6/12/2026
SkillSpector
Security scanner detecting vulnerabilities and malicious patterns in AI agent skills before installation, analyzing 26.1% vulnerability rate.
6/12/2026
I created an AI agent that finds investors that and does reachouts to investors
CLI tool using AI agents to find investor contacts and automate outreach emails; user reports 14% reply rate from 43 emails sent.
6/12/2026
Show HN: Atlas – local-first memory that re-evaluates beliefs when facts change
Open-source local-first cognitive memory using AGM belief revision on property graphs with propagation engine for updating downstream beliefs.
6/12/2026
ACM warns vibe coding skips core engineering practices
ACM warns that 'vibe coding' with AI skips fundamental software engineering practices and discipline.
6/12/2026
Show HN: Rust Implementation of Symphony
Desktop app that orchestrates Claude and Codex agents to autonomously work on Linear board issues in sandboxed workspaces.
6/12/2026
Show HN: An AI coaching team in Claude Code that's forbidden from judging you
Claude Code-based AI coaching system implementing the Intelligence Emotions model with five specialized coaches and local session memory.
6/12/2026
/architect: Reduce Fable tokens by 80%, Fable orchestrates/reviews, Codex builds
Claude and GPT-5.5 agents orchestrate code architecture and building tasks with 80% token reduction, running on existing subscriptions without API costs.
6/12/2026
Show HN: Loomabase – Offline-first sync for SQLite and PostgreSQL
Open-source offline-first sync engine for SQLite/PostgreSQL using Rust with CRDT conflict resolution at column granularity.
6/12/2026
Show HN: Rubric – test what your LLM agent did, not just what it said
Testing framework for LLM agents. Monitors tool calls, arguments, trace, latency—behavior not just output. Catch regressions in CI.
6/12/2026
ImpactArbiter – A PyTorch autograd trap for LLM-generated vLLM/SGLang invariants
PyTorch tool detecting silent failures in LLM-generated unit tests for vLLM/SGLang kernels. Addresses hallucination in test generation.
6/12/2026
What Is an LLM Control Plane?
Conceptual article on LLM control planes for production AI systems. Covers routing, budgets, privacy, fallbacks, provider management.
6/12/2026
Reasoning as Pattern Matching: Shared Mechanisms in Human and LLM Reasoning
arXiv paper abstract about pattern matching mechanisms shared between human and LLM reasoning. Limited content provided.
6/12/2026
If you are asking for human attention, demonstrate human effort
Discussion on team etiquette when sharing AI-generated code and documentation output. Addresses AI fatigue and integration practices.
6/12/2026
General purpose LLMs outperform specialized clinical AI on medical benchmarks
Nature Medicine study comparing frontier LLMs (GPT-5.2, Gemini 3.1 Pro, Claude Opus 4.6) to specialized clinical AI tools on medical benchmarks.
6/12/2026
Hacking Google with A.I. For $500k
Security researcher demonstrates AI-powered vulnerability discovery across Google infrastructure, finding 1,500 APIs and earning $500k in bounties.
6/12/2026
GatekeeperAI – self-hosted governance platform for AI apps your team is building
GatekeeperAI is self-hosted platform for enterprise governance of third-party and internal AI applications with security scanning.
6/12/2026
The 98% Problem: A Survey of Harness Engineering for AI Agents
Survey of harness engineering for AI agents, analyzing production deployment patterns and the layer below model selection.
6/12/2026
Ask HN: How to deal with agents constantly messing up padding/alignment in UIs?
Discussion of AI agents struggling with UI padding/alignment due to lack of visual feedback loops during generation.
6/12/2026
Cortex – Agent-Native Knowledge OS on Markdown (Karpathy's LLM Wiki, via MCP)
Cortex is agent-native knowledge OS on Markdown with MCP interface for AI agents to read/write/search knowledge graphs from file uploads.
6/12/2026
Measuring LLMs' impact on N-day exploits
Research measuring LLM impact on N-day vulnerability exploitation and real-world security harm from publicly disclosed bugs.
6/12/2026
A Fake Bug Report Hijacks Your AI Coding Agent – and Nothing Catches It
Tenet Security demonstrates 'Agentjacking' attack hijacking AI coding agents via fake bug reports through Sentry APIs.
6/12/2026
Run local agentic AI on the Mac using MLX (WWDC 2026) [video]
WWDC 2026 video on running local agentic AI on Mac using MLX framework. Developer tools for local AI agents.
6/12/2026
Ask HN: What do you do when you hit Claude subscription limits?
Discussion of Claude API rate limits and workarounds for multi-agent workflows. Community experiences optimizing LLM agent usage at scale.
6/12/2026
Show HN: Open infrastructure for Argentina's AI-run companies (36 NPM packages)
36 NPM packages for open infrastructure targeting AI companies in Argentina. Limited details provided.
6/12/2026
Small Context, High Parallelism: How To 10x Reduce Agentic Coding Costs
Optimization techniques for reducing API costs in agentic coding by parallelizing tool calls and minimizing context resubmission. Technical analysis of LLM agent efficiency.
