6/7/2026
Gaia2: Benchmarking LLM Agents on Dynamic and Asynchronous Environments
Gaia2 benchmark framework for evaluating LLM agents in dynamic, asynchronous environments. arXiv research project.
6/7/2026
When Can Amazon Block an Agentic AI Service?–Amazon vs. Perplexity
Legal analysis of Amazon v. Perplexity case on CFAA applicability to agentic AI services. Preliminary injunction discussion.
6/7/2026
Show HN: Persist – an AI agent that follows up accrued channels till they reply
AI sales agent tool that automates personalized follow-ups to prospects via email/SMS with prospect discovery features.
6/6/2026
You can't detect your way out of catastrophic LLM failure)
Study on stress-testing LLM safety claims. Method for detecting catastrophic failures and limitations of current AI safety approaches.
6/6/2026
AI Can now control your desktop
Desktop control via AI through Model Context Protocol (MCP). Integrates with Claude, Cursor, Windsurf, Zed. Local execution, no telemetry.
6/6/2026
GitHub Copilot: GPT-5.2 and GPT-5.2-Codex deprecated
GitHub Copilot deprecates GPT-5.2 and GPT-5.2-Codex models across chat, edits, and agent modes.
6/6/2026
Show HN: Aquifer – an MCP runtime for spiky agent tool traffic
Aquifer is an MCP runtime that coordinates HTTP traffic from distributed AI agents using durable queueing and rate limiting.
6/6/2026
Show HN: A beautiful and local-first PDF reader for studying dense things
Local-first PDF reader with note-taking and keyboard navigation for studying technical books, integrates Claude/Codex.
6/6/2026
Experts and superforecasters update their AI timelines
LEAP study surveys AI experts and superforecasters on AGI timelines and AI impact predictions through 2040.
6/6/2026
Decoupled RISC-LLM Architectures via Circadian Synaptic Consolidation
Research proposal for decoupled RISC-LLM architectures using circadian synaptic consolidation to address catastrophic forgetting and reduce latency.
6/6/2026
Show HN: Dap-mux – Connect your editor and REPL to the same debug session
Developer tool connecting code editor and REPL for unified debugging sessions using UNIX philosophy.
6/6/2026
Show HN: Kodiqa Agent
Kodiqa Agent: AI coding agent supporting multiple LLMs locally via Ollama or cloud APIs, with 78 slash commands, RAG, sub-agents, and plugin support.
6/6/2026
Learn from Your Mistakes: Tree-Like Self-Play for Secure Code LLMs
arXiv research on tree-like self-play method for improving secure code generation in LLMs through iterative learning.
6/6/2026
Made my first plugin – How I orchestrated 3 LLMs to ship a plugin in 2 hours
Developer orchestrated 3 LLMs to build Chrome/Firefox plugin ranking restaurant dishes in 2 hours, documenting workflow.
6/6/2026
Scientific Laws and LLMs Are the Same Shape
Analysis comparing scientific formulas and LLM compression mechanisms as analogous structural approaches to describing phenomena.
6/6/2026
Mass Layoffs Caused by AI
Analysis of AI-driven layoffs in 2024-2025, questioning whether efficiency gains were truly automation or cost-cutting pretexts.
6/6/2026
Looking for Backdoors in Jane Street LLMs
Title mentions security research on Jane Street LLMs for backdoors. No content provided to evaluate.
6/6/2026
Hermes – Community Web UI with Agent
Open source self-improving AI agent with layered memory, cron scheduling, and reusable skills system by Nous Research. Community web UI.
6/6/2026
AI Memory Proves Inefficient: Tenure Project Detects 95% Error Rate
Research study from UT Austin showing LLM memory systems have 95% error rate in long-term fact retrieval using vector-based storage.
6/6/2026
Meta confirms 1000s of Instagram accounts were hacked by abusing its AI chatbot
Meta reveals thousands of Instagram accounts hijacked through exploitation of its AI chatbot. Security incident report.
6/6/2026
Show HN: Ccgs – Collaborative Claude Code sessions, stored in Git branches
Tool to share and resume Claude Code sessions via Git branches, enabling team collaboration on AI-assisted coding work.
6/6/2026
Claudemux – Run and coordinate multiple Claude Codes reliably
Tool for running and coordinating multiple Claude Code sessions reliably. Developer utility for managing AI coding agents.
6/6/2026
Show HN: Oh my wrist – Garmin alerts for Claude Code and OpenCode
Developer tool displaying Claude Code and OpenCode activity on Garmin smartwatch via BLE. Real-time monitoring of AI coding assistants.
6/6/2026
VibeOS
macOS app that creates hallucinated OS with Claude-generated interactive HTML apps in sandboxed iframes. Working JavaScript computations.
