review
When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to…
A hijacked agent can only do what its tools allow. Why shell blocklists fail, how allowlists leak, and a test you can add without running a model.
"What used to take me six months, now takes me six hours," says Jorge Castro about using agents on his open source project. But AI has a catch: "You will be bus…
My popular mathematics book, “Six Math Essentials” (Six Maths Essentials in the UK), which was previously announced on this blog in February, is now available (…
CREA has defended modelled estimates linking Sasol’s Secunda emissions to about 1,000 premature deaths and R19bn in annual health-related economic damage, while…
Link to podcast In 2026, banks started telling James Governor they would no longer have humans review all of their code. AI code review went from experiment to …
What's changed Added x-claude-code-request-class , x-claude-code-agent-type , x-claude-code-prev-tool-durations , x-claude-code-compaction and x-claude-code-con…
I did not care for this one bit. I found it cloying, infantilising, and far too circuitous. It starts off as a bit of an overwritten melodrama, where every phra…
A decade ago, I read Ayoade's first book. In my review, I said: It's hard to know how to assess a book like this. It's far too clever for its own good - a frust…





