Research and exploration pathways▦

1. Read two pages on the same topic, one authored and one generated.
2. Identify three verifiable claims and their sources.
3. Locate date, author and status metadata.
4. Use a lab without confusing simulation with live AI.
5. Write a 200-word summary stating uncertainties.

Expected artifact : A reading worksheet comparing facts, interpretations and unsupported claims.

Success criterion : Beginners can identify primary evidence, explain agent roles and avoid unsupported conclusions.

1. Choose course-aligned sources and tasks.
2. Compare two contradictory generated answers.
3. Require justification using original evidence.
4. Explain generation limitations and media provenance.
5. Assess evidence-led reasoning with a transparent rubric.

Expected artifact : A lesson sequence, assessment rubric and source pack for learners.

Success criterion : Learners explain whether a citation supports a claim and distinguish synthesis from evidence.

1. Define genuine needs and user personas.
2. Choose HTML structure and UI without mandatory AI dependence.
3. Prototype with sourced content.
4. Test contrast, keyboard, performance and compatibility.
5. Document licenses, attribution and revisions.

Expected artifact : An accessible prototype with sources, fallback and a testing report.

Success criterion : Visitors reach the information without a model and generated content remains clearly identified.

1. Create a side-effect-free retrieval tool.
2. Document its arguments using JSON Schema.
3. Compare the roles of MCP, A2A and WebMCP.
4. Implement confirmation before consequential operations.
5. Test untrusted content and inspect traces.

Expected artifact : A documented tool schema, permission diagram and reproducible execution checks.

Success criterion : The tool rejects unauthorized operations and reports observed outcomes without claiming nonexistent connections.

1. Define research questions, hypotheses and metrics.
2. Build a dataset with known rights.
3. Include negative and contradictory cases.
4. Measure components, uncertainty and failures.
5. Publish methods, versions and limitations.

Expected artifact : A dated evaluation protocol, detailed results and failure analysis.

Success criterion : Results are reproducible with limitations rather than broad superiority claims.

1. Identify use cases and accountable roles.
2. Map data and tool permissions.
3. Evaluate privacy, security and accessibility.
4. Define metrics and approval gates.
5. Set up monitoring and correction processes.

Expected artifact : A risk register, permission matrix and review process.

Success criterion : The team can explain decisions, suspend automation and correct failures.