3 min
1. Framing
Title, abstract, introduction, and conclusion: question, source type, and claimed contribution.
Reading method
Ten minutes to avoid reading everything
The first pass asks whether a paper deserves more time. The lab adds evidence, critique, and an engineering decision.
10 min
10-minute triage
4 min
2. Evidence scan
Figures, method, and key results: evidence type and the strongest claim.
3 min
3. Verdict
Read deeper, watch, or skip—with one testable reason.
30 min
30-minute lab
Context
What engineering decision could this paper change? What do you expect going in?
First pass
Research question, paper category, and claimed contribution.
Evidence
Method, sample or dataset, metrics, key figure/table, and the claim → evidence link.
Critique
Validity threats, alternative explanations, transferability, and contradictions.
Memo
Decision, evidence, confidence, limitations, applicability, and next step—in 250 words.
Why the reference stays hidden until you try
Recall and committing to your own version require effort. The expert answer is therefore not a starting hint: reveal it after answering or explicitly choosing “I don't know.” A rubric then checks the shape of your reasoning, without an automated score. This is not a lock: the site is static, the reference texts reach your browser with the page, and finding them in the source is easy. The order holds because you keep it, not because anything prevents you.
The timer is a guide, not a gatekeeper
It pauses, resumes, and counts active time. It never locks the original, advances a stage, or penalizes a longer read.
What the protocol builds on
The structure adapts three-pass reading and engineering questions about claims, method, evidence, and limitations. Calibration uses active retrieval instead of passive rereading.