Spirith — Roadmap

Where Spirith would go after the hackathon, if the project shows enough potential and draws enough interest to continue. None of this is committed work; it is the direction, written down so the choice to continue can be made on something concrete.

In plain language

Every website you visit has a name like google.com. That name is not owned, it is rented. A company called a registrar rents it to you for about twelve dollars a year, and a big organisation called ICANN oversees the whole rental system. This system is called DNS, the Domain Name System. When you type a name, DNS looks up which computer it points to, like a phone book. If you stop paying rent, the name goes back on the market and somebody else can take it. Your website, your email and anything that pointed at that name simply break.

ENS, the Ethereum Name Service, is the same idea rebuilt on the Ethereum blockchain. Names end in .eth, like alice.eth. The difference is who runs the phone book. With DNS, the book is kept by companies and one person has to remember to pay the bill. With ENS, the book is a program on the blockchain that anyone can read and nobody can secretly edit. You still pay rent for a .eth name, but the key point is that the rent can be paid by anyone, including another program. That is what Spirith does. It holds money for a name and pays the rent automatically, forever, so no human has to remember.

Now the files. A website's pictures and text normally live on one specific computer, a server, and DNS points at that server. If the server dies or the bill goes unpaid, the files are gone. IPFS works differently. Instead of naming a computer, it names the file by a fingerprint of its contents, called a CID. Ask for that fingerprint and any computer in the world holding a copy can hand it over. The fingerprint never changes, so nobody can swap the file for a fake one.

An NFT is a blockchain record that says "token number 24 belongs to Alice" plus one web address where its picture and description live. That address is the weak spot. Most NFTs point at a plain .com, so the artwork lives or dies on one person's twelve-dollar-a-year bill. A well-built NFT points at the fingerprint instead, and then the only remaining job is making sure at least one computer keeps a copy. That costs cents a year, but somebody still has to pay, and nobody remembers to.

Spirith's idea is the same in every case: find the bill nobody remembers, and set up the standing order. First the .eth name. Then the copy of the files. Last, for the old contracts that can never be changed, the .com itself.

The thesis, in one line

Permanence has an invoice, and nobody set up the standing order. There are three such invoices, and Spirith pays each of them the same way: a per-item endowment, a payment anyone may trigger, two exits for the money, and a runway everyone can see. Phase 1 pays the .eth name and is the finished product for that bill. Phase 2 pays for the copy of the bytes an NFT points at. Phase 3 pays the .com for the contracts that can no longer be changed.

Phase 2 — File permanence

Paying for storage is permissionless and contract-payable, like an ENS renewal and unlike a DNS lease, so the runway mechanic transfers directly. A permanent collection points its tokenURI straight at an ipfs:// fingerprint, which every marketplace fetches itself and which can never expire or be forged. The one remaining bill is keeping a copy pinned, and that is what the Phase 2 endowment pays.

• The bytes are the cheap, solved part. Filecoin Onchain Cloud gives streaming payments that pause when storage proofs stop, at about $2.50 per TiB-month. The endowment pays that stream per collection, publicly, and anyone can top it up.
• A liveness oracle: walk any contract's tokenURI, follow every hop, hash the bytes, and publish whether the NFT is still actually there. Read-only, works on anyone's collection, and produces the strongest visual this project has: a wall of blue-chip names with red scores.
• Mint-funded perpetual care: a fraction of every mint goes into that token's storage fund at the moment of sale. The collector never decides, the artist never sets a reminder, and the collection reaches mint-out already funded. This converts the pitch from charity into infrastructure, and it is the business.

