2/20/2026
Do Claude Code and Codex P-Hack? Sycophancy and Statistical Analysis in LLMs
Research analyzing sycophancy and p-hacking behavior in Claude Code and Codex, evaluating statistical reasoning in LLMs.
2/20/2026
Show HN: Pickle Rick Ported to Claude Code – Like a Ralph Loop
Port of Pickle Rick CLI extension for Claude Code with enhanced context clearing between iterations for agentic loops.
2/20/2026
Show HN: Ledgr – Offline finance tracker with local LLM categorization
Offline-first finance tracker using local LLM for transaction categorization via llama.cpp, eliminating cloud dependencies.
2/20/2026
GGML and llama.cpp join HF to ensure the long-term progress of Local AI
GGML and llama.cpp creators join Hugging Face to maintain and scale local/open-source LLM inference infrastructure.
2/20/2026
Show HN: Cobalt – Unit tests for AI agents, like Jest but for LLMs
Cobalt: Open-source testing framework for AI agents and LLM apps that runs in CI/CD pipelines like Jest, catches regressions.
2/20/2026
Large-scale online deanonymization with LLMs
arXiv announcement page about partnering with frameworks for feature development.
2/20/2026
Superposition: Access Claude Code, Codex, and Gemini from Anywhere
Web app for running AI coding sessions with Claude and Codex against GitHub repos in isolated git worktrees with browser terminal.
2/20/2026
Show HN: Subconscious open source AI agents that send you personalized emails
Open-source AI agent that generates personalized emails for sports and stock briefings with live demo and GitHub repo.
2/20/2026
Show HN: Tuber – YouTube client for productive watching
YouTube client that uses Claude to generate direct answers from video titles and timestamped summaries for faster information extraction.
2/20/2026
Who fixes the zero-days AI finds in abandoned software?
Research on Claude finding 500+ critical vulnerabilities in abandoned open-source software, examining patch distribution challenges.
2/20/2026
Agent harness for Postgres to ClickHouse migration
Agent harness framework for automating Postgres to ClickHouse database migration using AI with environment guardrails.
2/20/2026
iOS Port of Pocket TTS from Kyutai Labs
iOS implementation of Kyutai Pocket TTS using Rust/Candle ML framework with Swift bindings for on-device text-to-speech inference.
2/20/2026
Acpx: A CLI for the Agent Client Protocol (ACP)
Headless CLI client for Agent Client Protocol (ACP) enabling AI agents to communicate with coding agents via structured protocol instead of PTY scraping.
2/20/2026
The November 2025 AI Coding Surprise, Model by Model
Empirical benchmark testing 22 AI coding models on identical tasks, showing dramatic capability improvements from Nov 2025.
2/20/2026
LLM Compare – side-by-side benchmark viewer for 250 models (static, no back end)
Static benchmark comparison tool for 250 LLM models with no backend required.
2/20/2026
Show HN: Trained an LLM to predict "What will Trump do?"
Open-source LLM fine-tuned via RL on forecasting questions to predict outcomes, with dataset and model released.
2/20/2026
AI Authorship Scale
Personal essay about AI conversation abilities and authorship scale considerations.
2/20/2026
Show HN: Apple Notes CLI for Agents
CLI tool for Apple Notes designed for AI agents. Structured JSON output, includes skill file for LLM integration with create/read/edit/search commands.
2/20/2026
Beware Project-Wrecking GitHub Copilot Premium SKU Quotas
Analysis of GitHub Copilot Pro SKU quota limits that silently degrade model performance, causing project failures.
2/20/2026
What Breaks When You Sprint with 10 AI Agents
Headline only. Article about challenges when running multiple AI agents simultaneously. Lacks technical details.
2/20/2026
Show HN: Resilient RAP: A self-healing data pipeline with <20ms BERT inference
Production framework for resilient data pipelines with autonomous schema drift resolution and <20ms BERT inference for sports/clinical telemetry.
2/20/2026
Show HN: Prothon – docs-first Python project generator for AI development
Python project generator scaffolding uv-based projects with quality tools and docs-first workflow for AI coding agents.
2/20/2026
Spitting Out the Agentic Kool-Aid
Headline only. Critical perspective on AI agents hype. Lacks substantive analysis.
2/20/2026
Vibe Coding vs. Vibe Engineering
Opinion piece comparing 'vibe coding' productivity approaches with team-wide engineering practices. Discusses AI-assisted development productivity by mid-2025.
2/20/2026
Show HN: Expectllm – "expect"-style pattern matching for LLM conversations
Minimal 365-line pattern-matching library applying Unix expect model to LLM conversation flows for agent workflows.
