What to use instead of Drift BET

The product

BET launched on 19 August 2024, announced on Drift's own updates page under the title "Introducing BET: The First Capital-Efficient Prediction Market on Solana". The acronym was explained in the post: Bullish on Everything. That URL now returns 404, and so does the follow-up post about the Breakpoint market; neither appears in the updates index today.

It was never a separate venue. A BET market was a Drift perpetual market carrying the contract type Prediction, with its price bounded between 0 and 1 and settled at one end or the other. Same program, same order book, same matching, same margin engine — and, as Drift's documentation put it, "prediction markets have the same fee structure as perp markets". That reuse was the whole product: the announcement's selling points were thirty-odd Solana tokens accepted as collateral rather than USDC alone, lending interest accruing on that collateral while a position was open, and a single margin account in which a long on an outcome and a short on Bitcoin offset each other.

Two things about it were not like the rest of this category, and both are worth knowing before you decide what replaces it.

Nothing about it was permissionless. Drift's security council — an elected multisig under Realms governance — initialized every market, on request, and the same council resolved every market by setting the oracle to a fixed 0 or 1 and an expiry date, after which the market went reduce-only. Permissionless market creation was promised in the launch post for "later this year", trailed again for early 2026, and never shipped: in the last published version of the code, initialize_prediction_market is still an admin instruction.

Settlement was not fully insulated. The documentation said plainly that where a market ran a shortfall its fee pool and insurance could not cover, the shortfall would be socialised into the settlement price — so a market resolved to 1 could settle below 1. That is a risk a fully collateralised YES/NO share does not carry, and it is the one thing every replacement in this list improves on.

What ended it was not a product decision. On 1 April 2026 Drift was drained. The protocol's own recovery update puts the stolen assets at $295,706,374.93 and timestamps the snapshot of affected positions at 18:31:47 UTC that day. The recovery package it describes is a transferable SPL recovery token worth one dollar of verified loss, funded by exchange revenue, up to $127.5 million committed by Tether and $20 million by other partners, with two independent audits required before relaunch. On 1 July 2026 the protocol relaunched under a new name, with USDT rather than USDC as the quote asset.

This is what we found on 19 September 2026. docs.drift.trade and app.drift.trade both redirect to the successor's domains. docs.drift.trade/prediction-markets/prediction-markets-intro redirects to a URL that answers HTTP 500. The successor's documentation runs to 39 pages and mentions prediction markets exactly once, on the migration page for developers, where it records that initialize_prediction_market is removed and ContractType.PREDICTION is now DEPRECATED_PREDICTION. The protocol repository has been renamed and archived read-only since 8 July 2026, still under Apache-2.0, so the BET code is readable and unrunnable at the same time.

None of that is an announcement. BET was never given one, which is why a review we read the same day still describes it as a working venue, records itself as last reviewed on 28 June 2026, and quotes an entry-tier taker fee of 0.0350 percent from a fee schedule dated 23 June 2026. Those were Drift's perpetual tiers, which BET shared; the successor's entry tier is 0.060 percent taker and a 0.0025 percent maker rebate, denominated in USDT. Both numbers are now facts about a perpetuals exchange, and neither describes anything you can take a position on.

What to use instead

BET was three products wearing one name — a self-custodied venue, a margin engine and an SDK — and only the first has a clean replacement.

If what you had was self-custody and no identity check, start with Limitless. It is the closest structural match left: binary YES/NO shares fully collateralised in USDC, matched on a central limit order book, wallet-only access, no KYC, and the same bar on US traders that Drift's terms carried. Makers pay nothing there; takers pay on a curve that peaks at the 50-cent midpoint. Myriad is the same accessibility with a different mechanism — an AMM instead of a book, which means a thin market still fills without somebody sitting on the other side, at 0.5% to the taker.

If what you had was the question set and the depth, Polymarket is where the volume is, across politics, sports, macro, crypto, science, culture and business, and its resolution runs through the UMA optimistic oracle rather than through an operator's multisig — which, given how BET ended, is not a small difference. It costs you an identity check BET never asked for, and collateral is USDC only.

If what you miss is the range, Futuur runs the broadest subject list in this catalogue and runs every question twice, in USDC or USDT and in play money, with no identity check. The mechanism is an LS-LMSR market maker, so again a curve rather than a book.

