Summer Fellowship / Curriculum

THE SIX-WEEK ARC

Two tracks run in parallel — Agentic AI and AP Statistics — and converge at joint events. Every week ends with a named, inspectable artifact, and each artifact compounds into a single project presented at Demo Day. No filler weeks. No throwaway work.

Week 1 / 6

Ship Something Real

Joint Kickoff · Tuesday

Both tracks open together: program vision, Demo Day preview, cohort norms, and accounts setup — then each track breaks out for its first build session.

Track 1 — Agentic AI

Zero to Shipped

Onboarding + first ship: every fellow deploys a live AI-powered web app by Friday.

You Will Learn To
Set up the full stack: GitHub, Next.js, Vercel, Claude API key
Make first Claude API calls; understand tokens, system prompts, temperature
Apply core prompt-engineering patterns (role, structure, examples, output formatting)
Ship: build → commit → deploy → share a public URL
Write a clear README and a one-sentence product description
Ships This Week

Live deployed app wrapping the Claude API (e.g., personalized study-buddy, recipe remixer, college-essay brainstormer) with a public URL + GitHub repo.

Tuesday Anchor

Track breakout after the joint kickoff: live-build a Claude-powered app from blank repo to deployed URL in 60 minutes.

Thursday Anchor

Prompt engineering dojo: structured drills, then fellows fork and extend the demo app; deploy clinic.

Async Work

Pre-read Anthropic's prompt-engineering docs. Finish and deploy the app; record a 60-second walkthrough; Friday standup post.

Checkpoint

Binary: live URL exists and works by Friday. Mentor rubric on README clarity (1–4).

Track 2 — AP Statistics

Data Has a Shape

Data intuition + descriptive statistics through real, messy datasets.

You Will Learn To
Distinguish categorical vs. quantitative variables; formulate investigative questions
Describe distributions: shape, center, spread, outliers (SOCS)
Build and interpret histograms, boxplots, and summary stats on a real dataset
Use AI tools to clean and explore data — and verify the AI's output by hand
Critique a misleading real-world graph
Ships This Week

“Data Autopsy” mini-report: pick one real dataset (open data, sports, Spotify), produce 3 visualizations + a 1-page written analysis with an investigative question.

Tuesday Anchor

Track breakout after the joint kickoff: “Lying with statistics” — dissect 5 viral misleading charts; intro to SOCS with a live class dataset collected from fellows in the room.

Thursday Anchor

Data exploration lab: messy real dataset, hands-on cleaning and visualization; Tracy Wallace models the “formulate questions” practice.

Async Work

Pre-read the chosen dataset's data dictionary. Finish the Data Autopsy; practice problems on the unit; Friday standup post.

Checkpoint

Rubric on the report (question quality, correct vocabulary, honest interpretation); 5-question concept check Thursday.

Week 2 / 6

Under the Hood

Track 1 — Agentic AI

What Makes an Agent an Agent

Agent fundamentals: the loop (perceive → reason → act), structured outputs, memory, evaluation.

You Will Learn To
Explain the agent loop and how it differs from a single LLM call
Implement structured outputs (JSON mode / tool schemas) reliably
Add conversation memory and basic state to an app (Supabase)
Write simple evals: define success criteria and test prompts systematically
Identify failure modes: hallucination, prompt injection, infinite loops
Ships This Week

v2 of the app, now agentic: persists user data in Supabase, uses structured outputs, and includes a written eval doc with 10+ test cases.

Tuesday Anchor

Live-code an agent loop from scratch (no framework) so the magic is demystified; Supabase setup lab.

Thursday Anchor

Eval workshop: fellows swap apps and try to break each other's agents; failure-mode taxonomy on the whiteboard.

Async Work

Integrate Supabase; write evals; pre-read Anthropic's tool-use docs for Week 3. Friday standup.

Checkpoint

Eval doc reviewed by a mentor; the app must pass its own evals live on Thursday.

Track 2 — AP Statistics

Randomness Isn't What You Think

Probability and randomness via simulation — casinos, sports streaks, and why intuition fails.

You Will Learn To
Estimate probabilities via simulation before formulas
Apply rules of probability, conditional probability, and independence
Interpret random variables and expected value through games of chance
Run Monte Carlo simulations in Python (with AI assistance) and by hand
Explain the Law of Large Numbers with evidence they generated themselves
Ships This Week

“Beat the House” project: design a casino-style game, compute its theoretical expected value, simulate 10,000 plays, and write up theory vs. simulation.

