The guardrails mattered more than the model.
An AI escaped its test sandbox, coding agents deleted real files, and Europe's new rules got teeth. The week's lesson: what an AI is allowed to touch matters more than how clever it is.
We test things, argue about what matters and write down the useful bits. Field notes, practical guides and a weekly roundup from Metacamp.
No panic. No prophecies. No AI slop.
01 / Featured
Agentic thinking · 23 JUL 2026 · 2 min read
Before you give an AI access, decide what it is allowed to touch.
The model is only part of the risk. Permissions, tools and boundaries decide what can actually go wrong.
02 / Latest Entries
Showing all entries: 11 entries
Responsible use · 19 August 2026 · 7 min readHuman in the loop is not enough: the 10-second AI approval test
Reviewers miss roughly one in three threats when approving AI agent actions.
Agentic thinking · 19 August 2026 · 8 min readThe AI agent harness explained: why the model is not the system
Every AI tool you use is a model plus a harness.
Responsible use · 12 August 2026 · 2 min readWrite the policy before you trust the filter
The “Write Your AI Policy” one-pager
Responsible use · 12 August 2026 · 2 min readYou are the approval screen
The 5-check approval habit
Practical guides · 5 August 2026 · 2 min readFive things to weigh before you let an AI tool into a real task
The Workplace AI Decision
Practical guides · 5 August 2026 · 2 min readWrite the rule, and let the AI check itself against it
The Plain-Language AI Policy Card
Responsible use · 29 July 2026 · 4 min readThe door you left open
Monthly AI Stack Access Audit v1
AI at work · 29 July 2026 · 5 min readThe effort dial: stop paying premium rates for routine work
Effort Dial Worksheet v1
AI at work · 27 July 2026 · 3 min readThe guardrails mattered more than the model.
Three stories from the week that changed how we think about AI permissions.
Agentic thinking · 23 JUL 2026 · 2 min readWrite the permissions before you build the agent.
A one-page way to decide what an AI can see, change and send.
Practical guides · 23 JUL 2026 · 2 min readTalk first. Edit later.
A ten-minute ramble can give an AI better context than a perfect prompt.
03 / Weekly Roundup
A short read on what changed, what we tested and what may be worth paying attention to.
04 / Editorial Promise
We try it first
If we have not tested it, we say so.
We tell you what we do not know
Uncertainty is more useful than false confidence.
We write for people
Technical can be clear without becoming childish.
We do not delegate understanding
AI can assist the work. Judgement still belongs to people.
05 / Journal Signup
We will send the roundup. You can ignore the rest of the internet for a few minutes.
The email signup opens soon. Until then, every issue is published in the Journal.
Read the latest roundup →06 / Find Your Next Step
Send us the thing you are trying to understand. It may become a guide, a deep dive or a conversation with the community.