business

Roadmap

Phase 1 ships now. Phase 2 unlocks teams. Phase 3 makes the platform.

v0.1 — Identity proof (now)

Before Phase 1 in the marketing sense, we ship a v0.1 slice whose only goal is to prove the identity: two people, two machines, one project, live activity visible on each side, automatic Recap when Threads end. The full design is in the v0.1 spec, with the Thread rename amendment on top.

Concretely, v0.1 includes:

• Project as a first-class object (create / invite)
• Node docking (npx manifoldone dock) with a pairing-code flow
• Activity stream from Node → Relay → Dashboard (view-only)
• Recap on Thread end via Manifoldone-managed LLM gateway
• GitHub OAuth for humans, JWT for Nodes

It does not include preview hosting, mobile UI, co-piloting, the prompt/MCP library, or Perforce.

v0.5 — Director Mode (between v0.1 and Phase 1)

Director Mode is the bidirectional Dashboard ↔ Node command channel: start, stop, and steer Threads remotely from the phone or web instead of only watching them. v0.1 keeps the Dashboard read-only; v0.5 promotes it to a control surface — initially against personal Nodes, eventually targeting Project-owned Nodes (build servers) in Phase 2. Director Mode earns its own slice spec when this work begins.

Phase 1 — Solo developer (≈ 0–6 months)

Goal: A solo developer can install Manifoldone, dock their PC, and use Claude Code from any of their devices with no perceptible loss of capability.

Milestones:

• One-line Node install (the parked simplification problem from the spec)
• WebSocket relay handling 10k concurrent Node connections
• Basic Memory (recaps from v0.1 + early prompt-saving UI)
• Web client (chat + Thread viewer + recap viewer)
• Mobile-friendly client (PWA)
• Vercel deploy of manifoldone.com
• Documentation and onboarding flow that doesn’t require a human handoff

The exit criterion: 50 weekly-active solo developers. We don’t ship Phase 2 until that number is real.

Phase 2 — Team coworking (≈ 6–12 months)

Goal: A team of 3–10 people, possibly with multiple Nodes, coworks on a project with full Thread sharing, role-based permissions, and Vibe PRs.

Major work:

• License virtualization or partnership-level resolution of the CC-per-user issue (the structural blocker — see Memory and Pricing)
• Multi-Node federation in the relay (already designed — Phase 2 hardens it)
• Approval flows for cross-user Thread subscription
• Memory MCP scoping and marketplace foundations
• Vibe PR full implementation (intent + chat + recap)
• Hand-off (one Node controlled by one director at a time, with explicit handoff)
• Preview & CI layer (web first; game CI begins in late Phase 2)
• Director console — full surface (v0.1 had a Thread viewer; v0.5 added Director Mode bidirectional commands; Phase 2 hardens the control surface for teams)

The exit criterion: 5 paying teams of 5+ developers each.

Phase 3 — Cloud Nodes, own AI, enterprise (≈ 12–24 months)

Goal: Teams who don’t want to manage their own Nodes can rent cloud-hosted ones. Teams that need a single subscription instead of per-user agent licenses can subscribe to the Manifoldone AI tier. Enterprises with security requirements have a deployment story.

Major work:

• Cloud-hosted Node offering (sandboxed, per-team)
• Manifoldone AI subscription (DeepSeek- or Llama-derived, tuned, hosted)
• Game CI pipelines for Unity / Unreal / Godot
• Enterprise governance (SSO, audit, on-prem relay)
• Skill / MCP marketplace
• Telemetry-driven feedback loop fully in production
• Co-piloting (multiple directors on the same Thread) — the natural extension of Phase 2 hand-off

The exit criterion: First seven-figure deal closed.