LEARN BY DOING.
This is the public home for material used in my AI training. Start with the kind of AI responsibility you are trying to build, then open the relevant course or session.
Open for everyoneUnderstand the three core course choices, where the AI Field Guide deepens each stage, and the difference between prompting exercises, self notes and case evidence.
Nothing yet. Choose Work With AI, Give AI Work or Run AI first. Use the AI Field Guide when you want deeper Prompt, Context, Harness, Loop or Graph Engineering.
Choose the core programme that matches the responsibility you want to learn, then use the relevant field guide when you want deeper explanation, experiments or implementation.
A clear starting point and an understanding of how the open learning library and field guide fit together.
How to read the activity labels
Only the clearly marked prompt or instruction should be copied into AI. Add the labelled case material when instructed.
Workbook questions and reflection cards are your notes. They are not prompts unless a session explicitly says otherwise.
Case material supplies facts. Later evidence stays closed until the exercise tells you to reveal it.
WORK WITH IT. GIVE IT WORK. RUN IT.
Choose one of three complete courses. Each has three arcs across 3 days and 2 nights: eight daytime sessions, two evening studios and a practical workbook. Start with the responsibility you want to understand.
Before you begin: bring a computer or phone with access to an AI chat for marked prompting exercises. A small non-sensitive work task is useful but optional. Each course has its own workbook and links to a fictional starter case if you do not have an example. No coding is required for the core participant route unless an activity says otherwise.
Work With AI
Use AI deliberately on real work: direct it, ground it in the right context, inspect the result, create useful outputs and preserve methods that work again.
Open course →Give AI Work
Move from asking for answers to controlled action, bounded responsibility, explicit authority, clean handoffs and escalation when AI should stop.
Open course →Run AI
Decide whether an AI capability is ready to rely on, then define ownership, evidence, monitoring, recovery, economics and the path to reduce or stop autonomy.
Open course →SEVEN CAPABILITIES. THREE COURSES.
USE → CONNECT → CREATE belong to Work With AI; ACT → DELEGATE → COORDINATE belong to Give AI Work; OPERATE belongs to Run AI. These are complementary capabilities, not a compulsory ladder or a contest for the most autonomy.
REAL WORK. VISIBLE PROOF. SOMETHING TO KEEP.
Training starts with a participant's real task, workflow or system, not a tour of tools. Experience a limitation, understand the underlying method, test the improvement and take a reusable playbook, blueprint or charter into your next piece of work.
Bring real work
Use a familiar task or the shared fictional case rather than disconnected demonstrations.
Learn through friction
See the failure or limitation before introducing the engineering model that explains it.
Prove the change
Use before/after evidence, second runs, counterexamples and decisions you can defend.
Keep what works
Leave with a reusable practice, artifact or operating rule instead of relying on one lucky answer.
GO DEEPER AT THE STAGE WHERE THE PROBLEM APPEARS.
9order's AI Field Guide records the deeper engineering behind the three courses. Prompt and Context belong to Work With AI; Harness and Loop belong to Give AI Work; Graph belongs to Run AI.
Engineer the collaboration
Prompt Engineering shapes intent and steering. Context Engineering shapes the reality, evidence and sources available to the model.
Open Work With AI field guides →Engineer delegated work
Harness Engineering makes the run dependable. Loop Engineering makes repeated execution autonomous without making it unbounded.
Open Give AI Work field guides →Engineer the operated system
Graph Engineering makes coordination, state, routes, authority, recovery and system outcomes explicit.
Open Run AI field guide →START FROM WHAT YOU NEED TO CHANGE.
You do not need to remember a course name or session number. Start with the work problem in front of you.
FROM THE LESSON TO THE BUILD.
ShoStack is the open method behind my product-building work: repository evidence, service contracts, safe implementation, verification and handoff. It is separate from the three AI courses, but a useful place to apply the Harness Engineering lessons.
PUBLIC SESSION DATES.
No confirmed public dates are currently posted. When a public workshop is announced, its date and registration details will appear here and with the relevant course. All published learning material remains open in the meantime.
TEACH WHAT YOU ACTUALLY USE.
I'm Sho. I build and operate AI-enabled systems, software and workflows, and teach from that work. The aim is practical judgment: direct the work, check the evidence, and make useful methods repeatable.
My work on products including Alamak.my, Language Gym and Aussie Campervans shapes what I teach. ShoStack shares the transferable engineering approach without exposing private implementations. For a concrete business application, explore ACOM.