Intent framing
State what role the response should perform: explain, compare, debug, plan, critique or evaluate.
Prompt Engineering Lab
Control the question before blaming the answer.
CortexLab treats prompting as an interface design problem. Specify intent, context, audience, depth, evidence and uncertainty expectations so the engine can choose the correct route.
Open Cortex workspaceContext boundary
Tell Cortex what source, project, article or domain should constrain the answer.
Depth control
Choose quick, standard, deep or research-level explanation.
Audience model
Beginner, student, engineer, stakeholder or interview answer.
Output contract
Paragraph, checklist, architecture, table, study plan, code walkthrough or critique.
Uncertainty rule
Require confidence and a teaching request when evidence is weak.
Counterexample request
Ask for failure cases, trade-offs and when not to use the method.
Evaluation request
Ask what metrics, baselines and tests would prove the proposed approach works.
Paraphrase depth
Control wording variation while preserving the grounded meaning.
Naturalness profile
Choose polished, conversational or relaxed phrasing without making false authorship claims.
Prompt patterns
Reusable structures for real engineering work.
teach me reinforcement learning from zero. start with intuition then math then an example then test me
Try in Cortex →debug this system design. list assumptions first then isolate the highest risk failure and propose tests
Try in Cortex →evaluate this idea against a simpler baseline. show evidence needed and reasons the idea could fail
Try in Cortex →answer like an ml engineer interview. definition then why it matters then example then trade off then validation
Try in Cortex →Naturalness and paraphrase controls change wording style. They do not prove human authorship and are not presented as a detector-bypass feature.