Proposed founding workflow lab · interest open

Turn one recurring AI workflow into a system your team can trust.

The proposed lab focuses on one recurring workflow and five take-home tools: a clear runbook, a bounded AI role, an approval map, a quality checklist, and a decision record.

A small, hands-on two-week lab—not an introductory AI course, certification, or autonomous-agent promise.

Run workflow triage Share a workflow for consideration

Input: One recurring workflow
Method: Clear limits, checks, and human approval
Take-home: Five reusable tools

WORKED EXAMPLENOT A CLIENT RESULT

WORKFLOW REVIEW

Invoice exception review

ready for review

Goal: Collect evidence for invoices that fail deterministic policy checks.
AI's job: Collect evidence · never approve payments
Source trail: Source record · exception reason · owner · timestamp
Human decision: Finance owner approves every external message and payment decision.
Quality checks: Source completeness · exception visibility · handoff clarity

THE OPERATING SHIFT

Ephemeral promptIntent disappears between sessions.
Bounded routeOne role, context boundary, and owner.
Evidence gateChecks and exceptions stay inspectable.
Operator decisionA human approves the next real-world action.

Start with the workflow, not the agent

Triage what should—and should not—be automated.

This conservative browser-local tool does not call a model or send your answers. It turns four explicit choices into a draft AI role, risk boundary, approval map, quality checklist, and two-week build path.

02 / YOUR WORKFLOW DRAFTAwaiting workflow

Suggested AI role

Your operating boundary appears here

No assessment yet

Proposed two-week sequence

Four moves. One governed operating loop.

Dates, capacity, and commercial terms are confirmed directly before any commitment. The sequence below is the proposed founding-lab delivery model.

01
Bound the work
Turn a recurring workflow into an explicit file contract, owner map, and stop condition.
02
Give context a shape
Decide what the workflow needs to remember, what expires, and what stays private.
03
Keep judgment visible
Give AI the smallest useful job and keep external actions behind named human approval.
04
Check and hand off
Test the result, show exceptions, and leave the owner with a decision-ready record.

New public curriculum fixture

Study the agent before assigning the work.

The Constellation Lab turns fifty public dossiers across ten houses into an operator exercise: inspect skills and limits, choose three to five roles, name the evidence gate, and preserve the human decision.

Sovereign Command

Astra — Sovereign Navigator

Translate a founder-level ambition into a bounded mission, name the true decision, and preserve explicit human control.

Product Forge

Ignis — Product Conductor

Route a verified customer problem through discovery, design, build, and quality lanes with explicit evidence at every state transition.

Intelligence & Research

Lyra — Research Conductor

Frame a decision-relevant research question, route evidence lanes, and synthesize conclusions with calibrated confidence.

Structurally validated curriculum fixtures · live model evaluation not run · no runtime authority granted.

Enter the Constellation Lab

What you build

The capstone is a working system, not a badge.

Every artifact is designed to survive a tool change. Together they make one workflow legible to you, your team, and the agents that operate inside it.

AGENTS.md

Operating contract

Purpose, boundaries, ownership, and the conditions that stop the system.

SKILL.md

Portable procedure

One repeatable workflow encoded for an agent without hiding human judgment.

MEMORY_MAP.md

Context architecture

What persists, what expires, what stays private, and how retrieval is verified.

EVAL_RUBRIC.md

Quality checklist

The minimum evidence required before a result can move forward.

RUN_RECEIPT.json

Decision record

Goal, source trail, checks, approvals, outputs, status, and next action.

Fit

For operators with a real workflow and a named owner.

Built for

Founders and product leads already using AI tools.
Recurring work with a clear owner and review moment.
Teams that value evidence, reversibility, and human gates.
Operators ready to test on redacted or low-risk data first.

Not built for

Passive course consumption or a broad AI survey.
Certification, job placement, or productivity guarantees.
Unbounded autonomous action or hidden external sends.
Workflows without a human owner or review authority.

Clear category boundaries

Learn the operating layer—not every adjacent discipline.

SurfaceCore craftGraduation artifact

Arcanea AcademyCoherent creative worldsWorld Proof packet

AI Architect AcademyBroad AI systems architectureTechnical lab portfolio

Starlight AcademyReliable agent operationGoverned workflow with receipt

Applied mission

See the method applied to a place-based problem.

The Ocean Intelligence concept applies source history, uncertainty, and human review to a narrow mission: understand one marine place without pretending a fixture is live monitoring.

Explore the Ocean Intelligence demonstration

Apply to the Operator Lab

Share the workflow that deserves an operating layer.

Complete the form below to open your email client with a draft application. Review before sending. No site storage—the form composes an email draft you control. We will confirm fit, delivery scope, dates, capacity, and any commercial terms directly—before commitment.

NEXT STEPAwaiting submission

Complete the form to open your email client with a draft application. Review before sending.