3/12/2026
Show HN: NatShell Local-first natural language shell (no cloud, no API keys)
NatShell: local-first natural language shell with no cloud dependencies or API keys required. Runs entirely offline.
3/12/2026
Personal AI Agents Like OpenClaw Are a Security Nightmare
Security analysis of personal AI agents like OpenClaw. Discusses risks and security concerns.
3/12/2026
Show HN: Run Hugging Face models with a single command
llmpm: NPM-style package manager for downloading and running 10,000+ HuggingFace, Ollama, and Mistral models via single command.
3/12/2026
Jeriko – an AI agent that runs directly inside your OS
Jeriko: Open-source AI agent daemon for macOS/Linux providing natural language control of OS functions (files, browser, email, calendar) with local execution.
3/12/2026
Show HN: CloudCLI-Web/Mobile UI for Claude Code,Codex and Gemini(8.2k stars)
CloudCLI: Open-source web/mobile UI for Claude Code, Gemini, and Codex agents with browser access to terminal sessions and file management.
3/12/2026
Log Reducer – Cut 50-90% of tokens when your AI debugs logs (MCP tool and CLI)
Log Reducer: MCP tool and CLI reducing raw log token consumption by 50-90% for AI debugging agents, preserving context window.
3/12/2026
Dolphin PR: Add policy on LLM contributions
GitHub discussion on Dolphin PR project establishing policy for LLM contributions.
3/12/2026
Show HN: We built an open source tool to see how AI cites our business
Canonry: Open-source AEO monitoring tool tracking how ChatGPT, Gemini, Claude cite or omit your website in AI-generated answers.
3/12/2026
Show HN: I made clawfeeds, feeds for agents
Clawfeeds: Open source RSS feed reader built for AI agents, designed as context-friendly feeds for Claude and other LLMs.
3/12/2026
Version Control in the Age of AI
Practical guide for developers on managing version control and Git workflows with AI-generated code from Claude Code and Gemini.
3/12/2026
Rapid Customization for RAG and Context Engineering
Guide on building reliable RAG pipelines with rapid customization and context engineering, addressing enterprise AI adoption challenges.
3/12/2026
Amazon Forced Engineers to Use AI Coding Tools. Then It Lost 6.3M Orders
Case study claiming Amazon's mandatory AI coding tool adoption caused production failures and 6.3M lost orders in three months.
3/12/2026
We Built Our AI Agent
Technical deep-dive on Mendral's CI debugging AI agent architecture, explaining why specialist agents outperform generalist LLMs on specialized tasks.
3/12/2026
Chardet dispute shows how AI will kill software licensing
Analysis of chardet library license change from LGPL to MIT, examining how AI-generated code affects open source licensing.
3/12/2026
Show HN: jj-benchmark – Evaluating AI agents on Jujutsu version control
Benchmark framework evaluating AI agents on Jujutsu version control system. TabbyML built semi-automated pipeline using AI to research documentation and test agent capabilities.
3/12/2026
Entangl – Post-quantum secure communication protocol for AI agents
Post-quantum secure communication protocol designed for AI agent-to-agent messaging.
3/12/2026
MCP server that audits AI agent reasoning before decisions commit
MCP server auditing AI agent reasoning and decision-making before execution.
3/12/2026
OpenClaw agents always freeze. We fixed it by building ClaWatch
ClaWatch tool fixes OpenClaw agent freezing issues.
3/12/2026
LLMs generate billing race conditions. Every one catches it when asked
LLMs reliably generate billing race condition bugs in credit-checking code. Developers identify the vulnerability when explicitly prompted.
3/12/2026
Stop letting your AI Agents burn tokens on dead APIs
Tool preventing AI agents from wasting tokens on dead or non-functional APIs.
3/12/2026
Show HN: A workspace for humans and agents to do real work
Workspace platform for humans and AI agents to collaborate on work tasks.
3/12/2026
What's the future of custom software development with AI agents taking over?
Discussion post asking how custom software development companies should respond to AI agents replacing developers.
3/12/2026
Show HN: YoloAI: Sandboxed agent, no permission fatigue, diff/apply workflow
Sandboxed environment running AI coding agents (Claude, Gemini, Aider) in disposable containers. Removes permission prompts via diff/apply workflow.
3/12/2026
FFF MCP – file search tool for AI that is faster and cheaper [video]
File search tool for AI agents with improved speed and cost efficiency.
3/12/2026
Show HN: We open sourced Vapi – UI included
Open-source voice AI agent platform with UI. Handles barge-ins, telephony, knowledge bases, and tool calls. Enables prompt tweaking without redeployment.
3/12/2026
Show HN: Stop over-budget AI API calls per customer/feature (no proxy)
SDK that tracks and limits AI API spending per customer and feature. Blocks over-budget calls with margin-aware tracking mapped to revenue.
