5/11/2026
Testing an AI Agent Harness over a Few Weekends
Technical writeup testing AI agent harness using Cursor and Kimi K2.5 to build apps from specifications with minimal manual coding.
5/11/2026
Discovering Reinforcement Learning Interfaces with Large Language Models
Research on using LLMs to discover reinforcement learning interfaces, combining RL with language model capabilities.
5/11/2026
AI Agents Show Long Context Gains with LCM, Specialized Applications Emerge
Brief report on AI agents demonstrating improvements with long context models (LCM) and emerging specialized applications.
5/11/2026
Show HN: Detecting deepfakes without sending your files to a cloud API
Deepfake detection tool that runs locally without sending files to cloud APIs.
5/11/2026
Vibe Coding Still Needs a Senior Engineer (For Now)
Security proxy for AI coding agents at OS level. Addresses safety concerns in agentic IDEs.
5/11/2026
Atlas – Pure Rust Inference Engine
Pure Rust inference engine for running machine learning models locally.
5/11/2026
Behind the Claude Frenzy That Ate Up All the Mac Minis
Story about Mac Mini hardware demand for Claude inference. Lacks technical detail.
5/11/2026
BaseLedger: An open-source API quota firewall for AI agents
Open-source API quota firewall for AI agents using Scala 3, Pekko, and PostgreSQL. Prevents budget bankruptcy via race-condition-safe ledger mechanics.
5/11/2026
Agile in the Age of AI
Essay on how Agile principles apply in AI-enabled software development, focusing on communication loops and expanded stakeholder models.
5/11/2026
Patterns for Reducing Friction in AI-Assisted Development
Framework proposing five patterns for effective AI-assisted development, drawing parallels from pair programming practices like onboarding and shared standards.
5/11/2026
Show HN: Studis – Turn product photos into social media ads with AI
Studis tool uses Gemini Flash and Claude to generate social media ad copy and visuals from product photos. Built with Laravel, Vue, and Pusher.
5/11/2026
What AI harnesses should learn from OAuth
Conceptual framework applying OAuth security principles to AI agent harnesses, treating agents as principals requiring proper authorization models.
5/11/2026
Where V-JEPA 2.1's Dense Features Hold Up (and Where They Don't)
Pre-registered robustness study of Meta's V-JEPA 2.1 vision model across four sizes, with production robotics deployment findings and failure mode analysis.
5/11/2026
I built an open-source visual AI agent builder – PandaFlow
PandaFlow: open-source visual builder for constructing multi-agent AI systems with graphical interface.
5/11/2026
Show HN: LLM Steganography to hide text inside another plausible text
Discussion on personalized LLMs tailored to individual users being more useful than general-purpose models.
5/11/2026
OMLX: Local LLM Server for Apple Silicon Macs
Local LLM inference server for Apple Silicon Macs with continuous batching, KV cache management, and menu bar control.
5/11/2026
Endy – orchestrate multiple coding-agent CLIs to cut LLM spend
Local control plane orchestrating multiple coding-agent CLIs (Codex, OpenCode, etc.) via tmux with task logging and unified command interface.
5/11/2026
Open Agent Memory Protocol 1.2 ships; 1.3 (draft) adds governed memory
Open Agent Memory Protocol 1.2 release with 1.3 draft adding governed memory with scope, sensitivity controls, and audit trails.
5/11/2026
DoneSpec – deterministic completion checks for AI coding agents
Deterministic completion verification layer for AI coding agents ensuring tasks meet explicit criteria via local, repeatable checks.
5/11/2026
Show HN: AI Agents in 30 Lines of YAML: Lowdefy v5.3
Lowdefy v5.3 enables AI agents in YAML config that call existing API endpoints as tools with streaming chat.
5/11/2026
Cotypist – AI Autocomplete for Mac
Cotypist: AI autocomplete tool for macOS with word prediction across all applications.
5/11/2026
How enterprises are scaling AI
Enterprise adoption insights from Philips, BBVA, Jetbrains, Scania. Focus on building trust and organizational conditions for AI scaling.
5/11/2026
AI: Apologies, I was only doing as instructed. (What Hollow is and isn't)
Hollow-agentOS: multi-agent system with three local LLMs running autonomously on consumer GPU, self-deploying tools without human input.
5/11/2026
Using AGENTS.md / Claude.md to discourage AI-generated code contributions
Discussion on using AGENTS.md and Claude.md files to discourage AI-generated code contributions.
5/11/2026
Timebook – Time tracking and invoicing your AI agent can use
Timebook: time tracking and invoicing tool designed to be used by AI agents.
