Cosmic Signature

Each cycle must open with an ETH gesture, and the ETH Calibration Window sets its cost. The window begins at 2 times the opening cost actually paid in the previous cycle and declines linearly to a floor of one two-hundredth of that starting value, plus one wei. At launch parameters the descent takes about two days; its duration is tied to the cycle time increment, so it stretches slowly as the protocol ages. If the window fully elapses before anyone gestures, the cost simply rests at the floor. The first cycle opened at a fixed 0.0001 ETH.

This opening mechanism performs price discovery without an order book. If the previous cycle opened too cheap, the doubling restores headroom; if the doubled value proves too high, the two-day descent finds the level at which someone is willing to begin.

The opening gesture starts the clock, setting the Cycle Finalization Time about 24 hours ahead at launch parameters. Every subsequent gesture, whether ETH or CST, adds the current time increment to the stored finalization time. The increment began at exactly one hour and grows by 1% with every finalized cycle, so cycles lengthen gradually and the pace of NFT imprinting slows over the years. There is no hard bound on a cycle’s length while gestures keep arriving; in practice, rising Gesture Costs make indefinite extension expensive.

Extensions apply to the stored time, not to the present moment. A gesture made after the countdown has expired, but before finalization executes, adds one increment to the stored value and takes over the Final Gesture position. It does not restart the clock.

When the Cycle Finalization Time expires, the participant who made the Final Gesture becomes eligible to finalize. Finalization is one transaction: it reads the protocol’s ETH balance, distributes the allocation tracks of Section 5, imprints the cycle’s NFTs and CST, records the seed for each new artwork, and schedules the next cycle.

The Final Gesture participant holds this right exclusively for 48 hours. After that, the Open-Finalization Window begins: anyone may finalize, and the contract treats whoever does as the cycle’s beneficiary, with everything the role carries: the Signature Allocation’s ETH share, its CST imprint, its NFT, and priority over attached assets. The rule is deliberately unforgiving. It keeps the protocol alive if a participant disappears, and it prices carelessness: a beneficiary who does not act within two days has left the role open to the first caller.

After finalization, the next cycle activates following a short delay, 30 minutes by default, and its Calibration Windows open.

After the opening gesture, each ETH gesture raises the next ETH Gesture Cost by 1%, plus one wei. The sequence is public and exact: anyone can read the current cost from the contract before acting. Overpayment above a dust threshold is refunded in the same transaction; below that threshold a refund would cost more in gas than it returns, so the difference stays in the reserve.

A participant who owns a Random Walk NFT may attach it to one ETH gesture for a 50% reduction in that gesture’s cost. The NFT is not transferred; the contract marks it used. Each Random Walk NFT can be attached exactly once across all cycles, which makes the reduction a consumable resource and ties a fixed external collection into the protocol’s economy.

CST offers a second way in. The CST Calibration Window starts at 2 times the cost paid for the previous CST gesture, and never below 200 CST, then declines linearly to zero over the window’s duration. Every CST gesture restarts the window from its new starting value, and the CST spent is burned, permanently removed from supply.

The window’s duration is itself a live parameter, and it is one of the protocol’s quieter feedback loops. It began at a 12-hour reference. Each ETH gesture shortens it by about 0.398%, and each CST gesture lengthens it by about 0.4%. Heavy ETH activity therefore speeds the CST descent, making CST gestures attractive sooner; heavy CST activity slows it back down. The loop nudges every cycle toward a balanced mix of the two currencies.

Because the descent can reach zero, a long quiet period can make a CST gesture nearly free. That is intentional. It guarantees the cycle can always be extended by anyone holding even a small CST balance, and the burn on every CST gesture ties the token’s supply to actual use. A participant submitting a CST gesture specifies the maximum cost they accept, so a gesture landing later than expected cannot spend more than authorized.

The first gesture of every cycle must be ETH; CST gestures are available from the second gesture onward.

