UofAi methodology
DAP is UofAi's AI-native method for teaching the skill that matters most now: taking real work, decomposing it, collaborating with AI through iteration and validation, and proving the result with portfolio evidence.
6
method components from diagnosis to credential
5
augmentation ladder rungs learners climb through practice
1
verified portfolio of real-world capability
What it is
DAP upgrades the existing UofAi learn → apply → ship proof-of-skill loop. Instead of teaching prompt tricks, each lesson and lab trains the transferable behaviors that experienced AI users display: task framing, critical feedback, multi-turn iteration, validation, and learning without dependence.
Learners decompose their real role into tasks, then identify where AI creates deskilling risk versus upskilling opportunity. The output is a personal Leverage Map: what to stop competing on, what to augment first, and where human judgment matters most.
Lessons and labs are sequenced through five collaboration modes: directive, feedback loop, task iteration, validation, and learning. The goal is to move learners beyond prompt fluency into high-success augmentation behavior.
Every rep follows the same loop: real task, learner attempt on a specific rung, coach feedback on the collaboration process, one targeted correction, and an immediate re-rep at higher difficulty.
Progress is measured with evidence: augmentation share, task value attempted, success rate, and critical-validation rate. Learners see their Augmentation Curve bend upward through demonstrated work, not vanity completion.
Certification is earned through a portfolio of real artifacts, longitudinal ledger evidence, and human-in-the-loop verification. It says what a learner demonstrably did, not what they watched.
The method teaches durable judgment: task decomposition, model selection, validation discipline, and recalibration when frontier models change. Recipes decay; augmentation capability transfers.
Why it is right for learning AI
Frontier-lab research points to a widening learning curve: experienced users attempt higher-value tasks, iterate more effectively, and get better results. DAP makes that accidental learning curve deliberate, measurable, and accessible before displacement risk becomes personal.
The augmentation ladder
1
Frame a task clearly enough that an AI hand-off can succeed.
Corrects: Under-specified prompts that produce generic output.
2
Read output critically and give targeted corrections.
Corrects: Accepting the first answer as finished work.
3
Co-develop across turns and steer the model toward intent.
Corrects: One-shot prompting instead of collaboration.
4
Stress-test claims, verify sources, and catch confident errors.
Corrects: Over-trust, hallucination risk, and weak provenance.
5
Use AI to extend your own expertise instead of replacing it.
Corrects: Dependence, deskilling, and shallow fluency.
How UofAi uses it
Each lesson now names the DAP rung being trained and asks the learner to produce a small process artifact: what they asked, how they corrected, what they validated, and what judgment transferred back to them.
Each lab now functions as a deliberate-practice loop: complete a real task, document AI collaboration, verify the result, reflect on the next rung, and submit proof that can compound into a Verified Capability Credential.
Build your first proof-of-skill artifact