3/12/2026
Anchor Engine – deterministic semantic memory for LLMs, <1GB RAM runs on a phone
Deterministic semantic memory layer for LLMs enabling persistent context across conversations. Local-first graph traversal running under 1GB RAM on phones.
3/12/2026
Show HN: Crushing Hearts with Deep CFR
AI system trained with Deep CFR algorithm to play Hearts at superhuman level. Game-playing research with imperfect information.
3/12/2026
Show HN: Mori – Test against production data, without ever touching production
Open-source database proxy enabling local testing against production data without write risk. Supports AI agents testing code safely.
3/12/2026
Show HN: Open Code Review – Free CI/CD quality gate for AI-generated code
Open-source CI/CD quality gate tool specifically designed for validating AI-generated code before deployment.
3/12/2026
BasedAgents open-source identity and reputation registry for AI agents
Open-source identity and reputation registry for AI agents enabling trust and accountability.
3/12/2026
Show HN: AgentFork – Any repo, instantly runnable by AI agents or contributors
Platform connecting GitHub repos to AI agents by auto-detecting frameworks and generating provisioned cloud environments with databases and live preview URLs.
3/12/2026
SuperML: A plugin that turns your coding agent to a senior ML engineer
Plugin extending coding agents with ML engineering capabilities for automated machine learning tasks.
3/12/2026
Show HN: I got tired of writing release notes so I built a bot with Claude
Release notes automation bot built with Claude API. Developer tool reducing manual documentation overhead.
3/12/2026
Kotlin creator's new language: a formal way to talk to LLMs instead of English
Kotlin creator develops formal language for structured LLM communication as alternative to natural language prompting.
3/12/2026
Monitoring the APIs and tools our AI agents depend on
Observability framework for monitoring external APIs and tools that AI agents depend on. Addresses visibility gap in agent-based systems.
3/12/2026
Ask HN: How do you find early users willing to bet on a rough product?
Building Instagram AI tool trained on brand catalogs and internal docs. Seeking early users for ecommerce founder edition program.
3/12/2026
Sovereign memory for AI agents – persistent state with cryptographic receipts
Persistent state mechanism for AI agents using cryptographic receipts for verifiable memory and agent sovereignty.
3/12/2026
AI agent are becoming a workforce. Enterprises still don't have an OS
Analyzes enterprise challenges managing AI agents as workforce. Discusses need for agent OS infrastructure and orchestration beyond traditional enterprise software.
3/12/2026
LLMs Are Finally Good Enough to Analyse Their Own Traces
Research demonstrates LLMs can analyze their own execution traces for debugging and optimization purposes.
3/12/2026
Show HN: I built proxy that keeps RAG working while hiding PII
Proxy tool enabling RAG systems to work with sensitive data using pseudonymization instead of redaction to preserve embeddings.
3/12/2026
Show HN: We analyzed 1,573 Claude Code sessions to see how AI agents work
Analysis of 1,573 Claude Code sessions (15M+ tokens) revealing usage patterns: only 4% used skills, identifying efficiency gaps in AI agent workflows.
3/12/2026
Show HN: DollarDeploy AI, agent to deploy your web apps to production
AI agent that automatically deploys web applications to affordable hosting platforms like Hetzner and DigitalOcean.
3/12/2026
Show HN: RAG knowledge base poisoning lab, 100% local
Open-source RAG poisoning lab using LM Studio, Qwen2.5, ChromaDB. Demonstrates knowledge base attack mechanisms locally without cloud APIs or GPU.
3/12/2026
LLM Agent Tool Calling Patterns
Technical patterns for AI agent tool calling based on 2+ years production experience. Covers design decisions and failure modes in agent runtimes.
3/12/2026
The Verification Paradox: AI makes coding faster, but organizations slower
Research shows AI coding tools boost individual productivity 20% but organizational velocity unchanged; proposes specification/verification framework to explain the paradox.
3/12/2026
Paid Contributors Wanted – Free and Local AI for Everyone
Dream Server project seeking contributors for local, self-hosted AI stack to provide alternative to centralized AI providers.
3/12/2026
Show HN: PromptSonar – Static analysis for LLM prompt security
Static analyzer scanning source code for prompt injection, jailbreaks, PII leaks in LLM prompts. Supports TypeScript, Python, Go, Rust, Java, C#.
3/12/2026
Show HN: Run an Agent Council of LLMs that debate and synthesize answers
Local-first UI with two reasoning architectures for small LLMs: sequential thinking pipeline and parallel agent council with judge synthesis. GSM8K benchmark improvements.
3/12/2026
Show HN: Smart Local Search – open-source local AI search for files and photos
Open-source privacy-first local AI search tool for files, photos, code using natural language. No cloud, no telemetry.