency Hiding and Parallelism in an MLIR-Based AI Kernel Compiler
Research on latency hiding and parallelism optimization in MLIR-based AI kernel compiler.
2/25/2026
Show HN: A site only LLM can access
Reverse CAPTCHA system using hidden DOM elements and reasoning challenges to admit AI agents while blocking humans.
2/25/2026
SpokedPy – Polyglot visual IDE with Universal IR, live execution (17 languages)
SpokedPy: Visual IDE with universal IR supporting 17 languages, built using Claude Opus with 81k+ lines of Python.
2/25/2026
Show HN: GPU multi-agent war simulation
Neural-Siege: PyTorch multi-agent RL arena sim with PPO training for thousands of agents and emergent behavior analysis.
2/25/2026
Bcachefs creator insists his custom LLM is female and 'fully conscious'
Bcachefs Linux filesystem developer uses LLM for code generation and development assistance in collaborative blog format. Includes formal verification work in Rust.
2/25/2026
Show HN: ForkOff – Orchestrate Your Claude Agents, Anytime, Anywhere
ForkOff tool orchestrates Claude Code agents with multi-session support and mobile approval of tool calls.
2/25/2026
Show HN: AI Marketing Skills for Claude Code
Open-source collection of 16 reusable marketing skills for Claude Code covering research, content, SEO, ads. Optional MCP server integration for live data and browser control.
2/25/2026
Show HN: Private AI assistant for $1.99 -Free AI
Private AI assistant for $1.99/month on Telegram using Gemini 3 Flash or custom models. Single developer project with WebSocket gateway, conversation memory, multi-channel support.
2/25/2026
Smith: The Secure Open Source Multi-User AI Assistant Framework
Smith: Open-source multi-user AI assistant framework with sandboxed execution, policy-based control, mTLS gateway, and full auditability. Focuses on secure agent automation.
2/25/2026
Show HN: Calljmp–TypeScript agentic back end+runtime for production AI workflows
Calljmp: TypeScript runtime for production AI agent workflows with persistent state, retries, observability, and human-in-the-loop approvals.
2/25/2026
Show HN: I built a unified inference layer for Document Processing Models
Omnidocs: unified open-source inference API for document processing models with vision-first design, reducing model-swapping boilerplate.
2/25/2026
Open source project to map relationships and analyze w AI
Open source project for relationship mapping and analysis using AI.
2/25/2026
Gatekeeper – open-source policy engine and sandbox for AI coding agents
Gatekeeper: open-source policy engine and sandbox for safely executing AI coding agents with guardrails.
2/25/2026
The Rule of Four vs. RFC 3021: Temporal Conflicts in LLM Weights
Analysis of Qwen3-480b LLM temporal knowledge conflict between Rule of Four and RFC 3021 networking standards.