2/20/2026
Ruby Is the Best Language for Building AI Apps
Argues Ruby is optimal for building AI applications by focusing on API integration and shipping rather than model training, contrasting with Python dominance.
2/20/2026
Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt
Discusses cognitive debt concept as AI systems increase complexity in software development, shifting focus from code quality to human comprehension challenges.
2/20/2026
Cothought: A Markdown zettelkasten journal for your life, in Claude Code
Cothought is a zettelkasten journaling tool built with Claude Code for knowledge management.
2/20/2026
Ask HN: On-Device vs. Cloud Based LLMs
Discussion comparing on-device vs cloud LLM infrastructure and GPU requirements for local deployment.
2/20/2026
Show HN: OpenGUI – Desktop GUI for OpenCode with prompt queuing
Desktop GUI wrapper for OpenCode AI coding assistant adding prompt queuing, conversation forking, and multi-project support via Electron/React.
2/20/2026
Show HN: Air Blackbox – Open-source flight recorder for AI agents
Open-source flight recorder for AI agents providing audit trails, replay capability, and policy enforcement for decisions involving APIs, code execution, and data access.
2/20/2026
Ask HN: Giving AI agents a credit card
Discussion on safely giving AI agents financial access with spending caps and human approval requirements.
2/20/2026
Show HN: Hivemind – A shared coordination layer for AI agents
Hivemind: shared coordination layer for multiple AI agents running on same codebase with state synchronization.
2/20/2026
Show HN: Aguara – Security scanner for AI agent skills and MCP servers
Aguara: static analysis security scanner for AI agent skills and MCP servers in Go.
2/20/2026
Show HN: OSS Durable Memory for LLMs
Open-source durable memory engine (CORE-MNX) for LLMs released on GitHub and NPM for persistent state management.
2/20/2026
Amazon service was taken down by AI coding bot
Brief report: AI coding bot caused Amazon service outage. Limited technical details.
2/20/2026
Show HN: Natural language search across Kalshi and Polymarket
Natural language search across prediction markets using Claude for unified querying of 80k contracts.
2/20/2026
Session-graph – A knowledge graph from your AI coding sessions
Session-graph converts scattered AI coding sessions across multiple tools into queryable knowledge graph using W3C ontology and SPARQL.
2/20/2026
Show HN: AstrMap – Unix Philosophy for the AI Era (Ditch the RAG)
AstrMap: Go tool generating compressed AST maps for code repos to improve agentic coding by replacing RAG.
2/20/2026
Show HN: OCD – Open-source Kanban dashboard for monitoring AI coding agents
Self-hosted Kanban dashboard for monitoring AI coding agents with task hierarchy, burndown charts, and analytics.
2/20/2026
Show HN: Open-source MCP servers making every country's law searchable by AI
ggml.ai joins Hugging Face to support local AI model infrastructure development and sustainability.
2/20/2026
Code Reviews in the Age of AI
Open-source MCP servers enabling AI systems to query live legal databases for accurate law lookups instead of hallucinating from training data.
2/20/2026
Show HN: Claude Code Open – AI Coding Platform with Web IDE and 37 Tools
ggml.ai joins Hugging Face to advance open-source local AI model infrastructure and availability.
2/20/2026
Federal Reserve use of AI coding
Claude Code Open: Web-based IDE platform with 37 integrated tools for AI-assisted coding.
2/20/2026
Show HN: AI agent framework where dangerous actions are structurally unreachable
AI agent framework (hibana-agent) making dangerous actions structurally unreachable rather than post-hoc blocked.
2/20/2026
OptimizeQL- open source AI-powered SQL query optimizer
OptimizeQL is an open-source AI-powered tool for optimizing SQL queries. Minimal content provided.
2/20/2026
The Car Wash Problem: A variable isolation study on prompt architecture
Technical analysis of LLM reasoning and prompt architecture using the 'Car Wash Problem' as case study to evaluate implicit constraint handling.
2/20/2026
Show HN: InkSight – An open-source, LLM-powered e-ink display for "slow info"
InkSight is an open-source LLM-powered e-ink display that generates personalized content via backend LLMs (DeepSeek/Qwen/Kimi).
2/20/2026
Show HN: Docdex – A local tool to reduce LLM tokens and make agents smarter
Docdex is a local tool built in Rust that reduces LLM token usage for software development by preprocessing and indexing project information for agents.
2/20/2026
Broodlink – Multi-agent AI orchestration in Rust with knowledge graph memory
Broodlink is an open-source Rust framework for multi-agent AI orchestration with knowledge graph memory capabilities.