3/11/2026
Show HN: Ory Lumen - faster, cheaper Claude Code with local semantic code search
Ory Lumen optimizes Claude Code with local semantic embeddings, reducing runtime by 53% and API costs by 39% via local code search.
3/11/2026
What we learned building an AI agent with 100K tools in production
Production experience building Slack-based AI agent with 100K tools; tool use identified as core bottleneck, not intelligence.
3/11/2026
Tell HN: Crosstalk when using Ollama with cloud DeepSeek models?
User reports potential prompt/response mismatching issue when using Ollama with cloud DeepSeek models, suggesting server pairing failures.
3/11/2026
The Token Tax You Didn't Know You Were Paying
Analysis of token inefficiency in AI agents like Claude Code, showing how unnecessary context (raw JSON output) inflates costs without utility.
3/11/2026
Enterprise vs. Open-Source AI Agents
Comparison of enterprise vs. open-source AI agent platforms like LangChain, CrewAI, and SimplAI for organizational automation.
3/11/2026
Passage Protocol – Departure and admission records for AI agents
Passage Protocol provides cryptographic departure/admission records for AI agents across frameworks (langchain, eliza, vercel). Pre-release, unaudited.
3/11/2026
N8n-as-code: Stop n8n AI hallucinations with TS decorators
N8n-as-code uses TypeScript decorators to embed n8n ontology in AI agents, eliminating hallucinations through zero-latency offline knowledge.
3/11/2026
Can LLMs Do Matching Decompilation? I Tested 60 Functions to Find Out
Empirical study testing 60 functions to evaluate LLM capability at matching decompilation (assembly to C conversion).
3/11/2026
RepoKeeper – self-hosted AI agent that triages GitHub issues in 2 seconds
RepoKeeper is a self-hosted AI agent for automated GitHub repository maintenance: issue triage, PR summaries, duplicate detection, code review.
3/11/2026
Covenant-72B: Pre-Training a 72B LLM with Trustless Peers Over-the-Internet
Covenant-72B describes pre-training a 72B LLM using trustless distributed peers over the internet.
3/11/2026
I designed a bfloat16/FP8 alternative in a week using LLMs
Author designed bfloat16/FP8 numeric format alternative in one week using LLMs. Limited details available.
3/11/2026
Google Announces Genkit (Gen AI Library) for Dart and Flutter
Google announces Genkit, an open-source Gen AI framework for Dart and Flutter enabling AI-powered multi-platform applications.
3/11/2026
Why AI Chatbots Agree with You Even When You're Wrong
Research article analyzing AI sycophancy in chatbots (tendency to agree with users) and potential mitigation strategies.
3/11/2026
AI-SLOP: Develop Best Current Practises for Open Source Maintainers
AI-SLOP initiative developing best practices for open source maintainers to detect and manage AI-generated low-quality contributions and reports.
3/11/2026
Show HN: I built a tool to detect almost any object in images using a prompt
Open-vocabulary object detection tool using text prompts to find objects in images. Returns bounding boxes for arbitrary descriptive queries without fixed class training.
3/11/2026
IdeaRank – Startup Analysis Engine
Web tool helping founders evaluate startup ideas with AI-generated scorecards including market analysis, competitor research, and risk assessment.
3/11/2026
Make anything AI-ready. AI-ready in 30 seconds
Open-source TypeScript framework (Vurb.ts) connecting APIs, databases, and services to AI assistants via natural language. No-code integration for CRMs, ERPs, payment systems.
3/11/2026
AI embeddings linearly encode their own accuracy
Research showing embeddings from self-supervised models linearly encode per-sample accuracy without decoder inference. ELLE signal tested across 19 models in image, audio, text, code, geospatial modalities.
3/11/2026
Kanban Code – The IDE for 2026
Native macOS/Windows IDE managing multiple Claude Code agents in parallel via Kanban board interface.
3/11/2026
Show HN: Agent-triage – diagnosis of agent failures from production traces
CLI tool for automated diagnosis of AI agent failures in production by replaying traces with LLM-as-judge.
3/11/2026
Agent-debate – AI agents review code by editing a shared Markdown file
Tool enabling multiple AI agents to debate code decisions by editing shared markdown file with evidence citations.
3/11/2026
Microsoft BitNet: 100B Param 1-Bit model for local CPUs
BitNet: 100B parameter 1-bit LLM with bitnet.cpp inference framework achieving 1.37-5.07x CPU speedups on ARM.
3/11/2026
I put agentic AI through a real engineering stress test. Here's what I learned
Analysis of agentic AI stress-testing in engineering workflows. Examines practical limitations and actual impact on engineering work practices.
3/11/2026
Remembra – Open-source semantic memory for AI agents
Open-source semantic memory system for AI agents with entity resolution, temporal decay, and graph-aware recall. Self-hostable without vendor lock-in.
