Coding with AI tools

Stage 3 of 10 · 7 lessons · ~2 weeks

Prompting for code: small tasks, clear specs

Turn work into small, checkable tasks with a written spec and a definition of done, plan before multi-file changes, reuse prompts that work and start each task with a clean context.

Start: Small, checkable tasks

Lessons· one lesson for each topic, in this order

  1. 1Small, checkable tasks
  2. 2Writing a task spec with a definition of done
  3. 3Pointing at a pattern to follow
  4. 4Plan before code
  5. 5Prompt templates for everyday tasks
  6. 6Saving prompts that work
  7. 7Fresh context for each task

Do this

  • Write a task spec with a definition of done for your next real task.
  • Split one feature into at least four small, checkable steps.
  • Use plan mode on a change that touches several files, and correct the plan before building.
  • Write your four templates: explain, bug, feature, refactor.
  • Save one prompt that has worked several times as a reusable prompt.
  • Start a fresh conversation for each unrelated task for a week and notice the difference.

Start with· follow one of these

Keep handy

Also worth your time

Before moving on: Four prompt templates of your own (explain, bug, feature, refactor), one saved as a reusable prompt in your tool, and one multi-file change delivered from a written spec and an approved plan in small, reviewed steps.

Tauseef's advice
Most bad AI output is a bad prompt in disguise: too big, too vague, no check. Spend two minutes on a spec and you will save twenty minutes of correcting. And when a conversation starts going round in circles, stop, clear it, and write the prompt you wish you had written first.

Mark this stage as not done?