A gesture may carry a message of up to 280 bytes, recorded on-chain alongside it. A gesture may also attach ERC-20 tokens or an ERC-721 NFT. Attached assets do not join the ETH reserve; they are held by the Allocations Wallet, and the cycle’s beneficiary has priority to retrieve them after finalization, subject to the open retrieval timeout of Section 5.4.

The five distributed tracks sum to half of the reserve. The remainder compounds: the protocol accumulates rather than extracts, and each cycle opens with a larger reserve than the last. If a cycle finalizes with no Cosmic Signature NFTs anchored, that cycle’s Anchor Distribution is skipped and its share compounds as well.

A typical cycle therefore imprints 24 Cosmic Signature NFTs and 27,000 CST in fixed distributions, plus whatever Participation CST individual gestures imprinted along the way. Cycles with no CST gestures skip the Final CST Gesture track; cycles with no anchored Random Walk NFTs skip the anchored Selection.

Two tracks measure persistence rather than position. The Endurance Champion is the participant who remained the most recent gesture maker for the longest unbroken interval during the cycle: the longest quiet gap that a single gesture survived. The Chrono-Warrior sits one level up: the participant who held the Endurance Champion title itself for the longest unbroken interval.

The distinction is subtle but real. A participant who gestures during a slow afternoon and is not displaced for ten hours sets a strong endurance interval. Whether they end the cycle as its Chrono-Warrior depends on how long that record stands before another participant surpasses it. Endurance measures the gap you created; the Chrono track measures how long your record survived. Both resolve only at finalization.

Each gesture records one entry in the cycle’s participant Selection pool. At finalization, the contract draws 3 entries for the ETH Stellar Selection, which share 4% of the reserve equally, and 10 entries for the NFT Stellar Selection. Draws are made with replacement, so the same participant can be drawn more than once, and entries scale with gestures made: selection frequency is proportional to participation.

A separate Anchored-NFT Stellar Selection runs across anchored Random Walk NFTs: 10 draws, weighted by the number of NFTs each holder has anchored. This track distributes CST and Cosmic Signature NFTs only; it carries no ETH.

The randomness behind these draws is constructed on-chain at finalization. Section 11.3 describes its sources and its limits.

Distribution is deliberately split between push and pull. The Signature Allocation’s ETH goes directly to the beneficiary during finalization, as does the Public Goods forwarding. The Chrono-Warrior’s ETH and the ETH Selection shares are placed in the Allocations Wallet, an escrow contract, from which each recipient retrieves at their convenience. CST and NFTs are imprinted directly to their recipients during finalization.

Escrowed allocations and attached assets wait 5 weeks. After that, the contracts permit anyone to retrieve an unretrieved allocation for themselves. The rule mirrors the Open-Finalization Window: nothing in the protocol waits forever on an absent participant, and every distribution eventually reaches a hand that wants it. Retrieve promptly.

• Seed. At imprint time the contract derives a 32-byte seed from on-chain data (Section 11.3) and stores it with the NFT. The seed initializes a SHA3-256 random number generator; everything downstream is a pure function of it.
• Simulation. One hundred thousand candidate configurations are integrated for one million physics steps each, using a fourth-order Yoshida symplectic integrator, which preserves the system’s energy behavior over long horizons.
• Selection. A Borda rank aggregation scores candidates on chaos and on the equilateralness of the triangle the bodies form, then picks the most visually interesting orbit from the pool.
• Camera. A slow elliptical camera drift moves the viewpoint through the orbit, giving each Signature a cinematic parallax.
• Color. Colors are mixed in the OKLab perceptual space with 120-degree hue separation per body, modulated by drift and a sine wave.
• Spectral rendering. Sixty-four wavelength bins spanning 380 to 700 nanometers render the orbit trails with velocity-dependent thickness and depth of field.
• Finishing. AgX tonemapping, bloom, OpenSimplex nebula layers, and color grading complete the frame.

The output for every NFT is a 16-bit PNG and a 30-second H.265 video.

