PONS, by the numbers

The Pons launchpad prints a fee every time somebody trades. Those fees buy PONS on the open market and send it to a dead address. So PONS is not a story, it is a claim on a repurchase, and a repurchase can be measured. This page measures it, every four hours, and says what the token is worth if the arithmetic is all you believe.

What it is worth

move the assumptions, the page follows

A token with no dividend and a permanent buyback is a share repurchase with extra steps. Value it the way you would value one: take the money actually being spent buying the thing back, decide what return you need, decide how fast you think that money dries up, and divide. Everything below the line is measured. Everything above it is a choice, and it is yours to change.

fair market value=B, the annual buybackr + d

What we measured

What it gives

What needs no assumption at all

Required return, r Decay of the flow, d Buyback measured over

The gas cliff

the one term a perpetuity cannot carry

Gas on Robinhood Chain stops being free on 29 September 2026. A single smooth decay rate cannot express a dated step down, and this page has already called that date the whole ballgame, so the model carries it as its own term: the buyback runs at today’s rate until the cliff, takes a cut on the day, and what survives decays for ever after at its own rate. The rate afterwards should be lower than the rate now, because the launches that only ever existed while gas was free are the first to go and whatever still pays a fee is stickier than whatever does not.

Cut to the buyback at the cliff Decay after it, d₂

Price against the model

Daily close on the PONS/WETH pool, against the model run on that day’s own buyback and float. The model line moves when you move the sliders; the last point is yesterday’s close, not the live price.

How it compares

launchpads and token-holder flows · swipe for every column

A multiple is worth nothing on its own. So here is the same arithmetic applied to the rest of the market: what each launchpad earns, how much of that reaches its token, and what the market charges for it. The column that matters is cap ÷ buyback. It is years of buyback, it is what a buyer is paying, and it is the same number as 1 / (r + d).

The platform

by UTC day, from Dune

Launches and graduations

since 4 August, when the current curve started

A Pons coin starts on a bonding curve and graduates when it fills that curve and its pool is opened, seeded with a fixed 285,714,285.71 tokens. Almost none of them make it. The rate is the cleanest measure of whether a launchpad is producing anything or just producing coins.

Launched against graduated

Coins deployed through the launchpad each day, and how many of them opened a pool.

Graduation rate

Two honest readings. The solid line is graduations on the day against launches on the day. The dashed line follows each day’s own coins and asks how many have graduated since, so the last few days are still filling in.

Wallets

every wallet on Robinhood Chain, not only Pons

The share of a day’s wallets that had never sent a transaction before is the number to watch. A healthy chain converts new wallets into returning ones. A subsidised one keeps minting first-timers who never come back.

The buyback

read from Robinhood Chain

Every fifteen minutes a keeper calls the splitter, it wraps a fixed clip of ETH, swaps it for PONS on the 1% pool and sends the tokens to 0x…dEaD. The clip size is set by an operator, not by revenue, so a quiet day is a decision, not an accident.

Every day on file

most recent first, * is a day its sources had not run past · swipe the table for every column

Method, and where every number comes from

The model

PONS pays no dividend. Its only cash flow is the platform’s own bid: protocol fees are spent buying PONS on the PONS/WETH pool and the tokens are burned. That is a share repurchase, so the token is worth the present value of every future buyback, spread over the float that will still be alive to receive it. With a buyback of B a year, a required return r and a flow that decays at d a year, that present value is B / (r + d). Divide by the float and you have a price.

freefloat’s published defaults are r = 40% and d = 40% a year, over a 7-day buyback window. Forty per cent is what we think a crypto-native buyer should demand from an asset whose revenue can go to nothing in a quarter. Forty per cent of decay is a launchpad assumption, not a Pons one: attention-driven fee businesses have never held a run-rate for a year. Both are guesses and the sliders exist so you can use yours. The window is a rolling seven days ending yesterday, not the last seven whole days: in a market moving this fast the freshest week is the only one that describes the bid running today. Taking the last seven complete days instead slides the window back past the 4 September step-up in the clip and reads the current bid about a fifth light. A day whose source stopped short of midnight counts for the share of the day it covers, not as a whole one.

The window never reaches back past 4 September 2026, the day the buyback clip went from about $85,000 a day to over a million. That is a regime break, not variance, and averaging across it produces a run-rate for a business that no longer exists. On 9 September a plain trailing seven days still caught 3 September, one day of the seven running at 7% of the current bid, and it cost $0.11 of published fair value. So the window is six days today and says so beside the slider. From 11 September seven trailing days fit inside the new regime on their own and the rule stops binding. It applies to the chart the same way, which is why the model line before 4 September still reads: a window entirely inside the old regime is left alone.