Tuesday Anchor

Probability lab with dice and cards → the same experiments as code; the hot-hand fallacy with real NBA data.

Thursday Anchor

Game-design studio: build and pitch casino games; peer play-testing generates the simulation data.

Async Work

Run the 10,000-trial simulation; write the theory-vs-simulation memo; AP practice set on probability. Friday standup.

Checkpoint

Project rubric + a 10-minute one-on-one with Tracy: explain expected value without notes.

Week 3 / 6

Connecting to the Real World

Joint Cross-Track Lab · Tuesday

“Agents Meet Data”: Stats fellows bring datasets and questions; AI fellows' agents analyze them live. AI fellows learn why evals need statistics; stats fellows see their toolkit applied to cutting-edge problems. Mixed pairs, one shared output each.

Track 1 — Agentic AI

Give Your Agent Hands

Tool use and MCP: agents that read and write real data and call real services.

You Will Learn To
Define tools with JSON schemas and handle tool-call rounds in code
Explain what MCP is and why standardized tool protocols matter
Build a working MCP server exposing 2–3 tools (e.g., to a database or public API)
Practice defensive design: validation, permissions, safe failure
Scope the Demo Day capstone (1-page spec, mentor-approved)
Ships This Week

Working MCP server (repo + demo video) connected to the app — the agent now takes real actions. Capstone spec approved.

Tuesday Anchor

Joint Cross-Track Lab — both tracks together.

Thursday Anchor

MCP build lab: from schema to working server; security discussion (what should an agent never be allowed to do?); capstone spec reviews.

Async Work

Finish the MCP server + demo video; draft the capstone spec before Thursday. Friday standup.

Checkpoint

MCP server demo (works or doesn't); spec quality gate — no fellow enters Week 4 without an approved capstone.

Track 2 — AP Statistics

The Shape of Chance

Distributions and sampling: normal and binomial models, sampling distributions, the Central Limit Theorem.

You Will Learn To
Use the normal and binomial distributions as models — and check when models fit
Construct sampling distributions empirically via simulation
Explain the Central Limit Theorem from their own simulated evidence
Understand bias vs. variability in sampling; design a simple survey or experiment
Launch real data collection for the final project
Ships This Week

CLT evidence portfolio: simulated sampling distributions at n = 5 / 30 / 100 with written interpretation. Final-project proposal approved (question + data-collection plan).

Tuesday Anchor

Joint Cross-Track Lab — both tracks together.

Thursday Anchor

Sampling Olympics: candy and penny sampling activities → code simulations; the CLT reveal; Tracy holds final-project proposal conferences.

Async Work

Run CLT simulations; submit the project proposal; AP practice set on distributions. Friday standup.

Checkpoint

Proposal conference with Tracy (go/no-go gate); concept check on sampling distributions.

Week 4 / 6

Orchestration & Inference

Track 1 — Agentic AI

Many Hands

Multi-agent workflows: orchestrators, subagents, routing, and when not to use multiple agents.

You Will Learn To
Design orchestrator–worker patterns; decompose a task across agents
Implement agent-to-agent handoffs with shared state
Compare single-agent vs. multi-agent cost, latency, and quality tradeoffs with data
Apply Week 2 eval discipline to a multi-agent pipeline
Make real progress on the capstone (core loop working by Friday)
Ships This Week

Multi-agent pipeline demo inside the capstone (e.g., researcher → writer → critic) + a 1-page tradeoff analysis with measured numbers.

Tuesday Anchor

Architecture studio: whiteboard capstone pipelines as a group; live-build an orchestrator; cost and latency measurement lab.

Thursday Anchor

Capstone work session with mentor rotations; mid-program retro: each fellow demos current state in 2 minutes.

Async Work

Capstone core loop complete; tradeoff write-up. Friday standup.

Checkpoint

Mid-program demo (2 minutes, recorded) — the “is this on track for Demo Day?” gate.

Track 2 — AP Statistics

How Sure Are We?

Statistical inference: confidence intervals and hypothesis tests for proportions and means.

You Will Learn To
Construct and interpret confidence intervals (proportions and means)
Frame null and alternative hypotheses; explain p-values correctly (and what they are not)
Connect simulation-based inference to formal z- and t-procedures
Identify Type I/II errors and their real-world costs
Apply inference to their own collected project data
Ships This Week

First formal inference on their own data: a correct confidence interval + hypothesis test on the final-project dataset, with plain-English interpretation.

Tuesday Anchor

“Is the coin fair?” — full inference logic via simulation first, formulas second; p-value misconceptions clinic.