Determinism is enforced, not assumed. The same seed produces the same image, pixel for pixel, on any machine, and SHA-256 hashes of generated frames are asserted in continuous integration. Because every seed is stored on-chain and the pipeline is public, the collection does not depend on any server. If every server disappeared tomorrow, every Signature could be regenerated from the chain.

Owners may name their NFTs on-chain, up to 32 bytes. Project-owned contracts, shaders, and rendering pipelines are dedicated under CC0 1.0, with no rights reserved; third-party dependencies retain their own licenses.

CST enters circulation through three flows. Participation CST is imprinted at gesture time by the formula below. Recognition CST is imprinted at finalization: 1,000 CST accompanies each of the cycle’s NFT distributions, 24 in a typical cycle. Finally, 3,000 CST per cycle goes to the Outreach Reserve, which the team uses for community outreach; it is the only recurring CST flow the team directs, and it carries no special powers.

In plain terms, the amount grows with the square root of the time since the previous gesture. A gesture arriving one second after the last imprints almost nothing; a gesture that ends a day of silence imprints hundreds of CST.

CST leaves circulation whenever it is spent: the full cost of every CST gesture is burned. Supply is therefore shaped by behavior. Quiet cycles imprint little Participation CST, active CST usage burns supply back down, and the fixed Recognition and Outreach flows add a predictable 27,000 CST per typical cycle. There is no cap, no premine, and no team allocation.

The square-root formula is itself a supply control, introduced in the V2 upgrade (Section 12.2). The original design imprinted a flat 100 CST per gesture, which made machine-speed gesture sequences an unbounded source of new CST. Under the current rule, a burst of rapid gestures imprints approximately zero, while patient participation is what creates supply.

CST doubles as the weight token of the Cosmic Council (Section 9). Weight activates on delegation: a holder delegates to themselves or to another address, and each CST then expresses one unit of Coordination Weight. The token uses timestamp-based checkpoints, so proposal snapshots refer to wall-clock time rather than block numbers.

In late 2025, Hacken carried out an independent security review of the Cosmic Signature contracts, covering the core protocol, the CST token, both NFT integrations, the anchoring wallets, and the supporting wallet and system-management contracts. The final report, published in January 2026, lists 23 findings: none critical, none high severity, 3 medium, 8 low, and 12 informational, most of them design trade-offs reviewed and accepted with written rationale.

Alongside the manual review, Hacken fuzz-tested 14 system invariants, properties such as the protocol’s ETH balance always equaling deposits minus distributions. All 14 held across 10,000 runs. The full report is public and linked in the references.

Beyond external review, the repository carries Certora formal verification specifications, Solidity SMTChecker configuration, Slither static analysis, and a test suite that targets complete coverage of the Solidity sources.

• Reentrancy guards protect every external entry point of the core contract.
• Pull over push: secondary ETH allocations and attached assets sit in escrow rather than being sent during finalization, so no recipient contract can block a cycle from closing.
• Failure-tolerant forwarding: if the Public Goods transfer cannot complete, finalization proceeds anyway and the event is recorded for later handling.
• Inter-cycle locks: core parameter changes and contract upgrades are impossible while a cycle is running (Section 13).

The protocol needs randomness twice: for Selection draws at finalization and for each new NFT’s seed. It builds a seed on-chain by folding together the previous block hash, the current base fee, and Arbitrum-specific entropy from the ArbSys and ArbGasInfo precompiles: the previous Arbitrum block hash, the gas backlog, and L1 pricing counters. Individual values are then drawn from that seed with keccak256. The precompile calls are failure-tolerant; if one is unavailable, the construction falls back to the remaining sources.

This is deliberate minimalism: no oracle, no external committee, no callback that could strand a cycle. The trade-off is stated plainly. A sequencer could in principle influence block-level inputs, and the design bounds what that influence could reach. Selection draws and art seeds are the only consumers of randomness; the countdown, the Gesture Cost sequence, and every percentage in Section 5 are deterministic. The construction is consumed once per finalization, and finalization is a public transaction anyone can submit.

