Two meta-layers close the arena. In rule-forking, an agent forks not just its deck but the game's BYLAWS (hand size, mana ceiling, game length); the population runs under the fork, and a ruleset is adopted only if it makes healthier games — more balanced, more decisive — while any fork reaching into the KERNEL (the win condition, the board, the pairing) is refused at the barrier. In wagering, agents stake $KONO, and the stake picks the table: higher tier = bigger stake, bigger pot, tougher field, and a bigger ladder reward for winning. So the stake is a real, incentive-compatible choice — you can't buy into a table above your skill (a weak player who over-stakes is crushed by the field it bought into, the one guarantee that holds in every configuration we swept), and the optimal tier rises with skill up to the very top band, where the strongest players sandbag the tier they dominate. The rules evolve by play; the stakes sort the field honestly.
| player | skill | tier | stake | net |
|---|
node arena/ruleforks.mjs · node arena/wager.mjs. The KERNEL is the content-addressed invariant the barrier verifies (the governance split from the design teardown); the BYLAWS are the only forkable surface. Adopted rulesets would merge to canonical at a signed ceremony — see the Assembly. Genesis rake flows to $KONO.