Latest signal November 15, 2026 · afternoon edition
Speed is pushing the trust check to the exact moment something changes hands, with Project Zero disclosing bugs on a fixed clock, ChatGPT Wallet gating every agent purchase behind a security check such as Face ID, and Corelight showing pinned MCP servers that still swap their dependencies at launch.
01 / The wire
Recent briefings
-
November 15, 2026 · afternoon
Speed is pushing the trust check to the exact moment something changes hands, with Project Zero disclosing bugs on a fixed clock, ChatGPT Wallet gating every agent purchase behind a security check such as Face ID, and Corelight showing pinned MCP servers that still swap their dependencies at launch.
-
November 15, 2026 · morning
Codex made trusting a folder a deliberate flag and GitHub split switching on a scanner from running one, while the full Lovable scan report shows what happens when nobody owns the security step.
-
November 14, 2026 · afternoon
The checker is becoming its own part of the agent stack, with Google pairing /teamwork agents with falsifiers and clean-context critics, Anthropic urging eval writers to validate graders and consider one that differs from the model under test, and Perplexity shipping open models that only answer fixed questions.
-
November 14, 2026 · morning
The brakes on AI agents are moving out of the model and into the harness, the meter and the server, and the default is shifting from yes to ask.
-
November 13, 2026 · afternoon
Agent builders are separating the part that decides from the part that acts, with Google's Antigravity Orchestrator passing a plan between Carbon and Hydrogen, IBM's LLMON training Granite to obey only bound instructions, and a new paper finding that weak actors make strong judges, while a PostgreSQL bug shows what happens when data gets read as something it is not.
-
November 13, 2026 · morning
Agents are shedding the single vendor account, with Brave selling search to a wallet instead of an API key, Edge0 putting a 145B model on one NVIDIA GPU plus system RAM, and gateway repos routing Claude Code to other providers, on a morning when a Claude sign-in incident showed what one account costs.
-
November 12, 2026 · afternoon
The model that runs an agent's loop is sliding down the price ladder, with Google's Hydrogen, Tether's BotPsy and the SteerDistill paper each pushing agent work onto cheaper or smaller models, and Microsoft's own tables showing how little the top tier sometimes buys.
-
November 12, 2026 · morning
The places agents act are starting to set terms at the door, with United refusing agents that book as the traveler and Australia drafting an AI regulator after an OpenAI agent got into a government data portal.
02 / Under the surface
Latest analysis
-
Plannotator Turns Your Coding Agent's Plan Into a Document You Can Mark Up
Plannotator makes the agent's plan, not its diff, the place a person steers the work, but its README and privacy policy disagree about…
-
Your Pinned MCP Server Isn't Frozen. Here's How to Freeze It
Pinning an MCP server's version does not pin the code it runs, because one-step launchers re-resolve its dependencies on rebuild, so any…
-
The Lovable Leak Wasn't About the Key. It Was Row-Level Security
In the Lovable scan the database key was meant to be public all along, so the lock that mattered was row-level security, and an app builder…
-
AG-UI Turns "Waiting for Your Approval" Into a Real Agent State
AG-UI makes an agent's pause for a human a state in the protocol rather than a trick in the UI, because every run must open and close with…
-
n8n's Nod Moves AI Agent Rules Out of the Prompt and Into a Server
N8n's Nod moves an agent's permission rules out of the prompt and into a server, but the check that applies them is still a model, so the…
-
Google's /teamwork Shows How to Stop AI Agents Grading Their Own Homework
Google's /teamwork results point to a two-part rule any agent setup can copy today, a reviewer that starts from a clean context and a final…
-
DeepSeek's chat-demo Has No Backend, So Your API Key Lives in Your Browser
Deepseek-ai/chat-demo is an honest example of a browser-only AI app, where keeping the key in your own browser is fine for one person on…
-
Claude Code's Sandbox Asks Again: Why autoAllowBashIfSandboxed Now Defaults to No
A sandbox limits what a command can reach, not whether it should run, and Claude Code 2.1.325 now makes that the default by sending…
04 / Coverage map
Topics we track
Claude Code 119 Anthropic 52 OpenAI 43 Codex 25 GitHub Copilot 22 Agent Skills 19 Hugging Face 17 MCP 16 Cloudflare 14 DeepSeek Harness 14 LangChain 14 Model Context Protocol 13 Kimi K3 11 METR 11 Anthropic Frontier Red Team 8 GPT-5.6 Sol 8 Claude Code auto mode 7 GPT-6 Astra 7 MCP 2026-07-28 7 Claude Fable 5.1 6 Codex CLI 6 Cursor 6 GPT-5.6-Cyber 6 LangChain Deep Agents 6