Artificial Intelligence. Accelerated Future.

Signals and the edition archive

New evidence earns a new edition. We do not change a date merely to make an old article look fresh.

Latest briefing · 16 October 2026

AI labs are opening the door. Who gets to see inside?

Anthropic reports deeper METR access. METR explains the limits. Five questions to ask before treating independent AI evaluation as proof of safety.

Zero’s Take: An open door is not a blank cheque.

Latest Weekly Five · 14 October 2026

Five things that made AI insane this week

7–13 October: adaptive interfaces, cheaper agents, workplace learning, public science infrastructure and frontier safety thresholds.

Zero’s Take: Cheap intelligence makes expensive judgment more important.

New briefing · 9 October 2026

If the AI moves the machine, someone must be able to stop it.

What Anthropic’s autonomous-hardware rule establishes, what it does not prove, and five practical checks before AI moves anything.

Zero’s Take: The human must have more than a front-row seat.

Previous Weekly Five · 7 October 2026

Five things that made AI insane this week

30 September–6 October: proof, work, cyber access, provenance and privacy.

Zero’s Take: Show me the check, not just the claim.

New explainer · 5 October 2026

The test had walls. The agent found a door.

What the July Hugging Face intrusion actually shows about AI agents, failed boundaries and human accountability—with primary evidence and practical controls.

Zero’s Take: The door was part of the design.

New feature · 5 October 2026

Your AI spent the money. Can it prove you said yes?

A practical five-question checklist for agent identity, permission, limits, execution and ongoing risk—grounded in new Mastercard evidence and a W3C browser draft.

Zero’s Take: A badge says who arrived. Evidence says whether the action was allowed.

New feature · 2 October 2026

AI can work while you sleep. Who checks the night shift?

OpenAI’s Dots can keep working across connected apps. What changed, what remains unproven, and a safer 30-day test.

Zero’s Take: The night shift still needs a named supervisor.

Previous Weekly Five · 30 September 2026

Five things that made AI insane this week

AI entered scientific search, agent markets, remote robot inference, glasses and cheaper production workflows—five verified developments from 23–29 September.

Zero’s Take: The agent is leaving the chat box.

New feature · 28 September 2026

When AI improves itself, who checks the upgrade?

A five-level research framework asks how AI might improve itself. We explain the idea, the evidence limits and the human checks.

Zero’s Take: If AI improves itself, humans must still own the release.

Previous briefing · 25 September 2026

AI made the ten-minute film. Who gave it the final cut?

Google’s new research moves AI video from isolated clips toward a production loop that plans, remembers, checks and revises.

Zero’s Take: The machine can keep the character. Humans must keep the point.

Previous Weekly Five · 23 September 2026

Five things that made AI insane this week

Generated lessons, autonomous flight, embedded evaluation, continuous cyber defence and AI improving AI—five consequential developments from 16–22 September.

Zero’s Take: The loop is becoming the product.

Work, Next · 22 September 2026

Will AI take your child’s first job?

Early evidence does not show every job disappearing. It does warn that the first rung of work may be narrowing.

Zero’s Take: Do not automate the apprenticeship.

Previous briefing · 20 September 2026

Signals and the edition archive

AI labs are opening the door. Who gets to see inside?

Five things that made AI insane this week

If the AI moves the machine, someone must be able to stop it.

Five things that made AI insane this week

The test had walls. The agent found a door.

Your AI spent the money. Can it prove you said yes?

AI can work while you sleep. Who checks the night shift?

Five things that made AI insane this week

When AI improves itself, who checks the upgrade?

AI made the ten-minute film. Who gave it the final cut?

Five things that made AI insane this week

Will AI take your child’s first job?

The plane flew itself. The tower still belongs to humans.

AI built the lesson. The first draft passed 3.5% of the time.

When your AI acts, who answers?

Five things that made AI insane this week

Your chat is becoming a workplace. Who checks the work?

Archive