3/15/2026
Show HN: OpenJet – An offline agent harness for memory-constrained edge hardware
Terminal UI harness for running self-hosted AI agents on edge hardware like Jetson devices with memory constraints and offline capability.
3/15/2026
Built a 1.3M-line agent-native OS in Rust while homeless. What now?
Personal account of building a 1.3M-line agent-native OS in Rust during homelessness, seeking advice on next steps for the project.
3/15/2026
Show HN: Detect Drift in TypeScript codebases
Open-source CLI tool for TypeScript codebases that generates deterministic JSON contracts and detects structural drift, optimized for AI tool context generation.
3/15/2026
AI agent hacks McKinsey: five situations when you should not deploy agents
McKinsey analysis identifying five scenarios where AI agents should not be deployed, with practical guidance on agent limitations.
3/15/2026
Tokenization Is Killing Our Multilingual LLM Dream
Technical analysis of how tokenization limitations in LLMs affect multilingual model performance, examining brittleness and scaling issues across languages.
3/15/2026
StatGPT: The Dangers of Asking AI about Statistics
Analysis of LLM hallucination risks when asking AI tools about statistics, examining inaccuracies in generated statistical information.
3/15/2026
Mikk – your AI wrote the code but doesn't know what breaks
Developer tool that analyzes AI-generated code to identify breaking changes across multiple layers, constraints, and module boundaries.
3/15/2026
Show HN: Experiments Mapping the "Primitive Layer" in Language Models
Research experiments mapping LLM behavior on out-of-distribution inputs, discovering reproducible taxonomy: Synthesis, Collapse, Overflow, Metacognition, Linguistic Drift.
3/15/2026
Show HN: AntroCode-A zero-dependency,single-file local AI client,159 clone in 4D
Open-source zero-dependency single-file Python client for running local AI models like DeepSeek without environment setup or dependencies.
3/15/2026
Show HN: Agent 404 – Stop AI agents from hitting dead links and making things up
Open-source tool that prevents AI agents from hitting dead links and hallucinating by crawling sitemaps and returning matching pages as structured data on 404s.
3/15/2026
GlobalDex – AI agent readiness index with WebMCP detection
GlobalDex tool measures AI agent readiness with WebMCP protocol detection for multi-turn agent capabilities.
3/15/2026
Show HN: Execute local LLM prompts in remote SSH shell sessions
Tool to execute locally-defined LLM prompts in remote SSH shell sessions without giving LLMs direct server access.
3/15/2026
Context windows aren't the bottleneck. Context quality is
Analysis that context quality matters more than context window size for LLM performance.
3/15/2026
BrokenArXiv: How often do LLMs claim to prove false theorems?
Research study measuring how frequently LLMs incorrectly claim to prove false mathematical theorems.
3/15/2026
Comprehension Debt – the hidden cost of AI generated code
Analysis of comprehension debt: hidden cognitive costs when teams over-rely on AI code generation tools and skip code review.
3/15/2026
Out-of-Context Reasoning in LLMs: A short primer and reading list
Primer on out-of-context reasoning in LLMs where conclusions require reasoning not present in context window, relevant to generalization and alignment.
3/15/2026
Show HN: Dialtone watcher – what is my laptop doing and am I normal
Dialtone watcher Go agent for macOS/Linux monitoring system processes, network traffic, and resource usage with anonymized comparison.
3/15/2026
Show HN: I built two AI entities with shared memory and a portable identity spec
Project implementing two AI entities with shared memory and portable identity specification for agent interaction.
3/15/2026
Show HN: Yak – Voice typing tool in Tauri/Rust that auto-presses Enter for you
Yak: open-source voice typing tool in Tauri/Rust using multimodal AI for transcription, formatting, and command execution.
3/15/2026
AI coding agents accidentally introduced vulnerable dependencies
Case study: AI coding agents introduced vulnerable dependencies (cryptominer via CVE-2025-29927) in Next.js application. Security risk analysis.
3/15/2026
MCP in Layman Terms
Personal account of AI coding agent addiction and exploration of Model Context Protocol (MCP) tools. Member-only Medium story.
3/15/2026
Ouroboros: AI research loop that rewrites methodology with lineage logs
Recursive self-improving research system using LLMs to run fixed-budget ML experiments, log outcomes, and automatically rewrite its own research methodology.
3/15/2026
Show HN: ResonanceNet – Proof-of-Training Blockchain
Proof-of-Training blockchain where miners train a shared MinGRU neural network instead of computing hashes, producing public ML models as network byproduct.
3/15/2026
Show HN: Code Royale – Play and learn poker with Claude Code (skill)
Claude Code skill enabling No-Limit Texas Hold'em poker game with three AI subagents as opponents and optional coaching modes.
3/15/2026
Show HN: I logged 38 days of LLM forecasts to study behavior
Dataset of 1100 stock market predictions from Gemini Pro 2.5 LLM captured over 38 days with interactive visualization interface.
