Rally: attention as the
scarce resource in token launches

1Introduction

Launching was solved. Getting anyone to give a shit wasn't.

Between 2023 and 2026 the cost of creating a liquid, tradeable token on a high-throughput chain fell effectively to zero. Bonding-curve launchpads removed the need for capital, counterparties, listing processes, and technical skill. This was a genuine achievement and it is now permanently commoditised: no venue can build a durable advantage on the act of issuance, because issuance is a solved problem available everywhere at the same price.

The consequence is a market with unbounded supply and fixed demand. Thousands of markets are created daily. The number of people willing to evaluate a new market has not grown at anything like the same rate, and the number of seconds any one of them can spend per market has collapsed. In practice, the overwhelming majority of launches are never seen by a human who was not already looking for them.

Every existing launchpad treats this as somebody else's problem. The platform mints the asset, takes its cut, and leaves distribution to the creator's ability to shout on social media. We take the opposite position: distribution is the product, and a launch venue that does not fund and allocate attention is selling the half of the problem that was already free.

Rally is deployed on Robinhood Chain, the Ethereum Layer 2 that opened to the public on 1 July 2026. The choice is not incidental. That chain is the sharpest live instance of the problem this paper describes: within three weeks of mainnet it had overtaken Base in daily active users, and at peak it was absorbing on the order of 18,000 token launches per day. An enormous, fast-growing venue with functionally unlimited issuance and no mechanism whatsoever for deciding what deserves to be seen is precisely where a marketpad should exist.

2The problem

2.1Supply is free, attention is not

Token supply is now produced at near-zero marginal cost with no natural limit. Attention is produced by humans, is strictly bounded, and does not scale. Any market where one input is free and unbounded while the complementary input is scarce and fixed will price the scarce input — except that current launchpads have no mechanism to price attention at all. It is allocated by recency, by luck, and by off-platform social capital.

The magnitudes are not marginal. On Robinhood Chain alone, peak issuance has run at roughly 18,000 new tokens per day. Assume — generously — that fifty thousand people actively hunt new launches on that chain, and that each is willing to spend ten seconds forming a first impression. That is about 500,000 seconds of evaluative attention per day chasing 18,000 markets: under thirty seconds of total human consideration available per launch, before accounting for the fact that attention is not distributed evenly but concentrates almost entirely on whatever is already visible.

The tail is therefore not underpriced. It is unpriced, and structurally unreachable. No amount of creator effort resolves this at the individual level, because the constraint is aggregate.

2.2The recency lottery

The default discovery surface is a newest-first feed. This has a specific and predictable consequence: since ranking is a function of creation time, and creation is free, the dominant strategy is to create constantly. Serial deployment is not an abuse of the newest-first feed; it is the rational response to it. The feed rewards volume of launches, so it receives volume of launches, and quality is diluted until discovery is worthless.

2.3The fee model selects for abandonment

Consider the canonical structure. A trader pays a fee; the platform takes a share; the creator takes a share; both are paid immediately and unconditionally; the value then leaves the system entirely. Three properties follow.

First, payment is front-loaded. Launch-day volume is typically the largest volume a market will ever see, so the creator's expected earnings are concentrated precisely in the window where the market has the least information about whether the project is real.

Second, payment is unconditional. Nothing about the creator's payout depends on the market existing tomorrow. The expected value of abandoning a market and deploying a new one is therefore strictly greater than the expected value of maintaining it, for any creator without an external reputational stake.

Third, value is extractive. Fee revenue exits to the platform's treasury and the creator's wallet. Nothing is recycled into the discovery layer that every launch depends on, which is why every launchpad's discovery layer is equally bad — none of them are funded.

3The marketpad model

3.1Definition

We define a marketpad as a launch venue in which issuance is the entry condition and the market itself is the product. A marketpad accepts that deployment is commoditised, prices it at approximately nothing, and competes instead on three functions: ranking markets by observed behaviour, funding the visibility of markets that earn it, and structuring payouts so that durability outperforms churn.

3.2Ranking by behaviour, not by age

Rally replaces the recency feed with boards ordered by Rally Score, a bounded composite recomputed continuously per market:

Two choices matter here. Velocity is used rather than absolute volume, so a small market accelerating can outrank a large market flatlining — the score measures whether something is happening, not whether something already happened. And retention is weighted significantly, which is the term that makes pure exit-liquidity launches structurally unable to rank.

