# Creator Workbench

**A two-day architecture lab.** Crisis Cat brings the product vision. Wicked Spooky brings the cloud design. Together, they turn an ambitious one-person social media team into a buildable, measurable AWS specification.

## The cast and the commission

**Crisis Cat — product designer and client.** Crisis Cat has designed the Creator Workbench: research, content, publishing, community, attribution, and a daily brief. The business starts with almost no money but expects thousands of subscribers if its marketing takes off.

**Wicked Spooky — the solution architect.** Wicked Spooky IT, also known as Second Act Sam, brings 30 years of IT experience and is studying AWS SAA-C03. In this scenario, Wicked Spooky takes responsibility for specifying the AWS environment Crisis Cat needs. “SAA” describes the architect's role and learning objective; it does not assert that the exam has already been passed.

WickedSpooky.ai hosts this training section inside its **existing Amplify application**. WickedSpookyIT.com is the delivery identity. The eventual `creator.wickedspooky.ai` product is a separate application outcome. These pages discuss and design that outcome; they do not operate it or provision its infrastructure.

## Two days, one mission

| Lesson | The question | The design artifact |
| --- | --- | --- |
| [Day 1 — The Creator Workbench](day-1.md) | What must the product do, and what is the simplest useful architecture? | Module specifications, human approval rules, local AI strategy, and Mermaid user journey |
| [Day 2 — Wicked Spooky's architecture review](day-2.md) | How do we deliver safely on today's budget and survive tomorrow's growth? | AWS feature review, alternatives, subscription guardrails, SAA mapping, and fail-forward plan |

## Our product design method

We apply the method already defined on [iLab](https://ilab.wickedspooky.ai/): **Discover → Plan → Execute → Certify → Impact**. Governed Discovery clarifies intent and evidence. AI-Augmented Insight helps us examine choices. Certified Execution checks work against agreed requirements. Continuous Learning records failures and repairs. Trusted Impact connects measured outcomes to the original goal.

| Stage | What it means in this lab | Evidence we will retain |
| --- | --- | --- |
| Discover | Translate Crisis Cat's pitch into requirements, constraints, unknowns, and authority | Product brief and prerequisite ledger |
| Plan | Compare architectures, assign ownership, and select the smallest credible design | Diagrams, decisions, cost model and acceptance criteria |
| Execute | Build bounded experiments when implementation labs begin | Versioned changes and reproducible test inputs |
| Certify | Verify approvals, tenant isolation, failure behavior, recovery and unit economics | Actual results, traces, cost measurements and restore records |
| Impact | Compare revenue, cost and useful outcomes; feed lessons into the next iteration | Daily brief, margin trend and fail-forward findings |

Method references: [Discover](https://ilab.wickedspooky.ai/discover.html), [Plan](https://ilab.wickedspooky.ai/plan.html), [Execute](https://ilab.wickedspooky.ai/execute.html), [Certify](https://ilab.wickedspooky.ai/certify.html), [Impact](https://ilab.wickedspooky.ai/impact.html). Here, Certify means internal acceptance; it does not mean third-party accreditation.

## How to read the case study

Architecture choices are narrated as completed **design decisions** so this can guide a build from scratch. Resources, integrations, incidents and performance results are not claimed complete. Crisis Cat's assumption that infrastructure and integrations already exist is a scenario input; Day 2 turns it into a readiness checklist to verify before delivery.

The [operations design case study](https://github.com/wickedspookyit-spec/wickedspooky.org/blob/main/public/docs/creator-workbench-case-study.md) holds the detailed cost model and all 14 SAA-C03 task statements. This training companion adds the client/architect story, the two-day discussion, and local-focused AI routing. It retains the human approval gate, headless event architecture, profitability floor, and evidence-first approach.

## The promise we are designing for

One person can prepare a week of content and follow its commercial results without becoming a cloud operator. Nothing is published without approval of the exact version. Routine work uses deterministic processing or a small model where useful. Accepted jobs have visible outcomes. Growth changes throughput and paid usage within tested limits; it does not remove approvals, platform rules, or cost controls.

Start with [Day 1 — The Creator Workbench](day-1.md).