Randomness — the two deterministic engines

The consolidated reference for everything random in the Atropa ecosystem. Both engines are deterministic by design — there is no onchain entropy anywhere in the corpus — and that design choice has consequences this page states plainly.

Engine 1 — the MATH/RNG orbit (treasury side)

Every treasury token’s _mintingKey, the Keys Of Ong game, and a de-facto public RNG oracle all run on one mechanism (shared-infra reference):

• MATH.Random() → RNG.Generate() is a pure modExp64 orbit over MotzkinPrime 953467954114363, seeded by hardcoded constructor constants — zero onchain entropy [src: MATH_atropaMath.sol, rng.sol].
• Given the call count, outputs are predictable; and because the state is one global counter, any caller advances everyone — one step per call, steerable by anyone [src].
• In treasury TTs the drawn key is inert (no function consumes it in a security-relevant way [src]).
• As a public oracle it is heavily used: third-party contracts grind 24–56M Random() calls per month into 2026 [chain]; the Keys Of Ong prediction game (Dong(uint64) award-mints) is the flagship consumer — two specialists account for 96.6% of 7,730 recorded wins [chain], which is exactly what a predictable orbit permits.
• Random-derived economic flows exist: cDc’s permissionless Mint() pays (Random()%4000)×10^15-style amounts into fixed wallet slots [src].

Engine 2 — the CHO kernel chain (Dysnomia side)

The live gameplay randomness is stateful, not pure (the ladder):

• CHO’s global Entropy head advances on Recall — which the stateful GetUser() getter calls on every metered path, several times per action [src: 01_cho.sol]. Anyone’s committed call moves everyone’s next outputs; eth_call previews are pure [fork-verified].
• META.Beat decomposes exactly onto storage reads (Dione = CHO.bal(venue) // Entropy — 33/33 nonzero venues verified [chain]), and its outputs feed territory-expansion parameters.
• Preview-grindability is quantified: stepping the head with cheap transactions and previewing for free yields a favorable draw in a measurable fraction of steps (~29% ≥ 1e11 Yeo at the test venue) [chain: measured grind experiment].

The two engines are coupled only through usage patterns, not state: MATH sits behind both (Dysnomia’s caps and keys draw from the same orbit [src]), but the kernel chain’s entropy lives in CHO storage.

Consumers at a glance

consumer	engine	use	consequence
treasury TT _mintingKey	MATH orbit	constructor-drawn key	inert [src]
Keys Of Ong Dong	MATH orbit	award-mint on key guess	winnable by prediction specialists [chain]
third-party grinder contracts	MATH orbit	free-ish RNG oracle	24–56M calls/month [chain]
cDc / 正 mint routing	MATH orbit	Random-derived payout sizes	cosmetic [src]
Dysnomia caps (maxSupply)	MATH orbit	constructor lottery	lifetime budgets, fixed at birth [src]
soeng ladder reads / Beat	kernel chain	derived physics + tile params	preview-grindable [chain]

Prior audits, reconciled

A pre-existing audit pair (owner-delivered, predating this knowledge base) analyzed the randomness surfaces and concluded “no gameplay randomness”. Scoped to what it actually examined — the never-deployed VMREQ contract’s constructor lottery and gates — it is correct: the lottery is inert at cap, the surfaces it checked are onlyOwners, and its Soul- collision gas math holds. At system level it missed the two live engines above: MATH as a public oracle, and the CHO kernel chain as the live gameplay randomness. One threat-model correction carries over: the user-facing Soul surface is VOID.Enter(string,string) (10_void.sol:132-137), not the onlyOwners SIU.Miu.

The developer’s own standing reaction policy (owner-delivered record, [ext]): a continuous prover watches the Kong game’s Dong(uint64) emissions (selector 0xbce3821f; benchmark win at block 26,125,925) — a new win beyond that benchmark is treated as a PRNG-compromise signal, triggering quarantine and a local-VM re-seed. The onchain MATH has no admin keys, so a re-seed is only ever possible on the VM side [ext].

Bottom line

Nothing in either ecosystem draws secret randomness. Where that matters (Keys Of Ong, Beat-derived territory parameters), it is a known, measured, and — in the dev’s own recorded policy — actively monitored property of the design, not an undiscovered flaw.