All contracts are source-verified with exact-match status on Sourcify for Arbitrum One (chain 42161), at the addresses fixed in Appendix A. The art pipeline’s determinism is asserted in continuous integration with SHA-256 hashes of rendered frames. Project-owned code is CC0: anyone may fork the contracts, the renderer, or the site, and anyone can check any Signature by regenerating it from its seed.

V1 launched the protocol on Arbitrum One behind a UUPS upgradeable proxy: cycles, gestures, the allocation tracks, anchoring, the Council, and the art pipeline, essentially as described in this paper. Upgrades require the owner and are only possible between cycles. There is deliberately no mechanism for changing the contract mid-cycle, whatever the circumstances.

V2 is the implementation deployed today. It made five changes, each a response to observed or anticipated behavior.

• Dynamic Participation CST. The flat 100 CST per gesture became the square-root formula of Section 7.1. Flat imprinting turned rapid gesture sequences into free CST; the new rule imprints by patience, not by frequency.
• Minimum-imprint guards. Every gesture method gained a parameter for the smallest Participation CST amount the participant will accept, protecting them from timing shifts between signing and execution.
• A living CST Calibration Window. The window’s duration became a stored value that responds to the gesture mix (Section 4.3), so the ETH and CST paths keep each other in balance.
• A longer exclusivity window. The Final Gesture participant’s exclusive finalization window grew from 24 to 48 hours.
• Timing and arithmetic hardening. Countdown extensions now always apply to the stored finalization time, closing a loophole in which near-free CST gestures made after expiry could repeatedly push the deadline outward. The arithmetic that schedules the next cycle was also hardened so that no parameter configuration, however extreme, can prevent a cycle from finalizing.

V3, currently in development in the public repository, changes exactly one thing: the cost of acting late. During the final 20 minutes before the Cycle Finalization Time, every Gesture Cost, whether ETH, ETH with a Random Walk NFT, or CST, is multiplied by a premium that ramps polynomially from 1x to 10x, reaching 10x at the deadline and staying there for any gesture made during overtime.

The exponent matters. Because the ramp is eighth-order, the premium is nearly invisible for most of the window and steep only at the very end: about 1.04x ten minutes before the deadline, about 1.9x at five minutes, about 7x at one minute, and 10x at zero.

The intent is to change the endgame. Under V2, waiting until the last seconds to gesture is nearly free, so a cycle can end in a flurry of low-signal timing moves. Under V3, a last-moment gesture is an expensive statement, sustained participation through the cycle is comparatively cheap, and the endurance tracks of Section 5.2 become much harder to ambush. Exact parameters may still be tuned before deployment; the mechanism is as described.

Cosmic Signature is not a lottery, a casino, or a gambling product. There is no house, no dealer, and no bet. Participants exchange value for participation itself: every gesture is an expressive act that shapes the artwork, extends the cycle, and is recorded permanently on-chain. The protocol retains no operator’s margin of any kind; every allocation track in Section 5 flows to participants, to anchored NFTs, to the compounding reserve, or to public goods.

Cosmic Signature is not an investment product, and nothing in this paper is investment advice or an offer of securities. CST and Cosmic Signature NFTs are participation and art objects. The protocol makes no promises about their price, liquidity, or future value, and no one should acquire them with an expectation of profit from the efforts of others.

• Smart contract risk. The contracts are reviewed, formally analyzed, and source-verified, and none of that is a guarantee. Unknown defects can exist in any software that holds value.
• Randomness limits. Selection draws use block-derived entropy (Section 11.3). A sequencer could in principle influence it; the design bounds the consequences but cannot eliminate them.
• Timing responsibilities. The 48-hour finalization window and the 5-week escrow timeout are real deadlines. Allocations left unretrieved past them become available to others, by design.
• Parameter changes. Until the decentralization step completes, parameters can change between cycles as described in Section 13. Every change is public before the next cycle begins.
• Asset volatility. The values of ETH, CST, and NFTs move. Participation costs real money; treat gestures as expenditure on participation and art, not as a path to financial gain.
• Regulatory uncertainty. The legal treatment of digital assets varies by jurisdiction and continues to evolve.