Idea intake
Build an Uber for pets
A deterministic demo run showing how an AI company moves from idea to MVP plan.
Autonomous startup war room
AI Company Simulator
Run progress0%
Demo company ready
Unsaved
Live discussion
Build an Uber for pets
Demo run0%
Monolith or microservices?We are forming a company around: Build an Uber for pets. I want an MVP plan in two weeks and a path to revenue.
Project plan
Execution flowchart
Company formation
Assemble agent roster and personalities
None
War room
Render real-time discussion timeline
Agent roster
Agentic coding
Show Codex activity and generated file previews
Architecture decision
Debate engine
Preserve disagreement and supervisor consensus
Timeline events
Memory
Persist decisions, tasks, metrics, and memory nodes
D1 schema
Quality
Define QA gates for stalled agents and unsafe output
Supervisor policy
Release checkpoint
Ship Build an Uber for pets as a modular monolith with clear internal boundaries, event logs, and extraction points.
Final synthesis and handoff are captured in company memory.
Interactive company map
Agent coordination graph
SVSupervisor
Ava North
CEO Agent
- Status
- Thinking
- Expertise
- Business strategy, sequencing, executive decisions
- Memory
- Prefers two-week MVPs and explicit kill criteria
We are forming a company around: Build an Uber for pets. I want an MVP plan in two weeks and a path to revenue.
Supervisor resolves the conflict: build a modular monolith now, define service boundaries for later extraction.
Project management layer
Autonomous Kanban
Backlog
Render real-time discussion timeline
Conversation can become hard to scan.
Show Codex activity and generated file previews
Generated code needs clear provenance.
Preserve disagreement and supervisor consensus
Agents may loop without a decision policy.
Persist decisions, tasks, metrics, and memory nodes
Audit history must survive reloads.
Define QA gates for stalled agents and unsafe output
Autonomy without guardrails damages trust.
In Progress
Assemble agent roster and personalities
Weak role prompts create bland discussion.
Review
Completed
Agentic coding integration
Engineer invokes Codex
Awaiting engineering kickoff...
AI debate engine
Monolith or microservices?
Owen Park
Start with a modular monolith. It keeps data consistency and velocity high while the domain is still moving.
Milo Chen
Users do not care about service topology. They care that the core workflow is useful and fast.
Supervisor decision
Ship Build an Uber for pets as a modular monolith with clear internal boundaries, event logs, and extraction points.
Minority view preserved: Investor wants a scale narrative, but accepts deferred service extraction.Shared company memory
Decision graph
Live war room, Kanban, debate, health, memory, and generated files for Build an Uber for pets.
decision nodeGenerated deliverables
Architecture, workflow, API, wireframes, MVP, and pitch content
Deliverable 1
System architecture
- React/vinext frontend renders the live simulator, company map, Kanban, debate, metrics, memory, and artifacts.
- Node-compatible app routes expose persistence APIs that map to D1 in hosted Sites environments.
- A future WebSocket gateway streams supervisor events, agent messages, task movements, and coding logs.
- OpenAI Agents SDK workers run role prompts, tool calls, debate rounds, and supervisor conflict resolution.
Performance metrics
Agent contribution scores
Ava NorthDecision quality
Milo ChenPrioritization
Rhea ValeUX clarity
Owen ParkBuild velocity
Nia BrooksRisk discovery
Sol ReyesLaunch leverage
Iris StoneCompliance coverage
Gabe HoltViability pressure