# What to use instead of Zeitgeist

The site, the app, the docs and the terms all load. There are no markets, no indexer and no working RPC. What we checked on 19 September 2026, and where to go.

*https://predictionmarkets.tools/alternatives/zeitgeist*

## The subject

| Field | Value |
| --- | --- |
| Product | Zeitgeist |
| Status | retired |
| Stopped serving | 2026-08-02 |
| Website | https://zeitgeist.pm |
| Category | Prediction Market Venues |

A GPL-licensed Substrate layer 1, deployed as a Polkadot parachain, whose runtime was the prediction market — markets, an AMM, an order book, oracle reporting and a ZTG-staked court of jurors, all as extrinsics.

## Replacements

*Ordered by how much of the original’s job each one covers, closest first. Paid placement does not affect this order.*

1. [Limitless](https://predictionmarkets.tools/tools/limitless.md) — Wallet-only, no identity check, users can open their own markets and take a creator fee, and shares are fully collateralised in USDC on an order book.
2. [Polymarket](https://predictionmarkets.tools/tools/polymarket.md) — The only venue here that resolves through an on-chain oracle with a challenge window and a token-holder vote — the closest thing left to Zeitgeist's court.
3. [Myriad](https://predictionmarkets.tools/tools/myriad.md) — An AMM across four chains, settled in stablecoins, wallet-only and the cheapest taker fee here — Zeitgeist's default scoring rule, on chains that run.
4. [Futuur](https://predictionmarkets.tools/tools/futuur.md) — The broadest subject list in this catalogue, in USDC or USDT with no identity check, and a resolution process that names its source for every market.
5. [Manifold](https://predictionmarkets.tools/tools/manifold.md) — Anyone opens a question and the creator resolves it, on MIT-licensed code — Zeitgeist's permissionless creation and readable rules, minus the collateral.

## The product

Zeitgeist was not an application on somebody else's chain. It was a purpose-built Substrate layer
1, deployed as a Polkadot parachain, whose runtime pallets *were* the prediction market: creating a
market, trading it, reporting the outcome and disputing that report were all extrinsics. ZTG was
the collateral, paid the bonds, and was what jurors locked to vote. The code is GPL-3.0-or-later
and public, which made it the one venue in this category whose rules could be read rather than
taken on trust.

The design was unusually explicit about mechanism, and the mechanism is why the page is worth
writing rather than deleting. A market named a scoring rule and the chain implemented three: an
AMM with liquidity pools anyone could join or exit, a central limit order book where makers posted
an implied price by naming two amounts to swap, and a parimutuel pool that simply divided the stake
among holders of the outcome that happened. Markets could be categorical or scalar, combinatorial
positions across several markets were part of the runtime, and futarchy — deciding a policy by the
price of conditional markets — was the stated reason the chain existed.

Resolution ran backwards from how most of this category does it. Every market named an oracle
account that posted a bond and reported the outcome after close. Anyone could dispute that report,
which sent the market to a **court**: jurors and delegators locked ZTG, selection was weighted by
the amount locked, votes were submitted as sealed hashes and then revealed, and the plurality won.
A losing side could appeal up to three times; at the fourth, a global dispute put the outcome to
every ZTG holder. Creating a market cost a bond rather than a fee — 200 ZTG for an advised market
an advisory committee could reject, 1,000 ZTG to skip the review — and reporting as a market's
oracle cost another 200 ZTG bond. Trading fees belonged to the market, not the protocol: a swap fee
to the liquidity providers capped by the runtime at 10%, an optional market-creator fee capped at
1%, and a 0.01% fee to exit a pool.

**Here is what we checked on 19 September 2026, and it is the reason this page exists.**

The marketing site, the app and the documentation all return HTTP 200. So do the terms. The
listing at `app.zeitgeist.pm/markets` returns 200 as well, and its server-rendered payload still
carries CMS metadata for a couple of hundred historic market ids — which is why the page looks
populated for a moment. The markets themselves are fetched in the browser, and this is where it
stops. The production build names exactly one live indexer, `processor.rpc-0.zeitgeist.pm/graphql`,
labelled "Polkadot (Live)". That host resolves and answers, but what it answers with is a parked
advertising page titled "processor.rpc-0.zeitgeist.pm - PDF Documents", carrying an `<h1>` reading
"Latest News" and a meta description about the Loeb Classical Library. A POST of a GraphQL query
returns 405; a GET returns 404. The staging indexer times out. The app's empty-state string, which
is what the listing therefore renders, is "No results!".