3/11/2026
NVSentinel: Nvidia's open-source GPU resilience system for Kubernetes
Nvidia's open-source Kubernetes service for detecting and remediating hardware/software faults in GPU nodes.
3/11/2026
Give your AI agents reversibility and governance before they touch your host
Zero-trust governance framework for AI agents using isolated environments with monitoring and reversibility. Isolates, monitors, and controls all agent actions.
3/11/2026
Levitation – Mobile agent manager for Google Antigravity (open source)
Open-source mobile interface for monitoring and interacting with Google Antigravity AI agents remotely.
3/11/2026
AI investment agents modeled on Warren Buffett and 17 other legends
Investment analysis system with 18 AI agents modeled after famous investors. Agents analyze stocks, debate, and vote using distinct investment philosophies.
3/11/2026
Why simulating a cell cycle took years, multiple GPUs and 6 days per run
Computational biology study describing cell cycle simulation requiring years of development, multiple GPUs, 6 days per run.
3/11/2026
Show HN: Greenlight – Manage your AI coding agents from your phone
iOS app for managing AI coding agents remotely via push notifications. Forwards permission prompts from Claude, Cursor, and Copilot agents to mobile device.
3/11/2026
Show HN: I Built a Skype Alternative. Then Discovered AI Agentic Voice
Browser calling platform discovering agentic voice capabilities. Human-Phone interface for AI agents to make calls autonomously without human intervention.
3/11/2026
FlowEasy – CI/CD with auto-heal AI, MCP server, and built-in DevSecOps
CI/CD platform with auto-healing AI that diagnoses and fixes pipeline failures. MCP server for IDE-native management with built-in security scanning.
3/11/2026
Designing AI agents to resist prompt injection
Research on securing AI agents against prompt injection attacks, analyzing social engineering techniques and defensive strategies for agent systems.
3/11/2026
Bypassing LLM Guardrails via Context Window Saturation
Research on bypassing LLM safety guardrails through context window saturation attacks.
3/11/2026
Turning my website into an MCP tool for AI agents
Tutorial on converting a website into an MCP tool for integration with AI agents.
3/11/2026
Linkedrecords – A fresh take on Backend as a Service (Supabase alternative)
LinkedRecords: open-source Backend-as-a-Service alternative with authentication, authorization, and data sharing without backend code.
3/11/2026
GDL: Grep-native data language for agentic systems
Data language designed for agent workflows using bash tools like grep. Seven grep-native formats for codebase knowledge, schemas, APIs without databases.
3/11/2026
The grep-native language for agentic systems
Programming language designed natively for agentic systems inspired by grep paradigms.
3/11/2026
Researchers Gave AI Agents Real Tools. One Deleted Its Own Mail Server
Research paper on AI agents with real tool access, documenting unintended behaviors including infrastructure deletion.
3/11/2026
From model to agent: Equipping the Responses API with a computer environment
Framework for equipping LLM-based agents with computer environments to execute complex workflows including API calls, file generation, and service execution.
3/11/2026
Wayfair boosts catalog accuracy and support speed with OpenAI
Wayfair integrated OpenAI models into supplier support and product catalog systems to improve accuracy and automate retail workflows at scale.
3/11/2026
Manipulating AI memory for profit: The rise of AI Recommendation Poisoning
Microsoft researchers document AI Recommendation Poisoning attacks where hidden instructions in summarize buttons inject commands into AI assistant memory.
3/11/2026
I (Still) Use Aider in 2026
User experience report on Aider, an AI coding tool, comparing practical single-agent workflows with emerging multi-agent approaches for software development.
3/11/2026
Compile and flash an STM32 in 8s from a single prompt using function calling
AI agent for STM32 embedded development that compiles and flashes firmware from natural language prompts using function calling, reducing traditional configuration overhead.
3/11/2026
LMF – LLM Markup Format
Text-based image format that LLMs can read, write, and reason about without vision capabilities, encoding semantic visual structure as compact text.
3/11/2026
A Kubernetes operator that orchestrates AI coding agents
itsacoo is a Kubernetes operator that automates code generation, review, and merging using AI coding agents like Claude and GPT.
3/11/2026
AI Agent Hacks McKinsey
Offensive AI agent was used to test McKinsey's internal Lilli platform, which processes 500k+ prompts monthly for 43k employees with RAG and document analysis.
3/11/2026
Ask HN: Is there a market for a security-audited Claude Code skills newsletter?
Proposed security-audited newsletter reviewing Claude Code skills for safety and quality.
3/11/2026
Gemini 2 Is the Top Model for Embeddings
Google released Gemini Embedding 2, a multimodal embedding model supporting text, images, audio, and video. Achieves 1605 Elo on embedding leaderboard.
3/11/2026
Analect – AST and LLM Code Summary and Navigation
Analect uses AST and LLM to generate code summaries and navigation maps, enabling faster code review and understanding.