The cliff term. A single smooth d cannot express a dated discontinuity, and the largest risk on this page is exactly that: gas stops being free on 29 September 2026. So the model runs in two stages. For the T years left before the cliff the buyback runs at today’s rate, decaying at d. At the cliff it takes a cut. What survives decays at d₂ for ever:

PV = B (1 − e^−(r+d)T) / (r + d) + B (1 − cut) e^−(r+d)T / (r + d₂)

The first term is not an extra nineteen days bolted on the front. It replaces the first T years of the perpetuity and is discounted at the same rate, which is why it carries the same r + d denominator. Set the cut to zero and d₂ to d and it collapses back to B / (r + d) to six decimal places. That is the check that this is the same model with one more term in it, not a different model.

Why the page leads with the cut rather than with a price. Every cell in that grid is a guess about an event that has not happened here or anywhere else, and a single number picked out of it would be a forecast dressed as a measurement. So the arithmetic is run backwards instead: given the market value, B, and a decay rate for the survivor, what cut at the cliff is today’s price already paying for? That is a reading of the price, not a view of ours, and it is the number at the top of the page.

What the model is still not. It has no terminal value beyond the perpetuity, no premium for the tokens locked outside the float, no discount for the operator’s power to shrink the clip, and no view on the price of ETH. Nothing in the data constrains d₂: it moves the answer by 10 to 15 cents a cell and it is a guess. And the whole grid assumes the buyback keeps paying out at its current share of a smaller revenue. It is a manual, non-immutable, single-key process, and a payout cut landing at the same time as the cliff is not modelled here or anywhere else.

The multiple, in one line

Every valuation is a multiple in the end, and ours is 1 / (r + d). At r = 40% and d = 40% that is 1.25: freefloat will pay $1.25 for a dollar a year of buyback, and no more. That one number does all the work on this page. The buyback is measured, the float is measured, and the only judgement left is how many years of the flow you are willing to pay for. Anyone who thinks the Pons bid lasts longer than we do should raise the multiple rather than argue with the arithmetic.

Why it is so low. The decay does it. At 40% a year the flow adds up to two and a half years of today’s buyback before it fades to nothing, and discounting what is left at 40% halves that again, to 1.25 years. Take the decay out altogether and the multiple is 2.5. Assume the bid is dead inside a year and it is 0.71. The ladder in the valuation section prints the whole range, so a reader can find their own number without doing any sums.

The comparison table applies exactly this multiple to every other launchpad: market value divided by the money reaching the token each year. Read backwards, a multiple is a forecast, and the decay priced column does that arithmetic for you. It is what the flow would have to do, at r = 40%, to justify today’s price. A negative number is a market betting the money grows. A positive one is a market betting it shrinks. PONS sits at the far positive end of that column, which is the entire disagreement in one cell.

How the comparison is built

Launchpad revenue is what the launchpad itself earns, from DefiLlama’s fee adapters, summed over the last 30 days. Only protocols DefiLlama files under Launchpad are counted, so an exchange’s trading business does not flatter its launchpad.
To the token is DefiLlama’s holders-revenue line for everything that team runs, over the last 7 days, annualised: the same 7-day window this page uses for PONS. It is the whole product line rather than the launchpad alone because that is what the token a reader would have to buy actually receives. For a pure-play launchpad, Pons and pump.fun included, the two are the same business.
Market cap is CoinGecko’s circulating figure. For PONS the table therefore carries CoinGecko’s cap, not the float we read off the chain ourselves; the two are within a per cent and the rest of this page uses ours.
A dash in the holders column means DefiLlama books nothing there. Usually that is a launchpad with no token, or one whose fees never reach a token. It is not a measured zero and should not be read as one.
Holders revenue is not always a buyback. Some of these protocols distribute to stakers instead, which is a different thing for a holder who does not stake. Pons is the simple case: the fees buy the token on one pool and burn it, and the burns are counted on chain.
Hyperliquid is not a launchpad. It sits at the bottom of the table as the benchmark every buyback token gets compared with, and it is labelled as such.

Graduation

A coin has graduated when the pool manager at 0x7ed5…ec7e initialises its pool, the event whose first topic is the token itself. That was checked on 9 September 2026 against four graduations the freefloat runner had recorded independently, and every one of them seeds the pool with the same 285,714,285.71 tokens. The denominator is the same set of launches the “coins launched” figure counts, so the rate is consistent with the rest of the page.