Phase 3 — DNS domains, for legacy contracts

The same endowment, applied to the .com that a decade of immutable tokenURIs still hard-codes. ENS renewal is permissionless and ERC-20-payable; DNS renewal is neither, and a smart contract cannot be a registrant. So Phase 3 is honest about trust rather than pretending to remove it:

• Prepay ten years at registration, endow the renewal, and name the caretaker. One human action every few years, funded and alerted by the endowment. Long term this is a foundation, the way archives have always worked.
• Host nothing. DNSLink turns a domain into two DNS records pointing at content-addressed files; the recurring cost collapses to registration plus zone hosting, which is free.
• Anchor ownership on-chain by importing the DNSSEC-signed domain into ENS and publishing the expected zone hash, so a hostile registrar is detected within the hour even if it cannot be prevented. In ENSv2 the .com can alias the endowed .eth and inherit its records, including the funding status.
• The escape hatch is the product: an on-chain fallback registry mapping (chain, contract) to the canonical frozen-metadata CID. Today a dead tokenURI gives wallets and marketplaces nowhere to look. Nobody has built this; it is small.
• The missing primitive: a renew(domain, years) service behind an HTTP 402 paywall that accepts USDC, calls a registrar API and writes the receipt on-chain. No contract on earth can renew a domain today. This is a standalone project in its own right, and it lives outside SpirithVault, whose two exits never include a registrar API.
• A migration tool that rebuilds a collection's entire URL space as a static tree from the chain alone, so the mirror can be regenerated forever without the artist or their hosting.

First patient: the author's own collect-code / EUCLID collection, immutable, SVG-only, and therefore unusually easy to freeze.

Protocol developments (any phase)

• Switchable yield adapters, per name. Today one IYieldAdapter is an immutable of the vault. Next: a registry of approved adapters, each endowment recording which one holds its excess, and a switchAdapter(label, to) that withdraws the name's position from one venue and deposits it in another in a single transaction, never leaving the earmark. Open design question: who may switch, the patrons by share-weighted vote, the name owner, or a keeper acting on the optimiser's recommendation. The load-bearing rules hold throughout: no admin key over funds, per-name isolation, reserve floor kept liquid during the switch.
• Adapter list and simulation in the dashboard. The name card lists available adapters with venue, rate range and risk notes, and lets a patron simulate the funded-until range under each before choosing; the agent's optimalCadence takes the adapter as an input.
• Liquidity-aware renewal fallback. If the yield venue cannot pay out on renewal day, renew for the longest block the liquid reserve covers instead of failing. A refreshRecord(label) that rewrites spirith.funded-until without a money movement belongs with it.
• What a withdrawal notice protects. Shares under notice stay in the earmark until the notice matures, so a renewal in that 30-day window is paid from them and the patron gets back what is left; observed on 2026-09-12 when a name with its whole endowment under notice was still renewable. That is the rule as specified (notice never blocks a renewal) and it keeps spirith.funded-until truthful, but a patron who has given notice may not expect to fund a six-year block on the way out. Options to weigh: exclude notice shares from optimalDuration so the vault buys only what the remaining earmark affords; renew at most one year while any notice is pending; or leave the rule and say so on the notice form. Not a change for the hackathon build.
• The six-year price for endowed names, paid yearly (ENS partnership). ENSv2 gives 43.75% off for paying six years upfront because the capital is committed; an endowed name commits it just as durably, on-chain and verifiably (the earmark, the 30-day withdrawal notice, the spirith.funded-until record). Proposal to ENS: a rent price oracle that charges the six-year rate on one-year renewals when the payer is a vault whose earmark covers at least six years. The endowment then keeps its whole balance earning while paying the discounted price, which removes the cadence trade-off and lowers the perpetuity threshold for a normal name from ~$129 to ~$112 at 4%. The registrar's oracle is already a replaceable component (setRentPriceOracle), so this is a policy change, not a protocol one.
• Real yield on mainnet. ERC4626Adapter over Aave's USDC token, already proven on a mainnet fork in Phase 1; mainnet deployment waits for ENSv2 mainnet.
• Aave Stable Vaults once their accounting chain reaches Ethereum: a contractually fixed rate is what lets the UI quote a date instead of a range.
• Subregistry endowments: one earmark covering an artist's whole subname tree.
• Endow as a gift, by email, for someone without a wallet.

Partners and funding

Programs that fund what Spirith builds, as they stood on 2026-09-07. Hackathon prizes are in 2; this list is for the company.

Distribution partners rather than funders: mint platforms (Art Blocks, fxhash, Zora, Manifold, Highlight) for mint-funded care, and pinning services (Pinata, Storacha, Filebase) for the bytes. Each is a conversation, not a program.