5/11/2026
Open-Source Models for Text Rendering and Image Editing
Open-source models for accurate text rendering and image editing in AI-generated images, addressing longstanding text generation quality issues.
5/11/2026
Meet the academics refusing to use generative AI
Profile of academics who refuse to use generative AI, citing concerns beyond job displacement.
5/11/2026
ProgramBench (Meta) Repro: variance across runs and findings
Meta/Stanford/Harvard released ProgramBench: benchmark for evaluating LLMs on code repository tasks using synthetic test generation and agent evaluation.
5/11/2026
GitHub Copilot is deprecating Grok Code Fast 1
GitHub Copilot deprecating Grok Code Fast 1 model on May 15th. Users must migrate to supported models.
5/11/2026
ASTro: AST-Based Reusable Optimization Framework
AST-based optimization framework for generating specialized code through partial evaluation.
5/11/2026
Silo: Isolated workspace manager for parallel agentic development
Workspace manager enabling parallel development with multiple AI agents in isolated Git worktrees.
5/11/2026
Show HN: AI agents who prevent context drift through gossip
Open-source multi-agent system using shared markdown wiki to maintain context and prevent agent drift.
5/11/2026
Semantic Phonons: Lattice Vibrations in AI Internals
Research exploring how language models represent semantic meaning through internal lattice vibrations analogy.
5/11/2026
Constrained Decoding: Forcing LLMs to Respect Your Taxonomy
Technique for constraining LLM outputs to valid categories in taxonomy-based classification tasks.
5/11/2026
Traceway: MIT-licensed observability stack you can self-host in ~90s
MIT-licensed OpenTelemetry observability platform combining logs, traces, metrics, and AI tracing.
5/11/2026
Nanonumpy: Build a tiny NumPy to understand vectorization
Educational Rust/Python package implementing NumPy-like operations to teach vectorization and SIMD concepts.
5/11/2026
Evolved antennas, LLM-generated code, and a potential antifuture
Discussion of evolved antennas created via evolutionary algorithms and parallels to LLM-generated code and potential risks.
5/11/2026
Make Claude Code Sing
Claude Code skill that generates a song from session transcripts using Gemini and Google Lyria, then plays it to speakers.
5/11/2026
Agent Patterns for AI Agent Development
Reference guide covering patterns and techniques for developers using AI coding assistants like Copilot and Claude Code.
5/11/2026
AI Agents Have Two Souls. You Only Control One
Definitional analysis of AI agents addressing vagueness in what constitutes an agent and resulting security implications.
5/11/2026
Predictions about AI and software development
Predictions on agentic coding future: multiplayer collaboration, better UX for AI artifacts, integrated human-AI tools for task tracking.
5/11/2026
State aware AI agent shitfest – with rules – that helps move code
THE_ROOM: Local-first Electron app for multi-model AI agent collaboration with code editing capabilities. Windows alpha.
5/11/2026
OpenAI launches DeployCo to help businesses build around intelligence
OpenAI launches OpenAI Deployment Company to help enterprises build and deploy AI systems, acquiring Tomoro for Forward Deployed Engineers.
5/11/2026
Reimplement GitHub for Agents
Self-hosted GitHub-compatible API server (agent-git-service) for agents and automation, supporting REST v3, GraphQL v4, and Git Smart HTTP.
5/11/2026
Does structured prompting change how LLMs reason, or just what they say?
Empirical study investigating structured prompting effects on LLM reasoning using Knowledge Innovation System protocol across ChatGPT, Claude, and Gemini.
5/11/2026
Show HN: ChatbotX, an open-source alternative to ManyChat
Open-source omnichannel chatbot platform for agentic workflows supporting APIs, CLI, and MCP protocols. Alternative to ManyChat and similar commercial platforms.
5/11/2026
When GPT 5.5 flags your chat for possible cybersecurity risk–ask it to help you
Anecdote about GPT flagging a fine-tuning session during a hackathon and using prompts to bypass restrictions. Lacks technical depth and verification.
5/11/2026
Beyond State-Wise Mirror Descent: Offline Policy Optimization with Parametric Policies
Theoretical research on offline reinforcement learning with general function approximation and parametric policies for large action spaces.
5/11/2026
Contextual Invertible World Models: A Neuro-Symbolic Agentic Framework for Colorectal Cancer Drug Response
Neuro-symbolic framework combining invertible world models with agents for drug response prediction in precision oncology.
5/11/2026
Decoding the Pulse of Reasoning VLMs in Multi-Image Understanding Tasks
Analysis of attention patterns in vision-language models during multi-image reasoning, revealing unfocused attention pulses and positional bias.