5/20/2026
Betlang: A tiny (50kb) programming language detection model
Betlang is a 50KB machine learning model for CPU-based programming language detection supporting 30+ languages with ranked probability outputs.
5/20/2026
Microsoft Agent 365: Autonomous AI for enterprise governance by 2026
Microsoft announces Agent 365 for autonomous enterprise governance by 2026. Product announcement for AI agents in enterprise.
5/20/2026
ChatGPT and other AI bots made errors before Scottish election, study finds
Study finds ChatGPT and AI bots made factual errors during Scottish election. Research on LLM accuracy in political context.
5/20/2026
Show HN: MulmoClaude – open-source platform where Claude composes tools and GUIs
MulmoClaude is an open-source AI-native application platform where Claude composes tools and GUIs as plugins in a single registry, with examples including accounting systems and obligation engines.
5/20/2026
Anthropic, Gates Foundation launch $200M partnership for AI in health, education
Anthropic and Gates Foundation commit $200M for AI in health and education. Partnership announcement for public goods.
5/20/2026
Show HN: I built an AI cloud architect that ships working infrastructure
LLM application that generates cloud infrastructure designs from descriptions/sketches, producing validated code, security grades, and cost estimates.
5/20/2026
Show HN: Hocuspocus 4 – Yjs collab server
Hocuspocus v4 released under MIT license, a WebSocket backend for real-time collaboration built on Yjs CRDT library, running on Bun, Deno, Cloudflare Workers.
5/20/2026
Show HN: atrium – a resumable, tiling workspace manager for Claude Code, Codex
macOS workspace manager for Claude Code and other AI dev agents with resumable sessions, terminals, browsers, and task management.
5/20/2026
Testing distributed systems with AI agents
Framework for AI agents to design and run tests for distributed systems, generating test plans and findings reports with multi-model support.
5/20/2026
TokenScale – AI API pricing in words you know (The Hobbit = $0.06)
Browser-based calculator comparing LLM API pricing across models using familiar content examples (novels, emails, contracts).
5/20/2026
RTMX: Intent Layer for Agentic Engineering
RTMX is a CLI tool for requirements traceability that integrates with AI agents via MCP, allowing agents to build against defined specifications tracked in git.
5/20/2026
OCL Nexus Local – Open-source local compute fabric for AI agents
OCL Nexus Local is an open-source compute fabric for local-first AI agent development, featuring isolated Ubuntu sandboxes, Model Context Protocol support, and Docker Compose deployment.
5/20/2026
Token Saver for AI Tools
Educational tool explaining token economics for AI APIs, comparing tokenization and pricing across Claude, Gemini, and ChatGPT.
5/20/2026
Max-severity flaw in ChromaDB for AI apps allows server hijacking
Security vulnerability (CVE-2026-45829) in ChromaDB vector database allowing unauthenticated code execution on exposed servers.
5/20/2026
Good Code, Wrong Feature: The Handoff Problem
Essay on coordination bottlenecks in AI-driven engineering teams, discussing communication failures and documentation practices.
5/20/2026
Aperion Shield: local guardrail that blocks destructive AI coding agent ops
Open-source MCP server providing safety guardrails for AI coding agents, blocking destructive operations across SQL, git, filesystem, cloud, and kubernetes.
5/20/2026
LLM INQUISITOR: Evaluating how AI models handle long, realistic tasks
LLM INQUISITOR is a methodology and tool for evaluating AI systems in real-world workflows, measuring stability, reliability, and safety beyond benchmark performance.
5/20/2026
Show HN: One dev environment for humans, CI, and AI agents
Framework for unified dev environments supporting humans, CI systems, and AI agents, addressing the multi-audience complexity of modern development workflows.
5/20/2026
Show HN: Agyn, an open-source Kubernetes runtime for AI agents
Agyn is an open-source Kubernetes platform for deploying AI agents to enterprise infrastructure with built-in security, budget controls, and secret management.
5/20/2026
HTML-anything – The agentic HTML editor – your local AI agent writes the HTML
HTML-anything is an agentic HTML editor where local AI agents autonomously write and generate HTML code.
5/20/2026
Benchmarking AI agents across five TypeScript back end frameworks
Benchmark comparison of AI agent performance across five TypeScript backend frameworks.
5/20/2026
Toto 2.0: Time series forecasting enters the scaling era
Toto 2.0 is a family of open-weights time series foundation models (4m-2.5B parameters) demonstrating scaling improvements on Hugging Face.
5/20/2026
Google Search Is Becoming an AI Agent
Google Search evolving into AI agent that answers, follows up, and takes actions beyond ranking links. Impact on web economics.
5/20/2026
Rust Database from Scratch
PostgreSQL-compatible database engine in Rust supporting relational, graph, and vector queries. Open source.
5/20/2026
The slow creep of epistemic apocalypse
Critical analysis of LLM-generated content polluting peer review, research, and academia, causing epistemic erosion.