The chain is no better. The same build names three production RPC endpoints. Two are on
`zeitgeist.api.onfinality.io`, which does not resolve at all. The third,
`wss://main.rpc.zeitgeist.pm/ws`, resolves to an address that fails the TLS handshake with an
unrecognised-name error. `rpc-0.zeitgeist.pm` times out and `bsr.zeitgeist.pm` refuses the
connection on port 443. Those are every endpoint the project points its own users at.

The repository is alive in the sense that it is not archived, and dead in the sense that nothing
has happened in it. The last commit on the default branch is dated 1 December 2025, and the last
runtime tag, `v0.6.2`, is dated 4 December 2025 — a tag rather than a published release, which is
itself a small sign of a process that stopped mid-step. There is no shutdown notice in the
repository, on the site, or anywhere we searched.

**The one date we can put on it is not the project's.** Two independent Polkadot trackers —
parachains.info and Subscan's relay-chain parachain page — both record para ID 2092 as a
**parathread** rather than an active parachain, and parachains.info records its last slot lease as
running from 29 September 2024 to 2 August 2026 and then expiring. A parachain without a lease is
not scheduled by the relay chain and stops producing blocks unless somebody buys coretime for it on
demand. That is the date in this page's header, and it is a secondary source: we could not read the
chain's last block ourselves, because doing that needs exactly the RPC endpoint that does not
answer. Treat 2 August 2026 as the best-evidenced end and not as an announcement, because there
was none.

One more thing a holder will want and will not find: on 19 September 2026 CoinGecko had no entry
for Zeitgeist at all — its coin endpoint answered `coin not found` and a search for the name
returned nothing. That is a fact about an aggregator rather than about the chain, but it is where
most people would go to look for a price, and there is not one there.

## What to use instead

Zeitgeist was four things at once — a wallet-only venue, a market factory, a mechanism-design
project and a dispute system — and they do not all land in the same place.

**If what you had was a wallet and no identity check**, [Limitless](https://predictionmarkets.tools/tools/limitless) is the
closest working venue. Shares are fully collateralised in USDC on Base, matched on a central limit
order book rather than a curve, access is a wallet and nothing else, and — the part that matters
most for a Zeitgeist user — anyone can create a market there and take a creator fee on it, which is
the one feature the rest of this list does not offer with real collateral behind it. Most markets
resolve automatically against Pyth, short-cadence crypto against a Chainlink TWAP, and everything
else by the team within 24 to 72 hours. [Myriad](https://predictionmarkets.tools/tools/myriad) is the same accessibility with
Zeitgeist's default mechanism: an AMM, across four chains, settled in stablecoins, at 0.5% to the
taker.

**If what you valued was the court**, [Polymarket](https://predictionmarkets.tools/tools/polymarket) is the only pick here that
keeps any of it. An outcome is proposed to the UMA optimistic oracle with a bond behind it and a
two-hour challenge window, and a proposal disputed twice escalates to a token-holder vote. It is
not a commit-reveal jury with three appeals, it is not weighted by a stake you chose to lock, and
the international platform now asks for an identity check Zeitgeist never did. It is still the only
resolution process in this catalogue that a third party can enter.

**If what you had was a habit of opening your own questions**, [Manifold](https://predictionmarkets.tools/tools/manifold) is the
honest answer and it comes with an honest caveat. Anyone creates a question, the creator resolves
it, the platform publishes its own Brier score, and the code is MIT-licensed and readable — which
is a fair share of what made Zeitgeist interesting. The currency is mana and it buys nothing. If
you were creating markets to learn something, that is a small loss; if you were creating them to
earn the creator fee, it is total.

**If you simply want a venue that will still be there next year**, [Futuur](https://predictionmarkets.tools/tools/futuur) runs the
broadest subject list in this catalogue, in USDC or USDT alongside a play-money book on the same
questions, with no identity check and an editorial team that names the public source it resolves
each market against.

**And if what you actually wanted was a chain you could read**, nothing in this catalogue replaces
it. The venues above are hosted products with terms of service. Their matching, their fee logic and
their resolution rules are described on documentation pages rather than compiled into something you
can diff.

## What you give up

**A market you can open without asking.** Zeitgeist let anyone create a permissionless market for a
1,000 ZTG bond and said in its own documentation that it would not take a badly written one down —
your money, your problem. Limitless has creator-created markets with a creator fee and is the
nearest equivalent; Manifold is unlimited and pays in play money. Everywhere else in this
catalogue, an operator decides which questions exist, and a question nobody at the venue wants is a
question that does not exist.

**Resolution that somebody other than the operator can overturn.** This is the real loss, and it is
worth stating precisely. On Myriad and Futuur the team resolves and there is no challenge
mechanism. On Limitless an oracle feed resolves the automated markets and the team resolves the
rest, with no on-chain challenge and no juror vote. Polymarket's UMA path is a genuine dispute
process but it is a bonded proposal and a token-holder vote, not a stake-weighted jury with three
rounds of appeal and a whole-token-holder backstop. Nothing you can trade today implements what
Zeitgeist's runtime implemented.

