1. Measure the run, not the demo
Lesson · 40 minutes
Efficiency is the time, the retries, and the checking on real items. A single impressive answer is not a measure.
Course 03 · Advanced
Mastering AI Performance Optimization: Algorithms for Efficiency
Make an AI-assisted process cheaper to run: smaller inputs, fewer retries, and a clear stop when the extra pass is not worth it.
Who it is for. People who already use AI in a workflow and want it to use less time, context, and rework.
What you will be able to do
- See where time and retries are actually spent.
- Shrink the input and the output before you reach for a larger model.
- Stop a loop when another pass will not change the decision.
Lessons
2. Shrink the input
Lesson · 45 minutes
Long context is not free. Paste the paragraphs that the task needs, not the whole archive. Name the fields you care about and leave the rest out.
3. Ask for a smaller output
Lesson · 40 minutes
Specify the shape: a table, five lines, or a yes-or-no with the reason in one sentence. An open request spends the run on prose you will delete.
4. Stop the loop
Lesson · 35 minutes
A second pass is justified when the first failed a named check: a missing source, a wrong name, a format the next step cannot read. “Make it better” is not a check.
5. Choose the smallest process that holds
Lesson · 40 minutes
Use the simplest setup that meets the check. A larger or slower setup is worth it only when the smaller one fails the same check on the same items.
6. Check your understanding
Quiz · 6 minutes