Stakeholder simulation service for the Epic Systems Engineering Judgment Workshop

---

pea stands for Product Engineer Agents.

It is the agent-side service for the Epic Systems Engineering Judgment Workshop. The separate workshop app owns the learner experience; pea provides the stakeholder simulations that learners talk to while they clarify ambiguous requirements, surface constraints, and practice engineering judgment.

The workshop thesis is simple: as AI takes on more implementation work, the scarce engineering skill becomes judgment. pea exists to help train that skill.

What Pea Does

pea simulates messy real-world stakeholders in structured exercises. These agents can act like:

• product managers
• executives
• support and operations teams
• users
• regulators
• other engineers

The point is not to help participants write code. The point is to create the kind of ambiguity, friction, and conflicting incentives that force participants to ask better questions before anything gets built.

What Pea Is Not

pea is not:

• a prompt engineering workshop
• an AI coding tutor
• a framework training project
• the primary workshop application

The learner-facing workshop app already exists. pea is the service that app uses to run stakeholder conversations, and it will also provide instructor/admin controls for managing agents, scenarios, and prompts.

Quick Start

bun install
bun run dev

See docs/getting-started.md for local setup and environment details.

Core Exercise Loop

Each scenario is meant to drive a critique loop:

1. Present an ambiguous situation.
2. Let participants ask clarification questions.
3. Require them to define the problem, constraints, assumptions, risks, and success criteria.
4. Reveal an implementation or proposed solution.
5. Critique the gap between intent and outcome.
6. Extract heuristics that improve future judgment.

Learning happens through repeated critique cycles, not lectures.

Simulation Principles

Stakeholder agents should:

• reveal information progressively instead of dumping requirements
• speak like stakeholders, not like system design interviewers
• have incomplete knowledge
• sometimes contradict themselves or introduce new constraints late
• surface hidden tradeoffs only when participants ask the right questions

Good exercises force participants to uncover missing requirements, conflicting incentives, rollout risks, UX concerns, and vague success criteria.

System Role

At a high level, pea is responsible for:

• stakeholder conversation runtime
• scenario and agent behavior configuration
• instructor/admin controls for prompts and simulations
• service endpoints the workshop app can call or embed

The workshop app remains responsible for cohort flow, learner UX, facilitation, and the broader curriculum experience.

Current Foundation

The current codebase is built on:

Layer	Technology
Runtime	Cloudflare Workers
Web stack	Remix 3 (alpha)
Package manager	Bun
Database	Cloudflare D1
Session/OAuth	Cloudflare KV
Stateful agents	Durable Objects
Testing	Playwright
Bundling	esbuild

Documentation

Document	Description
docs/product-overview.md	Purpose, scope, and workshop framing
docs/roadmap.md	High-level project phases and priorities
docs/getting-started.md	Local setup and development commands
docs/architecture/index.md	Runtime architecture and system boundaries
docs/environment-variables.md	Environment variable guidance
docs/cloudflare-offerings.md	Optional Cloudflare integrations
docs/agents/setup.md	Local development and verification expectations

---

_{Built with ❤️ by Epic Web}