**Scalar and combinatorial positions.** Every venue above is binary YES/NO. Scalar markets — where
the payout is a number in a range rather than one of two outcomes — and positions built across
several markets at once were runtime features here, and they are simply absent from the
replacements. If you were using them, you are not switching venues, you are changing what you can
express.

**Code instead of a terms page.** GPL-3.0-or-later meant the swap fee cap, the bond sizes, the
juror selection and the appeal count were constants you could look up and, in one case this
catalogue documented, catch disagreeing with the documentation. Four of the five picks above are
closed products; the one that is not pays in play money. What each of them charges and how each of
them resolves is a claim on a page, and the page can be edited.

**What you get back is that it settles.** Everything above is currently matching orders, taking
deposits and paying out. That sounds like a low bar until you have spent an afternoon on a site
that loads perfectly and cannot fill an order.

## Migration notes

**There is no announcement to plan against, so plan against the endpoints.** If you built anything
that read Zeitgeist — a dashboard, a bot, a research notebook — it is not failing loudly. The
GraphQL host still resolves and still returns HTTP 200, with an advertising page rather than data.
Any client that checks a status code and not a response shape is currently recording silence as
health. Make a schema mismatch a startup error before it prices something at zero.

**Getting a balance out needs a node, and there is no hosted one.** Every published endpoint is
down, so the only route we can see is running the node yourself from the archived source against
whatever relay-chain state exists, and then finding a way to move ZTG somewhere that will take it.
We have not tested this and we are not going to claim it works. If you hold a meaningful balance,
treat that as a project rather than an afternoon, and start by asking in the project's own channels
whether anyone is still running an archive node.

**An unresolved position has no resolver.** The court needs jurors, jurors need to lock ZTG, and
locking ZTG needs the chain. A market that closed without a reported outcome has no path to one
while this is true. There is no operator to write to, because the design's entire point was that
there was no operator to write to.

**The code is still readable and is the only reliable documentation.** The repository is public,
GPL-3.0-or-later, and holds the runtime constants — the swap-fee and creator-fee caps, the advisory
and permissionless bonds, the oracle bond, the appeal count. If you need to reconstruct how a
specific market should have settled, that is where the answer is. It has not moved since December
2025 and it is not going to.

**Nothing here transfers.** There is no migration path, no partner venue, no snapshot and no
recovery token. The replacements above are new accounts on new chains funded with different
collateral, and the contracts are not the same contracts. Anyone telling you otherwise is selling
you the replacement.

**Re-check before you act on this page.** A chain with public code and a live domain can come back:
somebody buys coretime, somebody redeploys an indexer, and the app in front of it starts working
again without a single line changing. Everything above describes 19 September 2026. If you are
reading this much later, spend two minutes on the market listing before you accept it.

## FAQ

### Is Zeitgeist shut down?

Nobody has said so. There is no notice on the site or the app, the repository is not archived, and we found no announcement anywhere. What we can report is that on 19 September 2026 the app listed no markets, the GraphQL indexer its production build queries served a parked advertising page, none of the three production RPC endpoints named in that build answered, and the Polkadot slot lease that scheduled the chain had expired on 2 August 2026 according to two independent trackers.

### Can I get my ZTG or my position out?

Not through anything the project publishes. Moving a balance on a Substrate chain needs an RPC node, and the endpoints named in the app's own configuration did not answer — one host does not resolve at all, one failed the TLS handshake, one timed out. Running your own node against the archived code is the only avenue we can see, and we have not tested whether it syncs.

### Why does the app look like it is working?

Because the parts that break quietly are not the parts you see. The pages are served from a static build and a hosted CMS, both of which are still paid for and still respond. Market data is fetched in the browser from a GraphQL indexer, and the host that indexer ran on now serves unrelated search-engine spam. The result is a complete-looking application whose market listing renders its empty state.

### Was Zeitgeist ever available in the United States?

No. The terms name Equipoise Corp. d/b/a Zeitgeist, registered under the laws of Panama, state that it has not sought licensing with any regulatory authority anywhere, and state that it does not accept clients from the United States and other jurisdictions where regulation prohibits it. No list beyond the United States is published. The page was still live on 19 September 2026 and carried the line "Last updated 2025".

### What happened to the court?

Nothing was announced about it either. The dispute mechanism — jurors and delegators locking ZTG, voting as sealed hashes and then revealing, plurality winning, up to three appeals before every ZTG holder votes — is still in the published runtime and is still the most interesting thing anyone in this category has built. It requires a chain that is producing blocks, and we could not find one.