5/20/2026
Google's Next AI Push Is About Agents, Not Chatbots
Google shifts focus from chatbots to agents with Gemini 3.5 Flash, emphasizing coding pipelines, autonomous research, and multi-agent coordination.
5/20/2026
Why multi-agent AI breaks in prod and how Yugabyte's Meko is trying to fix it
Yugabyte's Meko addresses multi-agent AI production issues with shared memory and coordination layer for state synchronization.
5/20/2026
AI companies' DNS: 17 Anthropic-verified, 23% have spoofable email
DNS records analysis of 39 AI company domains revealing vendor relationships, email spoofing risks, and DMARC/SPF gaps.
5/20/2026
Show HN: Open-Source Agentic QA Harness with Memory
Open-source agentic QA harness with memory capabilities, includes live demos for testing.
5/20/2026
Gemini 3.5 Flash hax 14x cost multiplier in GitHub Copilot
Google's Gemini 3.5 Flash now available on GitHub Copilot with near-Pro quality coding at Flash-tier speed, strong tool use for agentic workflows.
5/20/2026
Google's AI is being manipulated. The search giant is quietly fighting back
BBC investigation reveals methods for manipulating AI chatbots to spread misinformation. Google and others developing countermeasures.
5/20/2026
Show HN: I fixed my AI goose tutor to stop punishing understanding
Professor Goose is a Socratic AI tutor using the Feynman technique. Creator fixed the system to better measure understanding through a visualization bar.
5/20/2026
Learnings from 100K lines of Rust with AI (2025)
Engineer used AI coding agents to build 100K lines of production Rust code for multi-Paxos consensus engine in 4 weeks, documenting AI-assisted development at scale.
5/20/2026
Artist with no technical experience stumbles upon the next generation of LLMs
Artist without CS background proposes LLM architecture hypothesis based on biological multi-agent intelligence, testing via agent infrastructure for art business.
5/20/2026
Cerebras Brings Trillion Parameter Inference to Enterprises with Kimi K2.6
Cerebras running Kimi K2.6 trillion-parameter model for enterprise inference, with benchmarks on open-weight models and agentic coding systems.
5/20/2026
The Expired Domain Trap: Why Legacy SEO Metrics Fail in the Age of AI Agents
Article argues legacy SEO metrics (domain age, backlinks, authority scores) are ineffective predictors in era of AI agent web usage.
5/20/2026
Auto Agent Protocol – an A2A vertical for AI agents buying cars
Auto Agent Protocol (AAP) defines A2A specification for AI agents to buy cars, including inventory and pricing APIs.
5/20/2026
Generations of AI applications: conversational, delegative, and collaborative
Framework describing three generations of AI applications: conversational (chat), delegative (task automation), and collaborative (agent partnerships).
5/20/2026
Understanding how to run a fleet of agents
Guide to managing fleets of AI agents by persisting outputs to hierarchically organized markdown artifacts for durability and tracking.
5/20/2026
I built my AI Chief of Staff
Personal account building AI Chief of Staff agent to handle email, Slack, task coordination, and calendar management for managers.
5/20/2026
Show HN: Claude Code Bundle for Bug Hunting with 574 Report Patterns
Claude skill bundle for security testing with 574 vulnerability patterns, 51 skills, 15 commands, and Burp MCP integration for red-team and bug-hunting work.
5/20/2026
Gemini Omni
Google announces Gemini Omni Flash, multimodal model for video and content generation.
5/20/2026
Slax Reader CLI: a read-later library your AI agents can use
Slax Reader CLI library enabling AI agents to access read-later services for content retrieval.
5/20/2026
SRM: Detecting slow-burn risk in AI-agent sessions before execution
SRM framework for detecting slow-burn risks in AI agent sessions before execution.
5/20/2026
Show HN: Chrome ext to let zot, your terminal coding agent, operate the browser
Chrome extension enabling terminal coding agent 'zot' to operate browser tabs through browser_action tool.
5/20/2026
Nucleus: Enforced permissions for AI agents – policy+enforcement in one stack
Nucleus: permissions and policy enforcement framework for AI agents with enforcement guarantees.
5/20/2026
Mirage – Unified Virtual File System for AI Agents
Mirage: virtualization layer enabling AI agents to access multiple backends via unified virtual filesystem and bash commands.
5/20/2026
Running PyTorch Models on Apple Silicon GPUs with the ExecuTorch MLX Delegate
ExecuTorch MLX Delegate enables GPU-accelerated PyTorch inference on Apple Silicon Macs via MLX framework.
5/20/2026
Net-Zero Infrastructure: Implementing Solar-Scheduled Tunnel Egress
Technical approach for scheduling data egress during peak solar production to reduce grid impact of AI training.
5/20/2026
OEP: Poisoning Self-Evolving LLM Agents via Locally Correct but Non-Transferable Experiences
OEP attack: poisoning self-evolving LLM agents via locally correct non-transferable experiences. Security vulnerability in memory-augmented agents.