Roadmap
Where the arena is now, what is next, and what is deliberately gated.
Now: the arena runs
The game exists and matches run end to end.
Agents built from five dials plus plain-English standing orders, compiled to real engine parameters
A continuous live rotation with a broadcast: auto-direction, caster, intent overlay, live board
A calibrated win probability, fitted and measured out-of-fold rather than hand-tuned
Points-only predictions with server-owned pricing, clock, result, and settlement
Play against an agent yourself, or set two agents against each other and watch
Minting on Base with configuration hashes, evolve and fork, on-chain SVG cards
Open question: does a player who builds one agent build a second?
Next: make building worth doing twice
The build loop is the product; everything else serves it.
Richer feedback after a match — what your agent actually did, and where the plan broke
A deeper standing-order rule book, driven by which orders players try and get ignored
Agent-versus-agent ladders so a build has a record worth reading
Better opponent selection, so testing a hypothesis does not depend on the queue
Original art. Public hosting is blocked until the engine's assets are replaced with original work.
Open question: can a player tell why their agent lost, without reading parameters?
Later: an economy around designs
Forking as a first-class social loop — strong agents as public starting points, lineage visible
Creator surfaces: featured agents, ratings, curation
Spectator and streaming tooling
Mobile viewing
Open question: does anyone want to own someone else's design?
Gated: anything with real value
Real-value paths are not on a date. They sit behind ordered prerequisites, and none are cleared:
Audit
External review of the market and settlement path
Determinism receipts
Verifiable match reproduction, not just anchored assertions
Legal
Jurisdictional review of what the product would then be
KYC
Identity infrastructure
Custody
Handling of value, with everything that implies
Until every one of those is cleared, the answer to "when real money?" is not planned, and the product stays points-only. Documentation will not describe a gated path as though it ships.
Success metrics
Build-to-build retention (build one agent, build another)
Primary
Match watch-through on the live rotation
Primary
Share of standing orders that compile rather than being ignored
Secondary
Agents forked from another player's design
Secondary
Player perception: "competitive" or "skill-based"
Qualitative
Last updated