If it was your code that died, the Polymarket CLOB API is the direct target: books and prices read without a key, orders signed from a Polygon wallet. If you would rather not be in this position a third time, CCXT puts a unified, MIT-licensed API over seven event venues, so swapping a venue is a configuration change instead of a rewrite.

If what you had was the margin account, nothing here replaces it. No venue in this catalogue lends against your collateral while an event position is open, and none nets that position against a perpetual short. That combination was BET's entire pitch, and it is gone with the exchange it was built inside.

What you give up

One account, one collateral pool, one hedge. BET's reason to exist was that the same balance could back an outcome position, earn lending interest and margin a Bitcoin short in the same breath. Every venue above wants funded, idle stablecoin collateral in a wallet of its own. If you were running the prediction leg of a structured position on Drift, you are now running two accounts on two chains and reconciling them yourself.

Thirty-odd collateral tokens. Polymarket and Limitless take USDC. Futuur takes USDC or USDT. Holding SOL, an LST or a yield-bearing stable and posting it directly is not an option anywhere here — you sell first, and you eat the spread and the tax event on the way.

The fee level. BET charged basis points because it charged what a perpetual charged: 3.5 bps of notional at the entry tier. Polymarket and Limitless charge a coefficient on share count that peaks at the 50-cent midpoint, and Myriad and Futuur charge 0.5% and 1.25%. These are not measured the same way and a like-for-like comparison depends on where in the price range you trade — but the order of magnitude is the point, and it is not in your favour.

The order-type surface. Because BET was a perpetual market, it inherited limit, stop and reduce-only orders and the keeper network that filled them. Limitless and Polymarket are order books with a much plainer order set; Myriad and Futuur are automated market makers, where "stop loss" is something you implement yourself or do without.

What you get back, and it is not nothing. Every venue above fully collateralises the share you hold, so none of them can socialise a shortfall into your settlement price the way BET's documentation reserved the right to. And on Polymarket the outcome is proposed to an oracle with a challenge window and a token-holder vote behind it, rather than written by the same class of administrative key whose compromise ended the protocol.

Migration notes

Your claim is a snapshot, and the snapshot has a time on it. Drift's recovery plan sets it at 18:31:47 UTC on 1 April 2026, with assets valued at 16:06:00 UTC prices to keep the attack itself out of the valuation. The recovery token is transferable, represents one dollar of verified loss, and is a claim on a pool funded out of the relaunched exchange's revenue. It is not a refund and it is not dated. Read the published methodology against your own wallet rather than against the headline figure.

The quote asset changed, and the successor's own documentation flags it as the trap. Velocity settles in USDT. Its migration page says a desk that funds accounts with USDC, as it did on Drift, is depositing the wrong token. Nothing about that is specific to prediction markets, but it is the first thing that will bite anyone who follows a redirect and starts trading.

Old packages still install. @drift-labs/sdk was last published on 18 May 2026 and is still resolvable; the successor ships a separate package under a new name, with a different program id, a different client class and a different data API host. An integration that pins the old package will build, and will point at a program that no longer offers the instruction you are calling. Make the absence of a prediction market a startup error rather than an empty market list.

Your bookmark resolves, and that is the problem. app.drift.trade/BET redirects into the successor's application and answers 200 with an app shell. There is no banner, no 404 and no notice — which is exactly how a reader ends up believing the product is merely having a quiet day.

The code outlived the markets. The protocol repository is archived rather than deleted, under Apache-2.0, and still contains the prediction-market accounting: the 0-to-1 price bounds, the admin-only initialization, the settlement path and the shortfall socialisation. If you need to reconstruct how a specific position was valued, that is where the answer is, and it will not move again.

Jurisdiction is the one thing we could not re-check. Drift's launch post said only that BET was "available in all countries where Drift can be accessed", and Drift's own terms of use no longer resolve at their old address; the successor's terms and disclaimer pages return 200 but served no text to us without JavaScript on 19 September 2026. So we are not restating a restricted-country list here. Limitless and Myriad both bar US traders in their terms; Polymarket's international platform does the same while a separate, CFTC-designated venue serves the US. Check the one you pick against its own terms before you fund it.