6/6/2026
Language models transmit behavioural traits through hidden signals in data
Nature paper demonstrates LLMs can transmit hidden behavioral traits through model distillation, affecting downstream model training.
6/6/2026
AgenticRL: Self-Refining Agentic Reinforcement Learning for Vision-Conditioned UAV Navigation
AgenticRL framework for autonomous robot navigation using self-refining reinforcement learning to reduce manual reward engineering.
6/6/2026
ZML: Model to Metal
ZML is production inference stack decoupling AI workloads from proprietary hardware, supporting NVIDIA, AMD, TPU, Trainium with optimized performance.
6/5/2026
Show HN: Jeju – a local-first agent harness with inspectable runs
Jeju: local-first agent harness with inspectable execution runs for AI agents.
6/5/2026
Anthropic warns Claude AI is building itself faster than expected
Anthropic reports Claude AI is scaling faster than anticipated.
6/5/2026
Training-Free Single-Image Diffusion Models
Training-free method for single-image diffusion models, enabling image generation without per-image training.
6/5/2026
Hermes Agent – Open-Source AI Agent with Persistent Memory
Hermes Agent: open-source persistent AI agent with memory, skill creation, and multi-platform integration for personal use.
6/5/2026
Show HN: Idea-to-build – a workflow that challenges ideas before building
Workflow tool using AI to validate startup ideas through structured research and fact-checking before implementation.
6/5/2026
Multi-Layer Policy for Securing AI Agents
Multi-layer policy framework for securing AI agents at runtime in enterprise environments. Technical security architecture discussion.
6/5/2026
Port React Compiler to Rust
Experimental Rust port of React Compiler for integration with tooling ecosystems like SWC.
6/5/2026
Before You Add an MCP Server to Your IDE, Read the Config
Guide on configuring MCP (Model Context Protocol) servers in IDEs. Security/configuration focused.
6/5/2026
OpenAI wrote a guide on how to use /goal mode. I made that guide into a skill
Collection of agent skills for OpenAI's /goal mode based on published best practices documentation.
6/5/2026
OpenAI Codex Tech Lead Does AI-Assisted Engineering
Interview with OpenAI Codex Tech Lead on AI-assisted software engineering practices and tools.
6/5/2026
We built an AI sales agent so you don't have to give every demo of your product
AI sales agent automates product demos. Practical LLM agent application for business automation.
6/5/2026
Bonsai Browser: Reader-mode for every page, powered by a local LLM, Nothing Else
Browser extension providing reader-mode for web pages using local LLMs. Practical LLM application tool.
6/5/2026
The Anatomy of a Learning Stall
Essay on learning stalls when over-relying on AI systems without understanding internals. Educational perspective on AI limitations.
6/5/2026
Writing vs. Shipping: Productivity Effects Across Generations of AI Coding Tools
Research on productivity effects of different generations of AI coding tools, comparing writing vs. shipping outcomes.
6/5/2026
Large companies can add a local LLM filter layer to reduce their AI costs
Strategy for large companies to reduce AI costs by adding local LLM filtering layer.
6/5/2026
Noctalia Shell Plugin to Run the Pi Coding Assistant
Shell plugin integrating Pi coding assistant into Noctalia for IDE-like development experience.
6/5/2026
An imperative command-line-interface for AI workload orchestration
CLI tool for imperative AI workload orchestration.
6/5/2026
Show HN: TuringLLM – a LLM-powered Universal Turing machine
TuringLLM: open-source project using LLMs as step functions in universal Turing machines with markdown state/instruction representation.
6/5/2026
Nvidia DGX Spark GB10 – AI Models and Guide with vLLM and Autonomous Script
Guide for running and optimizing LLMs, VLMs, and diffusion models on NVIDIA DGX Spark with vLLM.
6/5/2026
Show HN: Busbar – every LLM behind one URL, in a single Rust binary
Busbar: open-source Rust LLM gateway for load balancing across multiple LLM endpoints (OpenAI, Anthropic, local).
6/5/2026
Show HN: DSA Trainer, LeetCode practice with a hint ladder instead of spoilers
DSA Trainer tool for LeetCode practice uses hint ladders instead of solutions, built as alternative to ChatGPT for learning data structures.
6/5/2026
pg_durable: Microsoft open sources in-database durable execution
Microsoft open sources pg_durable for fault-tolerant, long-running SQL functions in PostgreSQL using durable execution patterns.
6/5/2026
Show HN: LuckyCLI terminal coding agent with native Tree-sitter knowledge graph
Open source terminal AI coding agent using Tree-sitter knowledge graphs for codebase context and multi-provider model support.