Thinking with AI · Coming soon
Your prompt worked once, then failed. The fix is the context.
When the same request gets a good answer on Monday and a poor one on Tuesday, the model usually did not see what it needed. This course teaches you to decide what an AI sees for a task — instructions, examples, documents, constraints — and to save it as a context pack you and your team rerun.
The course: Context engineering: what replaces prompt engineering
$19 $39 early-bird price
This course is being built now. Join the early-access list to reserve the early-bird price and shape what it covers.
The course is still being built. Joining the list is free, and you pay only if you buy at launch, at the early-bird price.
If this sounds like your week
- A prompt that gave you a great answer last week gives you something generic today, and you cannot say what changed.
- You keep adding sentences to the prompt, and each fix breaks something that used to work.
- Your team uses the same tool and gets different quality, because each person prompts in their own way.
- You suspect "context engineering" is the answer but every article about it is written for people building agents.
What you do now, and where it stops working
Today you rerun the request until a good answer appears, or you paste a longer prompt from your saved-prompts folder. Neither tells you why the good answer happened, so you cannot repeat it on purpose or hand it to a colleague.
What changes after the course
- When an answer is bad, you check what the model could see before you touch the wording.
- Your three most common tasks each have a context pack you open and run, instead of a prompt you rewrite.
- A colleague can run your pack and get work of the same standard without asking you how.
- You hold one before-and-after comparison on your own task that shows what the context changed.
How it works: The context pack
- 1Diagnose: take one task that fails intermittently and sort the failure into missing instructions, missing examples, missing material or conflicting rules.
- 2Select: list everything the task could use, then cut to what the model needs for this step — more context is not better context.
- 3Package: write the pack as standing instructions, two worked examples, the source material and a definition of a finished answer.
- 4Test: run the pack against the old prompt on the same inputs and keep the version that holds up.
- 5Share: store the pack where your team works so the standard lives in a file, not in one person.
What you get
- Context design worksheet
- Context pack templates for writing, analysis and research tasks
- Diagnosis checklist for bad outputs
- Video walkthrough of a task rebuilt from prompt to context
This is for you if
- You use ChatGPT, Claude or Copilot for real work most days and the results are uneven.
- You lead a team and want everyone to get the same standard from the same tools.
- You have outgrown prompt tips and want to know why a request works.
Skip it if
- You want a library of copy-and-paste prompts. You build your own packs here.
- You are building retrieval systems or agent frameworks in code. The course covers the thinking, not the engineering stack.
- You use AI a few times a month. The method pays off on tasks you repeat.
Before you join
- Is context engineering just a new buzzword for prompting?
- Partly, and the course says so. The prompt is the few lines you type; the context is everything the model sees, including your files, examples and rules. The words matter less than the habit: you fix the inputs before you fix the phrasing.
- Will this still work when the models change next month?
- A context pack describes your task, not a model. Clear instructions, good examples and the right source material help every model you have used. The course has you retest each pack when you switch tools, so you see the difference instead of guessing.
- I am not a developer. Is this for people who code?
- No. Every exercise works in a chat window with documents you already have. Developers who use coding assistants will recognise the same idea in project instruction files, and the course shows that link.
- Who is teaching this, and has he done it?
- Jay West leads DevOps for the marketing organisation of a multi-location retailer. He wrote a copywriting doctrine that AI writing agents load before they draft: five schools of copywriting, graded by evidence, with their conflicts settled into written rules. That file is a context pack in daily use, and the course takes it apart.
- Can I not learn this from free guides?
- Many of the ideas are in free guides from the model makers, and the course points you to them. What a guide does not do is make you rebuild one of your own tasks and compare the result. That exercise is the course.
Rather have it built for you? Describe the project and the same work can be done for your team.