Lesson Plans

Format, all sessions (75–90 min): 15 min case story (Chris) → 15 min live demo including one deliberate failure-and-correction → 40 min hands-on (Aviva leads) → 10 min debrief + assignment.

Workshop 1 — AI as Your Analyst document systems

Case: a 1999-era investment memo template transformed into an institutional design system. Tokens (colors, type, tables) decided once in writing; AI applied them across documents; a validation loop — render, compare, actually read the output — because the first pass is never right.

Teach

1. Design tokens: decide the standard once, before generating anything.
2. The iterate-verify loop: prompt → output → inspect like a skeptical reviewer → correct → repeat. The skill is the inspection, not the prompt.
3. Division of labor: AI owns mechanics (format, structure, consistency); humans own judgment (what the memo argues, what the numbers mean).

Hands-on (Aviva)

Pairs get the deliberately ugly, fully fictional Seminole Crossing memo + the club token sheet. Restyle it with AI; pairs swap and QA each other's output against the tokens.

Artifact shipped: the club's one-page memo template — used for every deal pitch and coverage beat this year.
Assignment: restyle one document from your own life with AI; two minutes each next week on what AI got wrong and how you caught it.

Workshop 2 — AI as Your Research Desk 37 market studies

Case: a market research desk producing dozens of metro supply/demand studies through an AI pipeline. The lesson is the gates, not the speed: work decomposed into tickets, review checkpoints on every output, data quirks hunted (mis-scaled fields, survivorship bias), and the cardinal rule learned the hard way — never trust an AI claim you didn't verify yourself.

Teach

1. Decomposition: one giant "research the market" prompt fails; twenty small verifiable tickets succeed.
2. Review gates: define "done" before AI starts; check against it after.
3. Data skepticism: AI will confidently repeat a broken number — find the quirk in the source.
4. Provenance: every claim traces to a source you can name.

Hands-on (Aviva)

Each student + AI drafts a one-page market study for their coverage beat (assigned in the kickoff 1:1s): supply pipeline, rent trend, one demand signal, one risk — every number sourced. Live gate: pairs audit each other for one unverifiable claim. There is always one.

Artifact shipped: 9 draft coverage-beat one-pagers → after Chris's review gate, the website's first Coverage Universe content.
Assignment: fix gated claims; final due here in one week.

Workshop 3 — Build Night Aviva leads

Case (10 min, Chris): how this very site was stood up — tokens first, static pages, Cloudflare Pages + Access. The site they're using is the case study.

Hands-on — four teams, four tickets

1. Home + who we are — mission, the scoreboard, the analyst-class story
2. Coverage Universe — frame for the 9 beats; Workshop 2 output slots straight in
3. The Record — published work, Cohort Letter archive, placements
4. People + contact — 9 profiles, the professionals' bench

Each team drafts copy + layout intent with AI, then presents to a live IC-style gate: the room votes ship / revise.

Artifact shipped: v0 content for the entire public site in the Site Lab + a decision list for the real build.
Rolling assignment: best contributor per section keeps ownership of that page through launch.

Pre-session agenda template (Chris + Aviva, 30 min before)

Recaps

Aviva posts each session recap here within 48 hours: what shipped, what broke, attendance, and the one change for next session.