Thursday Anchor

Inference lab on fellows' own project data; error-types debate (FDA drug approval scenario); Tracy 1:1 rotations.

Async Work

Complete the inference write-up; timed AP practice FRQ (12 minutes); Friday standup.

Checkpoint

Timed FRQ scored with the real AP rubric; inference write-up feedback from Tracy.

Week 5 / 6

Make It Real

Joint Pitch Workshop · Thursday

Storytelling for technical work — 3-slide structure, demo choreography, handling Q&A. Fellows from each track present 90-second drafts to mixed-track peers for critique.

Track 1 — Agentic AI

From Project to Product

Real-world deployment: auth, persistence, error handling, real users, observability.

You Will Learn To
Add auth and per-user data (Supabase Auth + RLS)
Handle errors and edge cases gracefully; add basic logging and observability
Recruit 5+ real users; collect structured feedback
Iterate based on user evidence, not vibes
Draft the Demo Day narrative: problem → demo → what's next
Ships This Week

Production-ready capstone with auth, 5+ real user sessions logged, and a user-feedback synthesis doc.

Tuesday Anchor

Hardening lab: auth, RLS, error states; “demo your error messages” exercise; user-recruitment planning.

Thursday Anchor

Joint Pitch Workshop — both tracks together.

Async Work

User testing sprint (5+ users); iterate; draft the pitch. Friday standup.

Checkpoint

User-feedback doc reviewed; pitch draft v1 delivered at the joint workshop.

Track 2 — AP Statistics

Predicting the Future

Regression and prediction: linear models, residuals, and the line between correlation and causation.

You Will Learn To
Fit and interpret least-squares regression (slope, intercept, r, r², residuals)
Diagnose model fit with residual plots
Articulate correlation ≠ causation with study-design reasoning
Connect regression to ML: what is a model, loss, and prediction?
Complete all analysis for the final project
Ships This Week

Complete final-project analysis including a regression component + draft slides.

Tuesday Anchor

Regression lab on real paired data (e.g., screen time vs. sleep from the cohort survey); residual plot diagnostics.

Thursday Anchor

Joint Pitch Workshop — both tracks together.

Async Work

Finish analysis + slides; AP practice set on regression. Friday standup.

Checkpoint

Analysis QA checklist with Tracy; pitch draft v1 delivered at the joint workshop.

Week 6 / 6

Demo Day

Joint Demo Day · Thursday evening

All fellows present 3–4 minutes plus Q&A to families, mentors, and invited community guests on one shared stage — certificates, awards, and cohort celebration.

Track 1 — Agentic AI

Ship It, Show It

Polish, rehearse, present — and package the work for college applications.

You Will Learn To
Cut ruthlessly: a flawless 3-minute demo beats a buggy 6-minute one
Deliver a rehearsed live demo with a backup recording
Write the portfolio artifact: project page, README, demo video
Articulate skills gained in college-application language
Define a post-program roadmap (keep building, users, next features)
Ships This Week

Demo Day presentation + a public portfolio page (live URL, repo, video, write-up).

Tuesday Anchor

Dress rehearsal with full timing; bug-bash triage; portfolio-page workshop.

Thursday Anchor

Joint Demo Day — both tracks, one shared stage.

Async Work

Rehearse; finalize the portfolio; submit all artifacts by Wednesday evening.

Checkpoint

Demo Day rubric (working demo, clarity, Q&A) + complete portfolio = program completion.

Track 2 — AP Statistics

Defend Your Findings

Final project completion and public presentation of a defensible statistical study.

You Will Learn To
Present a complete statistical study: question → data → analysis → honest conclusions
Defend methodology under live Q&A (limitations, biases, what they'd do differently)
Produce a written report at AP-project standard
Map the project to AP exam units — concrete evidence of readiness
Plan the AP year: what's covered, what remains, study plan with Tracy
Ships This Week

Demo Day presentation + final written report (3–5 pages) + an AP-readiness self-assessment.

Tuesday Anchor

Dress rehearsal; methodology Q&A gauntlet (mentors play hostile reviewers); report polish.

Thursday Anchor

Joint Demo Day — both tracks, one shared stage.

Async Work

Rehearse; finalize the report; submit all artifacts by Wednesday evening.

Checkpoint

Demo Day rubric + report rubric + AP-readiness checklist signed off by Tracy.

Six Weeks. One Arc.

Every week ships. Every artifact compounds.

Submit ApplicationBack to the Fellowship