project-yaml · supabase_json

Bezel Team project-yaml

People with real product ideas need more than a coding agent or clickthrough demo. They need a guided production process that turns intent into reviewed screens, architecture, build-ready artifacts, implemented software, deployment, and iteration. Bezel is Dreamborn's AI-first product studio infrastructure for that workflow.

Planning Surface

Use this to decide what happens next.

Status

draft

open questions

Should Bezel Core be permissive open source, source-available, or delayed until Team has first users?
Is the first Team intake path GitHub Issues, Linear, or both?
Should Team initially run workers in Justin-managed infrastructure only, or support customer-connected workers in V1?
What is the first paid buyer profile: solo founder, small engineering team, or agency/product studio?

Agent Handoff

Start Here

Completion Evidence

No explicit evidence field yet. Require tests, screenshots, linked PRs, or reviewed outputs before marking complete.

Thesis

Open Questions

Should Bezel Core be permissive open source, source-available, or delayed until Team has first users?
Is the first Team intake path GitHub Issues, Linear, or both?
Should Team initially run workers in Justin-managed infrastructure only, or support customer-connected workers in V1?
What is the first paid buyer profile: solo founder, small engineering team, or agency/product studio?

Structured Payload

Machine-readable source fields

kind

bezel.project

users

problem	segment	success
Has a real product or internal tool idea but cannot translate it into production-ready UX, architecture, and implementation without hiring a full product/dev team.	founder_or_operator_with_product_idea	Can walk through a guided process, review concrete wireframes/screens, approve a build plan, and receive working production software rather than a clickthrough demo.
Understands the workflow pain deeply but lacks software design and engineering infrastructure.	small_business_or_domain_expert	Sees their workflow converted into screens, data model, build phases, and a hosted product they can actually use.
Needs delivery leverage for client product builds.	agency_or_consultant_later	Uses Bezel as a repeatable production system for design-to-build delivery after the first-party workflow is proven.

thesis

project

bezel-team

name

Bezel Team

status

intake

maturity

production_product

positioning

AI-first product development infrastructure for people who want real production software built

product line

Bezel

public naming

rule

External surfaces, docs, UI, repos, and customer language use Bezel only. RedKey is legacy/internal historical context and should not appear in product copy.

legacy names

RedKey

canonical name

Bezel

version

filename

project.yaml

deployment

rule

The product lives under Dreamborn as build.dreamborn.ai. Customer-facing language can say Dreamborn Build powered by Bezel or Bezel by Dreamborn; avoid RedKey.

company domain

dreamborn.ai

canonical domain

build.dreamborn.ai

public product name

Bezel

product surface name

Dreamborn Build

updated at

2026-05-04T12:49:12.473Z

monetization

model

Open-core distribution wedge plus hosted team SaaS

revenue rule

Do not over-invest in public Core unless it directly improves Team acquisition, trust, or onboarding.

suggested tiers

tier	buyer	price	value
Core	individual developers	free	local runner and proof-of-concept trust
Team	small engineering teams	paid per seat or per active repo	hosted dashboard, shared history, managed workers, review gates
Enterprise	regulated or larger teams	annual contract	SSO/RBAC, retention, audit receipts, private deployment, support

initial scope

in v1 team

Product owner intake for real software, not demo/clickthrough builds
Screen inventory, wireframe generation, and HITL approval flow
UX flow and state breakdown including empty/loading/error/success states
Architecture and production constraints artifact
Build plan from approved screens and architecture
Managed AI dev team execution against approved artifacts
Production deployment option on Dreamborn infrastructure
Review gates at Design, Architecture, Build Plan, PR/Implementation, and Deploy
AI-first opportunity map included in design phase

out v1 team

Generic coding-agent SaaS for engineering teams
Public Core CLI
Clickthrough-only prototypes as deliverables
Autonomous build before screen/design approval
Marketplace
Enterprise self-hosting
Public explanation of private ledger implementation details
Included full build/operation of all recommended support agents; those are expansion scope unless explicitly purchased

product split

private moat

name

Bezel Ledger

rule

Do not expose provider mechanics, topic identifiers, operator credentials, or private governance repair flows in public repos or product UI.

public language

tamper-evident audit receipts and durable execution history

private language

private ledger provider behind the RunLedger/ReceiptStore interface

bezel team paid

purpose

Production product-studio and delivery engine

includes

Guided product intake and problem framing
AI-assisted screen inventory and user-flow mapping
Human-reviewable wireframes and screen states before build
Architecture and production-readiness planning
Build-ready artifacts and acceptance criteria
Managed AI dev team execution with review gates
Production deployment on Dreamborn infrastructure or customer-connected hosting
Persistent product history, decisions, proof bundles, and iteration backlog
Managed product support and improvement cycles
AI-first opportunity map that recommends supporting agents and workflow shifts from SaaS to agentic operations
Optional paid agent-build expansion using Dreamborn/Bezel infrastructure

bezel core public

purpose

Future distribution wedge only after Bezel Product Studio workflow is repeatable

excludes

Hosted multi-tenant control plane
Private ledger implementation
Enterprise audit receipts
Managed worker fleet
Billing/metering
Team RBAC/SSO
Governance repair intelligence
Private agent memory and production reconciliation internals

includes

CLI: bezel init, bezel run, bezel watch
Local workflow file: bezel.workflow.md or project.yaml-derived config
GitHub Issues or Linear task adapter
Per-task worktree manager
One local agent runner provider to start
Local SQLite/Postgres event log
Basic run lifecycle: queued, running, blocked, review, done
Local dashboard
Proof bundle: changed files, test command output, summary, PR link when available
Provider interfaces: QueueProvider, RunLedger, ReceiptStore, TaskStateAdapter

license direction

TBD after business review; likely permissive for CLI/SDK or source-available for server pieces if copy risk is too high

open questions

Should Bezel Core be permissive open source, source-available, or delayed until Team has first users?
Is the first Team intake path GitHub Issues, Linear, or both?
Should Team initially run workers in Justin-managed infrastructure only, or support customer-connected workers in V1?
What is the first paid buyer profile: solo founder, small engineering team, or agency/product studio?

