The quality bar; three worked capstone shapes; live scoping.
All tracks — 4 hours, studio sprint
Capstone Sprint: Build a Personal AI-Enhanced Creative Workflow
Leave with a working draft of one real thing — a workflow, tiny tool, prompt system, process board, or prototype — that removes friction from your practice, plus the 1-page process explanation that proves your judgment shaped it.
Prepared demo artifact
Instructor backup / finished example: capstone-gallery.html. Use it if the live tool run fails, or to reveal the finished structure after critique.
01This Week at a Glance
You will make: a capstone draft + a 1-page process explanation (what it solves, AI's role, your role, time saved, disclosure).
Friction, shape, riskiest assumption — written down.
Build the spine; refine and document, with a workflow clinic between.
90-second rough shows; ethics, usefulness, craft.
The week in one line
Ship one true thing that gives you leverage without taking your voice. Thirteen weeks of techniques are worth nothing until one lives in your practice. The trap to avoid: automating your art instead of your friction.
02The Capstone Quality Bar
Your capstone passes if it answers "yes" to all four. Print these; check them at the end.
- Real friction. It solves an actual recurring problem in your practice — not a hypothetical, not someone else's.
- Documented human decisions. Selection, rejection, refinement, and judgment are visible in the process explanation.
- Honest disclosure. It states clearly what AI did and what you did.
- Demo-able to an employer. You could show it in an interview and explain every choice without hand-waving.
The rule
A capstone that's small, finished, and passes all four beats an ambitious half-built thing every time.
03The Capstone Shapes — Pick Your Lane
| Shape | What it is | Built from | Best for |
|---|---|---|---|
| Workflow / process system | A repeatable chain (brief → boards → shot list → checklist) | Weeks 2, 3, 5, 6 | Motion, animation, multi-stage pipelines |
| Prompt system / kit | A tuned, reusable prompt set for a recurring task | Weeks 2, 8, 10, 13 | UX, design, illustration, writing-heavy practices |
| Tiny coded tool | A single-file tool removing a specific friction | Week 11 | "I keep retyping/recalculating this" problems |
| Supervised automation | A spec'd agent workflow with checkpoints | Week 12 | Anyone drowning in file/admin chores |
| Process board / case study system | A documentation system that makes your process portfolio-ready | Weeks 4, 13 | Illustration, fine art, portfolio-building |
The fastest capstone
You've already drafted candidates: your Week 11 tiny-tool spec, your Week 12 workflow spec, your Week 5 motion pack, your Week 13 prompt pack. The fastest capstone is finishing one of those.
04The Build Sprint (today's structure)
- Scope ruthlessly: name the friction, pick the shape, cut scope to ~6 hours of work total.
- Riskiest assumption first: what sinks the whole capstone if it fails? Test that first, not the easy parts.
- Build the spine: get the core working end-to-end, ugly. A rough complete thing beats a polished fragment.
- Clinic: a classmate tests whether it actually saves time or improves output.
- Refine + document: fix what the clinic found; write the process explanation as you go.
- Prep the 90-second show: what it solves, the decision you're proudest of, AI vs. you.
05The Scoping Conversation
Run this with AI, then run the same scope past a human — humans catch the "still over-scoped" the model is too agreeable to push on.
The capstone advisorAct as a capstone advisor. My project idea: [your idea]. Interview me briefly, then output a one-week build plan: scope cut to what's achievable in ~6 hours, the AI-role/human-role split, the riskiest assumption to test first, and what "done" looks like. Be ruthless about scope — I'd rather finish something small than half-build something big.
06The 1-Page Process Explanation (the assessed deliverable)
This document is what makes the capstone yours. One page, these sections:
- The friction — the real recurring problem, with honest frequency.
- What I built — the workflow/tool/system, in two sentences.
- AI's role — specifically what AI did (drafted, generated, structured, automated).
- My role — the decisions, selections, rejections, and craft that are yours. The ✏️ moments.
- Proof it works — your acceptance check or time-saved estimate; ideally a clinic-partner's verdict.
- Disclosure — the honest line you'd attach if this went in a portfolio.
- What's next — the one thing you'd improve with more time.
Template starting points: workflow spec, tiny-tool spec, process documentation.
07Lab: Build Sprint + Workflow Clinic (the whole session)
- Scope (Hour 1): scoping conversation (AI + human); lock your shape and 6-hour scope; name your riskiest assumption.
- Sprint 1 (Hour 2): build the spine — riskiest assumption first, then core end-to-end. Ugly is fine.
- Workflow Clinic (~20 min): swap with a classmate; they actually try your thing. Does it save time / improve output? Would they use it? Where did it break?
- Sprint 2 (Hour 3): fix what the clinic found; write the process explanation as you go.
- Show-and-tell + peer review (Hour 4): 90-second rough shows; critique on ethics, usefulness, craft — not polish.
Zero-AI / planning capstone (fully equal)
A fully-specified workflow or tool design — the complete spec + process explanation + a paper walkthrough — is a legitimate capstone. Several of the strongest capstones are plans for things students will build later.
Came in cold today?
Pick the smallest shape (a prompt kit or tiny-tool spec), use a standing friction, and aim for "finished and honest" over "ambitious." A complete small capstone is a real capstone.
08Integrity Check
This week's four
- Does your capstone automate your friction or your art? If it's making your finished work, re-scope.
- Are your human decisions documented, not just claimed? Point to the ✏️ moments.
- Is your disclosure honest and specific enough to stand in an interview?
- Did a real human (clinic partner) confirm it does what you say? Untested "it saves time" is a hypothesis.
09Troubleshooting
| Problem | Fix |
|---|---|
| My idea is too big for today | Good problem. Cut to the spine: smallest version that solves the core friction. The rest is homework |
| I don't know which shape to pick | Pick the one closest to a draft you already have. Finishing beats starting |
| It works but feels unimpressive | "Impressive" isn't the bar — useful, finished, honest is. The two hours a week it saves IS the impressive part |
| The clinic partner couldn't use it | The most valuable result of the day. Fix the thing they tripped on — that's real product feedback |
| I'm building something that makes my art | Stop and re-scope. Aim the leverage at the friction around your art, not the art itself |
| I can't finish in the time | You don't have to — today's deliverable is a draft + process explanation. Document where you got to and what's next |
10Resources
- TemplateLinkedAI Workflow Spec Template (this course)The scaffold for workflow/automation capstones — steps, checkpoints, risks, disclosure.
- TemplateLinkedTiny Tool Spec Template (this course)For coded-tool capstones — user, problem, inputs, outputs, acceptance checks.
- TemplateLinkedProcess Documentation Template (this course)The backbone of your 1-page process explanation.
- ReferenceLinkedDemos & Visual Aids — Week 14 (this course)The three worked capstone shapes, in detail, with the scoping prompt.
- GuideLinkedBuilding Effective AI Agents — AnthropicIf your capstone is an automation: the simplest-pattern-that-works principle keeps your scope honest.
- RevisitLinkedYour own Weeks 5, 8, 11, 12, 13 deliverablesYour fastest capstone is finishing one of these. Re-read your own specs before starting.
Optional extension
- Build the second-smallest version: once the spine works, add the one feature your clinic partner most wanted — and only that one.
- Write the capstone's own case study (Week 13 method) — the extraction interview on your own build.
- Hard mode: design how your capstone fits a team's workflow, not just yours.
Next week: Demo Day & the Artist's AI Operating System → — present what you built, revise your Manifesto to v2, and build your system for staying current without drowning.