Gloomberg Launchpad
Where we are.
A launch product where you buy a pack rather than a token. Every pack costs the same and what varies is what is inside it. Nothing about any pack is knowable while the sale runs, escrowed backers stand under it at ninety cents on the dollar, and a sale that misses two thirds refunds everyone. The contracts have no owner and cannot be changed.
Why this exists.
Gloombles started as a collection. The launchpad is the pivot: from making one thing to building the thing that launches things. It is a curated launchpad, not an open one, and that word is the product rather than the marketing.
The need it goes after is real and badly served. A token launch today is a race, and the people who win it are the ones with the fastest bots and the earliest information. Everyone else buys the top of a curve from someone who bought the bottom of it a block earlier. Fair launches did not fix that; they moved the advantage to whoever could act quickest.
So the sale is built to make speed worthless. You buy a pack, not a token, and every pack holds the same bet whenever it is bought, because a pack draws what is inside it after the sale closes rather than taking it from a queue. Nothing about any pack can be read while buying is open. There is no early seat to win.
Sustainable means the platform takes 2.5% of a raise, not twenty. It is a business that earns a little from many launches rather than a lot from one, and it takes no token allocation at all. A launchpad that takes a fifth of the raise is taking the operating money the project needs to survive the thing it just launched.
There is no team allocation, because there is nothing left to allocate. The sale commits the supply it needs and burns the rest, including every token the packs never hand out. A partial sale does not leave a stack sitting on the contract and it does not hand one to the project either. Nobody walks away from launch day holding a pile pointed at the people who just bought.
What a project gets instead is 8.5% of the raise, in ETH, at graduation. That is the whole of it. It is a real constraint rather than a boast: a project has to launch at a size that actually funds it, because there is no treasury to sell into the market later if it does not.
What curated means in practice is not settled yet. The code enforces the part that can be enforced: the book cannot be switched off, five backers have to post before a sale opens, and nothing can be changed after deploy. Who gets to launch, who decides, and what the bar is are still being worked out. That is a real open question rather than a detail, and we would rather say so than describe a process we have not run.
And the proof runs on us first. GLOOM is customer number one, launching on the smallest of the five sizes, with the same blind packs, the same escrowed floor and the same refund rule any third party would get. Selling a launch mechanism you have never put your own token through is how most of these end badly. If it does not hold for us, it does not ship to anyone else.
The build.
- 00
The mechanism
Building← we are hereThe rules that make a blind sale with a floor under it possible at all. Eight commits have landed, and this is still where the work is: the most recent error found here was the pool's opening price, which would have priced every pack above what it draws on average. A mechanism is not finished when the code is written. It is finished when it has run on a live chain and behaved.
Nothing reveals until the sale bonds
No batch closes, no seed exists, and the odds view cannot move while anyone can still buy. This one gate is what makes both the floor and the refund safe. Without it, a standing bid on a blind pack fills only when it is wrong.
The claim clock starts at settlement
Settlement is nobody's job, so a holder must never lose allocation to a delay they could not cause.
Packs move while nobody knows anything
A holder can hand packs to anyone, but only before bonding, which is exactly the window in which no seed exists for either side to read.
Below two thirds, everyone refunds
A sale that does not reach its minimum unwinds: every pack refunds at what it cost, the whole token balance is burned, and the whales earn nothing at all.
The bid book
Five anchors post blind before a single pack sells or there is no launch. Escrow cannot be withdrawn, fills are buyer-initiated, and nobody profits from a failed sale.
- 01
Getting it on chain
BuildingWriting contracts and being able to actually deploy them are two different jobs, and the second one is where four findings surfaced, all from reading the contracts against each other rather than against the design. None would have been caught by the tests as they stood, and one meant the stack could not be put on chain at all. The scripts exist and the arithmetic is fixed. What is not done is watching them run.
The opening price was wrong, and the contract caught it
The pool's opening price was still the value from an older, cheaper pack. Against the current bonding target it implied an allocation that would have priced a pack above what it draws on average, so buyers would have been underwater by design. The curve's own deploy assertion refused to construct, which is exactly what that assertion is for. Corrected, and every round number the design assumed came back with it.
