The PAI-L Mentorship Framework: From Idea to Civic App

Phase 1: Problem Formulation & The "NotebookLM Sandbox"

• The Concept: Before building, students need to deeply understand the policy space without getting bogged down or misled by hallucinations.
• Mentorship Strategy: Have students gather primary source documents (city council minutes, historical policies, zoning laws) and upload them into NotebookLM. Teach them how to use NotebookLM to generate study guides, audio overviews, and query the text to isolate the human problems within a policy. The AI acts as an interactive literature review, letting the student focus entirely on problem formulation and finding gaps in civic inclusion.

Phase 2: Defining the Target with Google Deep Research

• The Concept: AI tools need quantifiable benchmarks to optimize against, but finding robust baseline data is difficult.
• Mentorship Strategy: Teach students to use Google Deep Research to scour the web for existing quantitative benchmarks, datasets, and civic precedents. Instead of spending weeks searching manually, Deep Research can synthesize the current state of a public problem (e.g., "What are the measurable factors in rural broadband rollout?") in minutes, giving the student a solid foundation to build their app's scoring rubric upon.

Phase 3: Rapid Prototyping with Antigravity

• The Concept: The barrier between idea and working tool has collapsed. Students don't need to learn Python; they need to learn how to direct an agentic coding assistant.
• Mentorship Strategy: Treat the student as the "Research Director" and Antigravity (or a comparable coding agent) as their Lead Developer. Mentor them on how to write clear, structured prompts. Instruct them to have the agent build interactive web apps, data visualization dashboards, or survey pipelines directly in their local environment. They iterate in real time—if a feature breaks, they simply ask the agent to fix the stack trace.

Phase 4: Stress-Testing & Auditing for Rigor using Gemini 1.5 Pro

• The Concept: Speed kills rigor. Fast prototyping can lead to fragile or biased civic tools if left unchecked.
• Mentorship Strategy: Teach students how to audit their own models. They can build rigorous evaluation pipelines where Gemini 1.5 Pro acts as a "Red Team." Students should create AI simulations that stress-test whether their civic tool or proposed policy can withstand intense scrutiny from diverse, competing stakeholders. Emphasize that the student’s job is to ensure the app complicates the narrative fairly, rather than indiscriminately optimizing for any single outcome or viewpoint.

Phase 5: Deploying Living, Forkable Civic Infrastructure

• The Concept: The end product is an open, dynamic tool that the community can use, replacing the static academic paper.
• Mentorship Strategy: Coach students to use Antigravity to seamlessly push their code to GitHub, or formalize their structured prompts into public Gemini Gems. For robust web applications, they can deploy using Google Cloud or Firebase. The ultimate goal is for their civic tool to be public, machine-verifiable, and ready for other researchers or community organizers to fork, debate, and adapt.