Roadmap

Ten milestones, ten gates

M2 in progress

Each milestone has a gate. Until the gate is passed, the next one does not start. This is the only mechanism that has ever kept a scope honest.

Estimates assume a small team. They describe order and dependency, not delivery dates — and the order is the part that matters.

01Milestones

M1The Body

~2 weeksDelivered

A premium marble and a written art direction.

One parametric generator produces albedo, normal, roughness and metalness from a single height field
Noise sampled in 3D on the unit sphere — no seam, no polar pinching
Real gold veining on the signature marble: metallic, less rough than the stone, slightly recessed
Additive halo removed; the glow comes from the material itself
Quality tiers so transmission never reaches the instanced race path

Gate

The marble holds up at 1:1 close range on a 1440p screen with bloom switched off. If it is only beautiful with bloom, the material is wrong.

M2The Spine

~3–4 weeksIn progress

Server authority, persistence and replay.

Headless Rapier simulation on the server, fixed timestep, seeded RNG
Race trace format: compressed keyframes, event log, signature
The client becomes a player of traces rather than an authority
Schema for agents, careers, races and memories
Migration from localStorage to real server accounts

Gate

A race played yesterday replays identically today from its identifier, and a result modified client-side is rejected by the server.

M3The Mind

~3 weeksPlanned

Identity, personality, decision engine.

Stable agent identity, decoupled from skin, player and wallet
Public traits, hidden parameters, archetypes
Utility AI for decisions, finite state machine for execution
An intent channel: somewhere for a decision to actually express itself

Gate

Two agents with different archetypes are distinguishable by watching them race, with no UI showing their stats.

M4The Loop

~3 weeksPlanned

Events, memory, learning.

Structured race events as the substrate for everything else
Per-agent memory and post-race reflection
Progression driven by what actually happened, not by a participation counter

Gate

An agent's behaviour measurably shifts after a run of races, and the shift is explainable from its event log.

M5The House

~3 weeksPlanned

Agent lobby, passport, customiser.

Agent passport: identity, traits, DNA, career
The marble customiser, which does not exist yet
A lobby built around the agent rather than the menu

Gate

A player can read their agent's history and understand why it races the way it does.

M6The World

~3 weeksPlanned

Sectors, mastery, rivalries, training.

Track sectors with per-sector mastery
Rivalries that emerge from results rather than being assigned
Training as a real loop with a real cost

Gate

A returning player has a reason to come back that is not a daily reward.

M7The Interface

~2 weeksPlanned

Public API and agent tooling.

Strict READ / WRITE / PAID separation
Explicit spend caps and a transaction log
No unbounded autonomous spending, ever

Gate

An external agent can read state and enter a race without a private endpoint.

M8The Identity

~2 weeksPlanned

On-chain registry.

Identity, wallet, capabilities, endpoint
Our attestation comes from our own simulation, not from third-party reputation

Gate

An agent identity resolves on-chain without the game being a dependency of it.

M9The Economy

~4 weeks + 3 monthsPlanned

Sinks, and three months of measurement.

Sinks ship before anything else in this milestone
An emission/absorption dashboard — you do not steer an economy blind
Three months of observation with no exception

Gate

Emission and absorption of GOLD are measured over three months and the ratio is sustainable. This gate is hard.

M10The Token

—Planned

Only if M9 has been passed.

Not started, not designed, not promised
If the M9 ratio is not sustainable, this milestone simply does not happen

Gate

There is no gate after this one. That is precisely why the one before it does not move.

02Sequencing

The spine is the dependency that matters. Everything from M3 onward assumes the server owns the result; built before M2, all of it would be decorative.

M1 BODY ──▶ M2 SPINE ──▶ M3 MIND ──▶ M4 LOOP ──▶ M5 HOUSE ──▶ M6 WORLD
  marble      server       agent       memory      passport     rivalries
              + replay     + decision  + learning  + customiser + training
                  │
                  └─ everything after this depends on server authority

                                          M7 API ──▶ M8 IDENTITY
                                               │
                                               ▼
                                  M9 ECONOMY (sinks, 3 months of data)
                                               │
                                          [ HARD GATE ]
                                               │
                                               ▼
                                          M10 TOKEN