6/12/2026
Show HN: Setup your Company Brain in minutes
Centralized vector database for ingesting multi-app data to support AI agents via MCP. Addresses context retrieval and memory management for agentic systems.
6/12/2026
AI agents that quickly estimate electronic devices' carbon footprints
UW research: AI agents estimate electronics carbon footprints, addressing gap in device sustainability information for consumers.
6/12/2026
Launch HN: BitBoard (YC P25) – Analytics Workspace for Agents
BitBoard: YC-backed agentic analytics workspace enabling agents and users to collaborate on live dashboards and data analysis.
6/12/2026
Sandboxes Are Not Security
Analysis arguing sandboxes alone are insufficient for AI agent security; agents can misuse valid credentials despite isolation. Critical infrastructure security for AI agents.
6/12/2026
Tech Writers May Be Solving the Wrong AI Problem
Documentation written for humans poorly serves AI systems consuming it. Explores gap between human and machine comprehension of implicit context.
6/12/2026
AI Splits the Market
Analysis of how AI technology impacts different software markets differently, challenging uniform disruption narratives. References NBER paper.
6/12/2026
Skill Workshop: Turn Agent Work into Reusable Skills
OpenClaw Skill Workshop: system for agents to learn and reuse procedures from repeated work with human review before application.
6/12/2026
An AI agent that must produce evidence before it can say "done"
Distill: self-hosted AI agent requiring physical evidence before task completion, learns reusable skills across sessions with versioned rollback.
6/12/2026
F-bombs don't make LLMs smarter
Research testing whether 'swearing' in prompts improves LLM reasoning. Null result study on prompt engineering effects on o1 model performance.
6/12/2026
Show HN: SharkClean MCP
MCP server for SharkClean robot vacuums enabling agentic control via Claude. Open-source tool integrating IoT devices with LLM agents through cloud API.
6/12/2026
Ask HN: Favorite prompts for improving LLM output?
Discussion thread asking for effective prompting techniques to improve Claude and GPT output quality and performance.
6/12/2026
Show HN: Tokenbrook Vale, a cozy office village for your AI agents
Tokenbrook Vale is open-source visualization tool for running multiple Claude AI agent instances in retro pixel art office environment.
6/12/2026
Show HN: Janus – MCP that collects context from browser and terminal
Janus MCP server collects browser and terminal context to simplify prompting AI coding agents with interaction history.
6/12/2026
Show HN: My open source agent built and launched its own business in 48 hours
SmithersBot open-source AI agent autonomously built and launched a business in 48 hours to solve payment verification for autonomous agents.
6/12/2026
Show HN: VibeClip – open-source AI video editor you control by chatting
VibeClip is open-source AI video editor controlled via natural language chat interface.
6/12/2026
Show HN: StackScope – I crawled over 40k indie launches to see what they ship
StackScope crawler analyzes 40k+ indie product launches to identify technology stacks, hosting, frameworks, and AI builder signals used in real applications.
6/12/2026
From ML engineer to AI-native: reskilling toward an edge
Essay on ML engineer career transition to AI-native roles, featuring AI agents automating data pipeline scaffolding and model boilerplate with concrete 90-day reskilling plan.
6/12/2026
Claude dashboard – local web interface
Local web dashboard for Claude Code sessions with live monitoring, history, usage charts, and token cost analysis.
6/12/2026
Show HN: Apodex-1.0 – a deep-research agent team that verifies its own evidence
Apodex-1.0 is a deep-research AI agent team that performs science-based verification of its own evidence. Limited technical details provided in excerpt.
6/12/2026
Show HN: I had Fable clone a game, it turned it into an arena combat AI trainer
Developer used Fable to clone game Soldat and built neural network AI bot trainer from simulated matches, with live dashboard and replay system. Demonstrates LLM code generation capabilities.
6/12/2026
Show HN: TetherDust – Self-hosted AI Analytics Engineer (open source)
TetherDust is a self-hosted open-source AI Analytics Engineer tool. Title-only submission with no technical details.
6/12/2026
Why watch Claude "discombobulate" when you could watch low res YouTube videos?
Tool that converts videos to ANSI pixel art and displays them in Claude Code statusline during agent execution.
6/12/2026
How to automate Instagram engagements with computer vision (and get banned)
Demonstration of automating Instagram engagement using computer vision to overcome DOM obfuscation.
6/12/2026
AI is about to get fast, and it's never going to slow down
Analysis of Claude Fable 5 achieving top performance across multiple benchmarks (Artificial Analysis, LMArena, SWE-Bench Pro, FrontierCode), discussing acceleration pace of frontier AI models.
6/12/2026
Bio input based, instead of vision based, physical AI for industrial bio
Physical AI system for autonomous industrial bioprocess operations using bio-input sensing instead of vision, requiring safety grading mechanisms for autonomous decision loops.