The Survivors board — markets still scoring above threshold after seven days — is the surface that has no analogue on existing venues, because on existing venues almost nothing is designed to survive seven days.

3.3Attention as a funded protocol resource

If attention is the scarce input, a venue must acquire and allocate it deliberately rather than hoping creators do it externally for free. Rally maintains an Attention Pool: a protocol-owned budget, funded continuously from fee flow, spent on featured placement, trader rebate campaigns, liquidity incentives, and creator grants.

The Attention Pool is what converts the fee architecture from a payout schedule into a loop. Fees fund visibility; visibility produces volume; volume produces fees. The allocation of the pool is governed by staked token holders, which gives the token a function that is not speculative: it directs a real budget.

4Fee economy

4.1Design goals

• Creator earnings should be an increasing function of market durability, not of launch-day volume.
• A meaningful share of fee revenue should remain inside the ecosystem rather than exiting it.
• Value extracted by adversarial behaviour should be recaptured, not merely discouraged.
• Token value accrual should be funded by revenue, not by emission.

4.2The split

A 1% fee applies to every swap on the pad. It is divided as follows.

Table 1 — base fee allocation. Buckets sum to 100%; redistribution is internal.

Note the aggregate: 40% of gross fee revenue is recycled (25% Reserve plus 15% Attention Pool base), rising above 40% whenever creators underperform, whenever vesting is forfeited, and whenever the anti-sniper mechanism fires. Only 8% is structurally extractive.

4.3Creator Score

A flat creator fee pays identical amounts to a creator who deployed four hundred tokens this morning and a creator who built a market people still trade next month. This is the central pricing error we are correcting. Creator share is a function:

Survival measures days above the liquidity floor. Retention measures holders still holding after seven days as a fraction of peak holders. SustainedVolume is rolling seven-day volume over launch-day volume — a market that traded once and stopped scores near zero on this term by construction. Conduct is a penalty term covering bundled supply, oversized creator sells, and liquidity manipulation; it is hard-zeroed on a detected rug.

The redistribution property is the important one. Every basis point a weak creator fails to earn is not saved by the protocol — it is routed to the Attention Pool, which buys visibility for other launches. Weak launches directly subsidise the discovery of strong ones.

4.4Conditional vesting

Merit weighting alone is insufficient, because Creator Score is necessarily backward-looking and a creator can extract launch-day fees before any score has accumulated. Rally therefore splits creator payouts across time:

1. 40% of accrued creator fees stream immediately and are claimable in real time.
2. 60% streams linearly across 30 days, conditional on the market remaining above the liquidity floor and on no conduct flag being raised.
3. Unstreamed remainder on a failed condition is forfeited to the Attention Pool.

This inverts the temporal incentive. Under the canonical model, expected creator earnings are maximised at t=0 and decline monotonically. Under Rally, the majority of expected earnings sit thirty days out and are contingent on the market being alive to collect them.

4.5Reserve, buyback, and burn

The Rally Reserve accumulates 25% of gross fees in ETH and executes weekly TWAP purchases of $RALLY on the open market. Of each purchase, 60% is burned permanently and 40% is distributed to stakers. Because total supply is fixed at 1,000,000,000 and the contract exposes no mint function, burn is the only direction supply can move.

We state the obvious caveat directly: this is a mechanism linking protocol revenue to token supply, not a price guarantee. If the pad does no volume, there is no revenue, no buyback, and no burn.

5The $RALLY token

5.1Supply and allocation

Total supply is fixed at 1,000,000,000 $RALLY, an ERC-20 deployed on Robinhood Chain (chain ID 4663). The contract has no mint function, no pause or blacklist hook, no upgradeable proxy, and ownership is renounced at deployment. There is no inflation: every incentive in the protocol is paid from the fixed allocation or from realised fee revenue, denominated in ETH.

Table 2 — supply allocation and unlock. Full schedule in Tokenomics.

Insider allocation is 24% combined (team plus treasury) and none of it is liquid at TGE. By the time the first team token unlocks, the pad has been live for twelve months and its fee economy has been observable in public for the same period.

5.2Emission

The Rally Engine distributes 250,000,000 $RALLY over four years — 100M, 75M, 45M, 30M by year — split 40% to creators by Creator Score, 30% to traders by wash-filtered volume, 20% to liquidity providers by depth-weighted uptime, and 10% to referrals and quests.

