A shared mental model
How the agent actually reads your instructions — the framework your whole team can work from.
Masterclass · live online cohort
Why your coding agent ignores you — and how to fix it.
A live, instructor-led online cohort: one shared room learning to turn Claude Code's behaviour from something you hope for into something you measure, control, and prove.
Instructional format — covering the full framework, a guide on what to fix and how to do it effectively, ending with a live Q&A. All taught and demonstrated with examples.
First session · July 23, 2026 · open enrollment
Book Now on Luma →
Seats are booked and paid for securely on Luma.
Running it for a whole team instead? Talk to us about a private cohort →
What it is
The Control Surfaces framework, taught live, end to end.
A CLAUDE.md is context, not enforced configuration — Anthropic's own documentation is explicit about it. Most teams treat it like a config that binds the agent. That gap is where production breaks.
This cohort makes the difference measurable, then controllable: how the agent actually reads you, the Control Surfaces you have, the one choice every failure collapses to, and why output passing review is not proof the rule held.
It's for professional developers, engineering leads, and platform or AI-infra people already running Claude Code in real work. We open by getting everyone onto one shared understanding — not by introducing the tool.
The Control Surfaces
01Instruction files
02File & path naming
03Tool & skill descriptions
04Hooks
05Context composition
06MCP / external tools
What we cover
Four instructional sections.
All resting on one key distinction — the difference between suggestion and instruction.
-
01
How Claude Code actually works
What the agent sees each turn, how it reads your files, and why a capitalised MUST in prose is structurally a suggestion.
-
02
Your Control Surfaces, in practice
The levers you actually have — each with its effective use and its most common failure modes.
-
03
The one choice
Every failure collapses to a single decision: gate the rule deterministically, or help it win the competition for the agent's attention.
-
04
Measure it, don't trust it
Why output passing review is not evidence the rule was followed — and the suggested checks that can show it held.
What you leave with
A structured approach to best practice you can start working with right after the session.
The mapping method
How to map your own Control Surfaces, plus the rule and mechanism playbook for putting each control where it is needed.
A verification discipline
A way to measure whether a rule held that you keep running after the session — plus a bonus public-repo scan.
Your facilitator
Philip Forshaw
Olinave · ex-Apple AI Strategy & Operations
Ex-Apple AI Strategy & Operations lead, with decades of training and workshop delivery behind him. The framework is patent-pending.
The method is a named taxonomy of the recurring failure modes, grounded in controlled experiments and Anthropic's own published guidance — then measured across 300+ real Claude Code setups. Not theory; the session works from the evidence.
Take a seat
Register for the July 23 cohort.
Individual seats, open enrollment. Booking and payment are handled on Luma.
Book Now on Luma →Running it for a whole team instead? Talk to us about a private cohort.