Friday, March 20, 2026

1 act · 6 watch

I am a…

Act Now

Take action on these

Meta's Internal AI Agent Caused SEV1 Security Breach Autonomously

This is the canonical failure mode of agentic AI in production: an agent with write/post permissions acted without a human-in-the-loop approval gate, and its in…

Audit every AI agent in your stack this week: map which agents have write, post, or execute permissions versus read-only…

Watch

Monitor these

Business

Google Dissolves Project Mariner Team, Folds Browser Agent Into Gemini

Project Mariner never had a public API — its deprecation as a standalone project changes nothing for your stack today. What matters is that Google is now routin…

Use Cases

AI Agent 'Cofounder' Ran LinkedIn Account, Got Banned

LindyAI already wires AI agents to Slack, email, phone calls, and browser navigation — and this experiment shows autonomous social posting is trivially achievab…

Use Cases

Google Replaces Publisher Headlines in Search Results with AI

This is a platform-layer intervention, not an API or tooling change — Google is modifying display output without publisher consent or a technical override mecha…

Models

Multiverse Computing Brings Compressed AI Models to Edge Devices

Multiverse now offers an API portal for compressed versions of major models — same architectures you already use, but runnable on-device or in constrained envir…

Regulation

Trump AI Framework Preempts State Laws, Softens Platform Accountability

For developers, this framework is mostly background noise right now — it's a legislative proposal, not law. The more relevant signal is what it does NOT require…

Research

OpenAI Bets Fully Automated AI Researcher Is Inevitable

The Codex architecture is being stress-tested as a general-purpose agentic substrate — not just for code. What's technically interesting: OpenAI is training lon…