The schedule decays deliberately and steeply. Emissions exist to bootstrap a floor that has no participants yet; a protocol that still requires them in year four has not found a market. From year three onward, the dominant source of participant income is intended to be fee revenue, not emission.

5.3Staking

Staked $RALLY becomes sRALLY, with lock-tier multipliers of 1.0×, 1.6×, 2.5× and 4.0× at flexible, 30, 90 and 365 days respectively. sRALLY receives the 12% fee bucket plus 40% of every buyback, paid in ETH, and carries the vote weight that allocates the Attention Pool each week. Early exit is permitted at a 25% penalty on principal, which is burned.

5.4Value accrual

$RALLY accrues value through three channels, all of which are downstream of realised volume rather than of narrative: revenue distribution in ETH to stakers, supply reduction via burned buybacks, and governance rights over the Attention Pool budget. None of the three requires new token issuance to function.

6Incentive analysis

The test of a fee design is what the rational strategy becomes under it. We compare four participant types under the canonical model and under Rally.

Table 3 — dominant strategy comparison.

The claim we are making is narrow and falsifiable: under this fee structure, abandoning a market is more expensive than maintaining one for any creator whose market has non-trivial residual volume, and the cost of abandonment is transferred to the discovery budget rather than retained by the protocol. We are not claiming this eliminates low-effort launches. It prices them.

7Attack surface

Any scoring system that gates money invites manipulation of its inputs. We enumerate the principal vectors and the mitigations, including where mitigation is partial.

Table 4 — attack vectors and residual exposure.

We treat detection as an ongoing adversarial process rather than a shipped feature. Publishing the weights makes them targets; not publishing them makes the fee split unauditable. We choose auditability and accept the cost.

8Governance

Three phases. Phase 1 places parameters under a core multisig while the mechanism is tuned against live behaviour. Phase 2 hands weekly Attention Pool allocation to sRALLY gauge votes — real control over a real budget, scoped narrowly. Phase 3 moves fee splits and treasury spend to on-chain proposals with a timelock.

Scoring weights are permanently excluded from token-holder governance. The parameters that resist manipulation cannot be governed by the parties with the most to gain from relaxing them; this is a deliberate limit on decentralisation and we state it rather than obscure it.

9Roadmap

1. Phase 0 — Complete. Fee mechanics designed and tested. Unannounced test deploys run in production; one exceeded $1M in volume within an hour without any announcement.
2. Phase 1 — Launch. Pad opens. Boards, Creator Score, conditional vesting and the Attention Pool live from block one.
3. Phase 2 — Token. $RALLY TGE via fair launch on the pad. Staking, Reserve buybacks, Engine epoch one.
4. Phase 3 — Gauges. sRALLY governance over Attention Pool allocation. Trencher Airdrop Season 1 distribution.
5. Phase 4 — Open floor. Public scoring API, third-party board clients, integrations for external terminals.
6. Phase 5 — Parameter governance. Timelocked on-chain control of fee splits and treasury.

10Risks & limitations

• Demand risk. Every mechanism here is proportional to platform volume. Low volume means low fees, no buyback, and an Attention Pool with nothing in it. The design does not create demand; it allocates it.
• Complexity risk. Merit weighting and conditional vesting are harder to reason about than a flat fee. Some creators will choose a simpler venue for that reason alone.
• Adversarial risk. Scoring gates money and will be attacked continuously. Several mitigations in §7 are explicitly partial.
• Concentration risk. If sRALLY concentrates, Attention Pool allocation concentrates with it. Per-market caps reduce but do not remove this.
• Regulatory risk. Token launch venues operate in an unsettled and jurisdictionally inconsistent environment.
• Contract risk. Audits reduce the probability of critical failure. They do not eliminate it.

11Conclusion

The launchpad category optimised the wrong variable for three years. Issuance became free, which was the goal, and then remained the entire product, which was the mistake. What scarcity remains sits on the demand side — attention, durability, and the willingness of anyone to care about a market after the first hour.

Rally's answer is not a better deploy button. It is a fee architecture in which the creator is paid for durability rather than for launching, in which the value that weak launches fail to earn funds the discovery of strong ones, and in which 40% of gross revenue stays inside the system instead of leaving it. A fixed supply of one billion $RALLY governs that budget and captures the revenue it generates.

The next launchpad advantage will come from what happens to the money after the trade.

AParameter appendix