Build your AI Agent skills, one step at a time
Start with a small task, prepare its context, then learn to check and resume the work. Follow the linked Cue guides for the full exercises and examples.
This learning route is free to read. Cue features and outside AI services follow their own plans and limits. plans and limits ↗
The learning path
Optional background videos. Selection uses catalogue metadata and summaries; videos were not reviewed in full, and regional playback is unverified.
-
Stage 1: Start with a result you can check
Choose one small task and define an answer you can verify.
Check: Can you compare the answer with the facts you supplied?
Try your first Agent task ↗Optional background video[1hr Talk] Intro to Large Language Models · Andrej Karpathy
Background on language models; catalogue metadata and summary reviewed only.
Watch on CueTube -
Stage 2: Give the Agent reliable context
Prepare the source material, limits and unresolved questions before handing over a task.
Check: Can another person trace a key fact to its source?
Build a recording-to-context packet ↗Optional background videoAttention in transformers, step-by-step | Deep Learning Chapter 6 · 3Blue1Brown
Optional explanation of attention; it is not a tutorial for the linked workflow. -
Stage 3: Resolve a contradiction before acting
Spot where two notes disagree and identify the decision that still needs confirmation.
Check: Does the handoff distinguish a reported change from an approved action?
Practice resolving conflicting notes ↗ -
Stage 4: Resume, then verify the remaining work
Use a recovery brief to continue from the actual state of a task.
Check: Can you distinguish work that changed, tests that ran and results that were published?
Practice resuming an interrupted Agent ↗Optional background video🚀 TDD, Where Did It All Go Wrong (Ian Cooper) · DevTernity Conference
Optional background on testing code; metadata and summary reviewed only.