The pre-flight check could never pass
It asserted two things about the same value that had drifted apart, so no correct deploy could satisfy both. The two now move together.
Nothing deployed the curve or the book
The two contracts each need the other's address, and the book calls into the curve while being constructed, so only one order is possible, and the book's address has to be predicted before it exists. A wrong prediction would not revert: the sale would graduate normally with the whales' entire fee pool sent nowhere. Now written, and asserted.
The path that pays the whales had no coverage
The book's test suite never called graduation, so the transfer that pays out twelve percent of the raise was never exercised. It is now, end to end, against the real wired pair.
- 02
Before testnet
NextDocumentation and hygiene. None of it is a correctness blocker, but one item is what an operator would actually follow on deploy day.
Rewrite the deploy configuration example
It still asks for a value nothing reads and its deploy order points at a script that no longer exists. An operator following it cannot reach step four. This is the item that bites.
Clear stale comments in the source
Several headers still describe an earlier design: an older pack price, a fee leg that paid two places, a jackpot share that has since moved. The constants are right; the prose around them is not.
Regenerate the specification
The published spec quotes the pre-fix economics. Anything citing the old buyer multiple is wrong.
- 03
Testnet
LaterA full rehearsal on Sepolia, covering both the sale that works and the sale that fails.
Deploy the whole stack in order
Staking, then the beacon, then the flywheel, then the launchpad, then pre-flight against all of it.
Prove the address prediction on a real broadcast
This is the single highest-value thing testnet buys. It is asserted in the script and covered by a test, but it has never run against a live chain, and the failure mode is silent and permanent.
Run a sale end to end
Five anchors post blind, buyers buy, someone exits at ninety percent, the sale bonds, drand settles it, it graduates, the whales claim, the remainder burns.
Run a sale that fails
Below the minimum: everyone refunds, the seller who exited early is made whole, and the book drains to exactly zero.
- 04
Mainnet readiness
LaterThe last irreversible values, and the one thing code cannot solve.
Set the fee-ramp bounds on the day
They are immutable and denominated in dollars, so they must be computed against the real ETH price at deploy time.
Prove every payout address can be paid
A sink that cannot receive ETH bricks graduation permanently on a contract with no owner.
Recruit five anchors
This is the real gate, and it is a people problem rather than a code one. If only four post, nothing happens at all.
Publish the anchor list
Addresses, Terminal ids, holding age. Code enforces the honest path; only disclosure catches the dishonest one.
- 05
Third-party launches
LaterEverything needed for a project that is not us. Not required for launch one.
Make the curve generic
Its constants become deploy parameters, so one contract set serves every rung and every project.
Make the fee split configurable
Only matters for a project taking the full stack rather than the sale alone.
Still to decide.
Four questions that are judgement rather than engineering. Each has a recommendation and none of them is applied yet. One of them is permanent once a launch is live.
Fifteen seats in the book, or twenty
Per launchFifteen for the first launch, widening to twenty at the larger rungs. Returns are set by the fee divided by the escrow, so fewer seats at the same size pays each whale more. What it costs is floor depth.
What the launch reserve may do
Permanent once liveIt funds a seat when the book is short of five anchors. It should earn no fee and should not count toward the five, or it becomes a way for the platform to seed its own launches.
One jackpot winner or several
Per launchOne at the smaller rungs, where a single prize reads as a prize rather than a rebate. Split at the larger ones.
Whether the operations share is enough
Per launchShip as it stands, run three launches, and set the next rung from measured cost rather than a guess.
The honest risk.
Five anchors have to post before a sale can open, and if only four do, nothing happens at all. That is the hardest stop in the whole design, and no amount of code removes it. The launch reserve exists to cover a missing seat, and its rules are the one open decision that cannot be revisited once a launch is live.
This page is a snapshot, not a promise. Dates are not given because they would be invented; the order is real and the status is current as of the date above.