What you will be able to do
- Predict, roughly, whether a task is easy or hard for a model before you try it.
- Turn a vague ask into a brief the model can actually execute.
- Give the model the context it needs instead of hoping it guesses.
- Tell the difference between a model that is wrong and a prompt that was underspecified.
- Set up a base operating prompt and a working memory you reuse every day.
Chapters
- The prediction machine07:00
- Context is the whole game12:00
- Prompting as thinking10:00
- Where it breaks08:00
- Your base setup09:00
What you build
Build your base operating prompt and a working memory file.
Harness module
Core reasoning