Opening soon
Custom GPTs, Claude Projects & AI Assistants
Define, instruct and test a custom GPT or Claude Project built for one clearly scoped job.
6 chapters · 29 lessons
Want a note when it opens?
Join the waitlist and we'll email you when the course is ready.
What you'll cover
The chapters
The plan for the full course. Lessons open when the course does.
- 01The AI assistant landscape and its parts4 lessons
- 02Defining what your assistant should do5 lessons
- 03Writing effective system instructions5 lessons
- 04Knowledge bases and managing context5 lessons
- 05Building and testing your first custom GPT5 lessons
- 06Building with Claude Projects and other platforms5 lessons
Self-check
How much do you already know?
5 quick questions. Pick an answer to see why it's right or wrong. Nothing is saved.
Question 1 of 5
What does RAG stand for?
Question 2 of 5
Which of the following is an example of a leading indicator for an AI assistant?
Question 3 of 5
What is Goodhart's Law as it applies to AI assistant metrics?
Question 4 of 5
Why is it important that iteration triggers require a metric to fall below a threshold for multiple consecutive checks rather than just once?
Question 5 of 5
A team wants to set an iteration trigger for their assistant's output accuracy metric. They currently have no baseline data. What should they do first?
In the news
Recent stories on this topic
From NewsPals, our sister newsroom.
- Apple’s Siri AI May Orchestrate ChatGPT and Claude, Not Beat Them (opens in new tab)
A leaked Apple build suggests Siri could cooperate with outside models, or let them take the wheel entirely.
- Claude Shows Its Work: What Anthropic's Public Mental Health System Prompts Teach Builders About Safe AI Design (opens in new tab)
While competitors lock their instructions in a vault, Anthropic publishes Claude's global mental health guidance , giving every builder a rare, concrete look at how to engineer bounded AI behavior in sensitive contexts.
Key terms
Words you'll meet in this course
- System instructions
- Standing text that tells an assistant its role, rules and response format; the model reads it before every user message.
- Knowledge base
- The documents you attach to an assistant so it can answer from your material rather than only from its training.
- Retrieval-augmented generation (RAG)
- Retrieving relevant passages from your documents and adding them to the prompt, so the model can answer from that material.
- Context window
- The amount of text, measured in tokens, that a model can consider at once, including instructions, attached files and its own reply. Read the explainer
- Custom GPT
- A version of ChatGPT configured with your own instructions, files and optional tools, which you can share with others.
- Guardrails
- Rules and checks that keep an assistant within its intended scope and stop it from producing harmful or off-limits output.
- Conversation design
- Planning how an assistant opens, asks clarifying questions, hands off and ends a conversation, so exchanges stay useful and clear.
Help us choose
Which course should open next?
Pick the one you'd take first. Your vote helps us decide what to finish next.