Why the series starts on 4 August. Before that, Pons coins were deployed straight into a Uniswap V3 pool: there was no curve to fill and nothing to graduate, so a graduation count for July would read 100% and mean nothing. The current bonding curve and its pool manager begin in early August, and that is where the chart begins.

The buyback figure

The daily buyback is the market value, at that day’s close, of the PONS retired that day. One definition for the whole series, including the days before the splitter’s swaps were readable. The splitter’s own measured spend is carried alongside as a cross-check rather than mixed in; over the current window the two agree closely, and the page prints the gap so you can see when they stop agreeing. The launch burn of 13 to 19 July was not a buyback and is kept off the buyback chart, though it stays in the supply retired to date.

Sources

Turnover, protocol fees, creator earnings, coins launched, transactions and the hourly burn: the public adam_tehc/pons and the-robinhood-trenches Dune queries, read as their last cached run.
Active and new wallets, and the pool price and the splitter’s spend: two Dune queries freefloat wrote and runs itself. They are public: wallets, pool.
Supply, tokens burned, the pool price and the splitter’s balance: read live from the Robinhood Chain RPC at build time. Token 0x39dB…4571, splitter 0x5795…c324, pool 0x10cc…26ba.
Graduations: a third freefloat query, 8655495, which also produces the launch count the rest of the page uses.
A second read on fees: DefiLlama’s Pons adapter, plotted as its own line in the platform chart and never mixed into the Dune columns. It lands same-day where Dune’s revenue query lands a day later, and the two run a few per cent apart; if they ever diverge badly, that is worth knowing and the page will show it.
The ETH price: Hyperliquid’s public info endpoint for the live figure, the same one the Hyperliquid page reads; for a past day, the volume-weighted price of the WETH that actually traded on the pool that day, from Dune’s hourly price table. On 10 September 2026 that day figure was corrected: it had been dividing priced volume by all volume, so any hour Dune had not priced yet dragged the day down, and 9 September published at $1,592 against a real $2,499. Only the volume that has a price behind it is weighted now, and a day with no price at all shows a dash instead of a number.

Query runs behind this build:

Caveats we are not going to bury

The buyback has been running at its current clip since 4 September 2026. Every run-rate on this page is a handful of days of evidence. And the clip is a dial an operator turns by hand: it moved between 10 and 56 WETH an hour inside a single day earlier this month. No window models that, and it is a larger source of error than the choice between six days, seven or thirty.
A day is only whole when the source behind it ran after the day ended. The public Dune queries refresh at about 23:20 UTC, so for part of the day “yesterday” is still missing its last minutes. Such a day is marked with an asterisk in the table, and in the run-rate it counts for the share of itself it actually covers rather than as a whole day. Before 10 September 2026 it was counted as a whole day, which read the buyback about 4% light. The weighting is per source: the buyback is weighted by the burn feed’s own cut-off, not by whichever query happened to run earliest.
Dune’s revenue query reads a dataset somebody uploads by hand. On 14 September 2026 that upload stopped, and four days later it had still not resumed, so the fee columns sat empty while the rest of the market carried on quoting DefiLlama. Where Dune has published nothing, the table now shows DefiLlama’s own figure marked †, and launches measured by us off the chain marked •. Neither is ever written into a Dune total or into the run rate, and the marks say which is which. DefiLlama runs consistently below Dune, 5 to 9% on protocol fees and 2 to 6% on creator earnings over the last week, so the two are close but not the same measurement. Creator earnings from DefiLlama are its fees less its revenue, a subtraction of two published series rather than a figure it publishes directly.
Graduation is measured from the pool being opened, not from any claim by the launchpad. A coin that graduates and is abandoned an hour later still counts.
Protocol revenue on Dune runs well ahead of the buyback that reaches the pool. The page prints that ratio and does not pretend to know why: it could be a backlog, a different definition of revenue, or fees that were never destined for the buyback.
The float is total supply less the dead address. It makes no allowance for tokens that are locked, vesting or held by the team, so the fair value per token is, if anything, generous.
Wallet counts are chain-wide and sybil-heavy by nature. The sybil contracts filtered out are adam_tehc’s list, not ours.
Every figure is free of gas costs until 29 September 2026. That is the whole ballgame.

Not advice, not a recommendation, nothing for sale. freefloat holds a PONS position on Hyperliquid, disclosed and updated on the Hyperliquid page. A model is not a forecast and a buyback is not a promise.