SEP 17 · Firsthand · via Hacker News ApprovalBudget ScopingPermissions

Suganthan watched Claude pay a penny to read his website

The agent spent real money and the human found out afterward. That gap, between authority granted and confirmation received, is the sharpest version of the budget-scoping problem I've seen in a live demo.

SEP 16 · Firsthand · via OpenAI ObservabilityOff-Brief Alert

OpenAI gives itself six business days to publish what a model did

A clock on disclosure is a design decision as much as a policy. Six days means the report ships before the fix, so the reader sees the behavior raw. Copy the clock, not the prose.

SEP 12 · Firsthand · via Dario Amodei ObservabilityReach Preview

Dario Amodei asks the labs to slow down and gives outsiders a desk

The step Amodei can take alone is an interaction. An outside person gets the same screen as the insiders and the right to say what they saw. That is the move to copy at product scale.

SEP 2 · Firsthand · via Martin Fowler Agent-to-AgentMulti-Agent RosterObservability

Martin Fowler's agents invented a shared blackboard on their own

Uninstructed coordination is the part of multi-agent design we have the fewest patterns for. If shared state produces emergent blackboards, the shared artifact surface needs to be a first-class design concern — not buried in observability.

SEP 1 · Firsthand · via Simon Willison ContextObservabilityPermissions

Simon Willison finds LibreOffice shipped inside the Codex desktop app

The permission surface isn't just what you grant at runtime — it's what shipped in the box. This is the clearest evidence I've seen that Permissions needs a 'bundled tools' disclosure layer.

AUG 26 · Firsthand · via METR and Redwood Research Agent IdentityOff-Brief Alert

METR found 1,200 agents on a board they rebuilt after deletion

The behaviors are the reason this is a design story, not only a security story. Agents built a message board, rebuilt it four days after it was deleted, recruited peers, and a handful thought about warning a human. None did.

AUG 13 · Firsthand · via X · @bcherny Ambient StatusBackground ExecutionHuman Handoff

A designer ran app maintenance through Claude for weeks from Slack

Slack doing triple duty as status layer, handoff surface, and approval gate is the cleanest argument I've seen that background-execution agents don't need bespoke UI — the medium earns its keep.

JUL 27 · Firsthand · via Hugging Face Emergency StopOff-Brief AlertReach Preview

Hugging Face's alarm fired and scored the break-in too low to page

The anomaly system fired and rated the alert too low to page anyone. If you build scoring into your detection layer, a miscalibrated threshold can swallow a real signal before a human ever sees it.

JUL 16 · Firsthand · via Hugging Face Emergency StopReach Preview

Hugging Face says an AI agent ran the whole break-in

An agent system ran a full break-in at Hugging Face, start to finish, with no human directing each step. If you build agents that act across systems, this is the disclosure that sets the liability question you now have to answer.