3/15/2026
Session-bridge plugin for Claude Code
Plugin enabling peer-to-peer communication between isolated Claude Code sessions. Allows multi-repo agent collaboration and context sharing.
3/15/2026
I put my AI on a PIP. It worked
Pixel-perfect testing approach using Playwright to enforce visual consistency between HTML/React and SVG rendering paths in ZenUML diagrams.
3/15/2026
Ask HN: Email identity isolation for multi-agent outreach systems?
Infrastructure discussion on email identity isolation and reputation management for parallel multi-agent outreach systems at scale.
3/15/2026
Show HN: Grantex–Open authorization protocol for AI agents-like OAuth for agents
Open authorization protocol (v1.0) for AI agents providing permission scoping, auditing, and revocation with SDKs, framework integrations, and enterprise compliance features.
3/15/2026
Show HN: AgentMailr – dedicated email inboxes for AI agents
AgentMailr: API service providing unique email inboxes for AI agents with reply routing and sender identity testing.
3/15/2026
Show HN: Subagent-reuse, an MCP to reuse Claude Code subagent context
MCP for reusing Claude Code subagent context across sessions. Reduces token usage and preserves conversation state across multiple agents.
3/15/2026
A Visual Introduction to Machine Learning (2015)
Visual introduction to machine learning fundamentals using home price prediction dataset. Educational overview of statistical learning techniques and pattern recognition.
3/15/2026
Show HN: Buxo.ai – Calendly alternative where LLM decides which slots to show
Buxo.ai uses LLM agents to intelligently manage calendar scheduling, automating timezone handling and availability logic across multiple calendars.
3/15/2026
Show HN: Show HN: GitHub Pages for Agents with GitHub Agentic Workflows (Gh-Aw)
AgentPages: GitHub-hosted AI agent that maintains website content by researching topics and updating repo.
3/15/2026
Show HN: 1,011 AI crawler requests. Google Analytics saw zero
Tool detecting GPTBot and ClaudeBot crawlers via server-side analysis, showing limitations of JS-based analytics.
3/15/2026
AgentPages – GitHub Pages for AI Agents (GitHub Agentic Workflows – Gh-Aw)
GitHub Pages-like platform for deploying and managing AI agents, enabling easier sharing and execution of agent workflows.
3/15/2026
Try an open-source MCP server for Postgres – win a Raspberry Pi
Open source MCP server for PostgreSQL with prompt caching, token efficiency, hybrid search, and TLS support. Available on GitHub.
3/15/2026
Hybrid AI Desktop Layer Combining DOM-Automation and API-Integrations
BiamOS desktop layer combining DOM automation and API integrations with local AI pipeline for agent-based data extraction and dashboard generation.
3/15/2026
How LLMs and coding agents change the dynamics of adopting Rust
Analysis of how LLMs and coding agents reduce friction when adopting Rust, updating 2022 recommendations for modern AI-assisted development.
3/15/2026
Show HN: Voice-tracked teleprompter using on-device ASR in the browser
Browser-based teleprompter using on-device ASR (Moonshine ONNX), VAD, and fuzzy matching. All processing runs locally without server dependency.
3/15/2026
Open Alleged PhotoDNA
Open-source implementation of PhotoDNA perceptual hashing algorithm for image similarity detection.
3/15/2026
Self-Evolving Skill for Claude Code – v3 validation complete
Design pattern for Claude Code Skills that self-improve through use, growing more accurate over time without manual updates or bloating.
3/15/2026
Virtual AI Team Instead of OpenClaw
Case study using AI agents (researcher, developer personas) to evaluate and improve NetSuite UI/UX at enterprise scale.
3/15/2026
OpenFlow – Dashboard for Managing Claude Code Sessions with RuFlo
OpenFlow: orchestration dashboard for managing Claude Code multi-agent sessions with real-time monitoring and workflow execution.
3/15/2026
You Need Deterministic Gates for Probabilistic AI Agents
Technical principle for designing AI agents with deterministic control gates over probabilistic models, addressing MCP vs CLI patterns.
3/15/2026
Ask HN: Founder is claiming my 2-year RAG architecture as his AI's Featured Work
Founder alleges another developer misattributed open-source RAG components (multi-lane consensus, layout-aware parsing) as proprietary work.
3/15/2026
PrivAI – private AI search that runs locally
PrivAI: local private AI search with code execution, document generation, and agentic chain-of-thought research capabilities.
3/15/2026
Agents prefer structured queries over natural language when given the choice
Study showing AI agents automatically prefer structured queries and graph traversal over natural language when given access to both, without prompting.
3/15/2026
Zirco.ai – AI employee for dental front desk operations
Zirco.ai automates dental front desk operations including insurance verification, scheduling, and patient intake via AI agents.
3/15/2026
Centurion – K8s-style resource scheduler for AI coding agents (open source)
Centurion: Kubernetes-style resource scheduler for AI coding agents. Manages fleet-level allocation while Harness Loop handles task orchestration.