product strategy

sequencing

Build the outcome-oriented product studio first for people who want products built. Extract developer/team infrastructure later once the workflow is proven.

public wedge

Bezel Team/Core later, only after the product-studio workflow is repeatable

core principle

Public Core demonstrates the agent-runner workflow. Team owns the daily operational value and monetization.

primary product

Dreamborn Build / Bezel Product Studio

production not demo

true

value proposition

one liner

Dreamborn Build turns a product idea into production software through multi-model research, screen-by-screen design, AI dev execution, and managed hosting.

team promise

Rent an AI product team with a production process: design the product, review the screens, approve the plan, build the software, deploy it, and keep iterating.

not the pitch

Not a demo harness
Not a prompt collection
Not a blockchain product in public positioning
Not another chat UI for code generation

developer promise

Not the primary wedge. Developers may later use Bezel infrastructure, but V1 serves product owners who want real software delivered.

orientation system

principle

The customer should always know where they are in the product-build process and what they are doing next.

action rule

Every review screen must have one primary next action and one clear escape: request changes, ask question, or save for later.

artifact rule

Orientation data should come from phase/gate status and artifact approval state, not static copy.

shell pattern

Persistent project progress rail plus current-step header plus next-action panel.

required questions

Where am I?
What phase is this?
What did Bezel just produce or learn?
What decision/action is needed from me?
What happens next after I approve or request changes?
What is blocked?

architecture direction

ledger plane

Receipt writes go through an interface; local/dev uses a database event log, Team uses private Bezel Ledger.

control plane

Bezel Team owns team state, run visibility, policy, review, billing, and durable history.

execution plane

Workers execute in isolated repo workspaces with scoped secrets and explicit lifecycle reporting.

core abstractions

TaskSource
QueueProvider
WorkspaceManager
AgentRunner
RunLedger
ReceiptStore
ReviewGate
NotificationSink
PolicyEngine

ai first transformation

examples

Instead of building only a CRM dashboard, identify a follow-up agent, lead research agent, and stale-opportunity recovery agent.
Instead of building only a support portal, identify a triage agent, knowledge-gap agent, escalation agent, and customer-summary agent.
Instead of building only an ops dashboard, identify monitoring agents, exception-resolution agents, and approval workflows.

principle

Every product design pass should identify where the customer can move from SaaS-shaped human workflows to AI-first workflows.

core deliverable

AI-first opportunity map: recommended agents, workflow automations, human review gates, and data/permission requirements discovered during product design.

business boundary

Lightweight agent suggestions and workflow concepts are included in the design process. Building and operating the actual support agents is additional scope/pricing because it uses Dreamborn/Bezel infrastructure beyond the base product build.

customer visible language

Bezel does not only ask what app you want. It asks what work should stop being manual SaaS work and become agent-supported work.

multi model collaboration

gates

intake synthesis
AI-first opportunity map
screen map and wireframe critique
architecture review
build plan review
implementation/proof review
release readiness review
iteration strategy

principle

Every major product decision should support multi-model research and collaboration, not a single-model black box.

artifact rule

Raw model outputs are evidence. The product should save both the model-review evidence and a human-readable synthesis artifact before the next gate.

product behavior

Bezel should ask models to disagree, compare alternatives, critique assumptions, and produce synthesized recommendations at key gates.

model roles initial

role	purpose	model family
product_strategy_critic	Challenge positioning, scope, user value, and product clarity.	Claude
market_and_systems_researcher	Broaden research, compare patterns, suggest AI-first workflow shifts, and stress-test market assumptions.	Gemini
implementation_reviewer	Translate decisions into executable artifacts, testable implementation plans, and code review.	OpenAI/Codex

customer visible language

Dreamborn researches your product with multiple AI perspectives, shows the tradeoffs, and turns the best answer into build-ready artifacts.

production non negotiables

Production product from day one; no fake task loops, canned demo states, or clickthrough-only deliverables.
Every build starts from approved human-reviewable screens/wireframes, not only prose requirements.
Every screen has states: default, empty, loading, error, permission/locked where relevant, and success/complete.
Build cannot start until product intent, screen map, key wireframes, architecture, and plan are approved.
No external product surface uses RedKey naming.
Private ledger mechanics stay hidden behind Bezel Ledger abstractions.
Human review and stop controls are first-class at every phase.
Secrets are never passed through repo files, prompts, or public logs.
Design artifacts must distinguish included product-build scope from optional supporting-agent expansion scope.
Major gates should include model comparison evidence and a synthesized recommendation, not a single unchallenged model answer.
Every customer-facing screen must show current phase, current decision, and next action.

immediate artifacts to create next

MODEL_COLLABORATION_PROTOCOL.json
AI_FIRST_OPPORTUNITY_MAP.json
INTENT.json for Bezel Product Studio
SCREEN_MAP.json for HITL wireframe/screen breakdown
WIREFRAME_SPEC.json defining screen state requirements and review gates
REQUIREMENTS.json for production V1
ARCHITECTURE.json separating product studio, design layer, execution layer, hosting, and private ledger
POLICY.json with naming, production, secrets, and no-clickthrough-deliverable rules
PLAN.json with M-00 substrate audit before implementation dispatch