What we will do
To decide whether you can trust an AI agent you need to know who controls its three parts: the model (the weights that predict text), the harness (the code around the model that gives it tools, permissions, memory and a loop) and the runtime (the machine it runs on, which decides what it can touch).
After a ten-minute introduction, everyone builds an agent on their own laptop with our Agent Dev kit. Claude Code writes each step in front of you; you decide, watch and test. By the end your agent has tools and a loop, has been tricked by a planted instruction, has been protected against it, has had its model swapped and its runtime fenced in. You leave with a table of who controls what in the agents you already use.
Who it is for
Anyone curious about how these things actually work: people who use AI at work, people who decide whether their team may, people who want to look under the hood before trusting it. No programming knowledge needed.
What to bring
- A laptop you can install software on, and its charger.
- Claude Code set up before the day. The Agent Dev kit walks you through it, including a near.ai key for the models your agent will call. Plan an hour for this; your confirmation email has the link.
When and where
Friday 6 November 2026, in person. The start time and the venue are being finalised and will be in your confirmation email.
There are 20 places, given out first come, first served. When you sign up you join the queue, and we confirm your place by email within 24 hours.
Sign up
Two ways, both free. All we need is an email address; once it is confirmed you are in the queue, and we confirm your place within 24 hours.
With your Google account
Google tells us your email address and nothing else, and your place is taken straight away.
With your email address
What we keep, and for how long: see the privacy policy.