On November 3, SEC staff called Canton Coin a digital commodity in a no-action letter from the Division of Corporation Finance. It is a staff position, not a Commission ruling: staff will not recommend enforcement when validators and apps mint CC without registration. A day later, the first trades of a tokenized US stock on Canton ran through Tradecraft. Canton is usually pitched as the chain for Wall Street. Here are eight apps that already run on it: what each one does, who funded it and what the numbers say. Six are trading venues, one is a money market and one brings bitcoin to the network.

The backdrop: CC trades at $0.10 with a $4.1B market cap (CoinGecko), about half its February high. Canton DEXs did $817M in volume over the last 30 days (DefiLlama), and most of it went through one venue.

1/ Temple (Institutional Order Book)

Temple runs a private, non-custodial 24/7 order book exchange for institutions. It went live on January 8 for approved institutional counterparties, starting with the CC/USDCx pair. It leads Canton by a wide margin: $698M in volume over the last 30 days, about 85% of all Canton DEX volume, and $1.87B since December 2025 (DefiLlama).

Funding: a $5M seed in October 2025, then a strategic round led by SBI Group in May 2026, amount not disclosed.

2/ Rocky Exchange (Spot and Perps)

Rocky Exchange offers spot and perpetual trading on Canton. Orders match on an off-chain order book and settle onchain, and users connect through Rocky's own Canton wallet extension. It holds the biggest all-time spot number on the network, $2.09B (DefiLlama). Recent activity is quieter: $55.0M over the last 30 days, behind Temple and Cantex.

Funding: not disclosed.

3/ Cantex (AMM Exchange)

Cantex is a self-custodial AMM exchange built for institutions. Users swap Canton-native and bridged assets against liquidity pools, with atomic onchain settlement and private execution. It did $63.9M over the last 30 days, second on Canton after Temple, and $404M all-time (DefiLlama).

Funding: not disclosed.

4/ Cancore (Cross-Chain Trading)

Cancore connects Canton to the liquidity that already sits on public chains. It trades Canton assets directly against Ethereum, Arbitrum, BNB Chain, Robinhood Chain and Tron, with atomic, self-custodial settlement and no bridge in the middle. Live on Canton mainnet since April, it reports $86M+ settled across 160,000+ trades.

Funding: a $7.2M seed announced on October 27, the freshest round on this list.

5/ Tradecraft (AMM and Liquidity Layer)

Tradecraft is an AMM and liquidity layer: users swap Canton-native and bridged assets, or provide liquidity and earn trading fees. Its TVL is small, $1.7M (DefiLlama), but it carried the network's biggest moment this week. On November 4 the first trades of tokenized Canton Strategic Holdings stock (Nasdaq: CNTN) ran through a permissioned Tradecraft pool on TSV, a venue for tokenized US stocks: Five North sold, DRW Cumberland bought and Flowdesk provided liquidity, under the SEC's new Innovation Exemption (press release).

Funding: not disclosed.

6/ Ekiden (Derivatives)

Ekiden is a derivatives exchange for trading firms, funds and market makers. It combines an order book with RFQ and settles onchain, and calls itself Canton's first order book derivatives exchange. Its first testnet week in June drew 1,502 active users and $8.9M in volume. Mainnet opened in September (Canton Foundation), still invite-only, and the team counted $5.87M in volume over its first 20 days.

Funding: a $2M seed in May 2026.

7/ Alpend (Private Money Market)

Alpend is a lending and borrowing protocol for institutions on Canton. Users supply assets to earn yield or borrow against collateral. Privacy is the selling point: positions are confidential by default, and settlement is onchain. TVL is $2.4M (DefiLlama).

Funding: not disclosed.

8/ CBTC by BitSafe (Bitcoin on Canton)

CBTC is a Bitcoin-backed asset on Canton issued by BitSafe, the way bitcoin enters the network. BitSafe also builds tools for institutions to launch, move and manage assets on Canton, including the open-source Decentralization Manager. CBTC TVL is $3.6M (DefiLlama).

Funding: not disclosed.

The pattern: money and activity on Canton sit in trading. Six of the eight are exchanges of some kind, one of them handles about 85% of DEX volume, and only three teams have disclosed a round: Cancore ($7.2M), Temple ($5M plus an SBI-led round) and Ekiden ($2M). Lending and bitcoin are still small, at $2.4M and $3.6M in TVL. The SEC letter and the first tokenized stock trade give Canton a regulatory story; the onchain numbers show how early its app layer still is. Volume and TVL are from DefiLlama as of November 5, other metrics are company-reported. Do your own research, nothing here is financial advice.

Which one are you watching?

Nine early-stage projects that hit our radar, real products rather than memecoins. The theme: the boring parts of finance. Crypto spent years inventing new ways to trade. This week's names rebuild the services a normal economy actually runs on: a loan, a shop till, a team budget, a fund manager, a will and a risk report. Each one swaps a back office for rules written onchain. Follower counts are small on purpose, you are seeing these early. With a link to each.

1/ Unlloo (Reputation Lending)

Almost every loan in DeFi asks you to lock up more than you borrow. Unlloo lends USDC, USDT and DAI with no collateral at all. A borrower posts a request with a maximum APR, lenders compete with offers at or below it, up to 200 per loan, and an onchain reputation score decides who gets credit. Funds sit in onchain escrow until the loan is filled. Live on Ethereum mainnet. With no collateral, lenders can lose money, and the project says so on its front page.

2/ Birch Hill (RWA Credit)

Most onchain yield still comes from lending to other traders. Birch Hill lends against real buildings. Its first vault, live on Base through Morpho since October 13, offers USDC lending backed by Groma's apartment buildings in Boston and Providence, with the asset value published onchain by Chronicle. Access can be restricted depending on who you are. On October 28 it announced a $4M seed led by Multicoin, with ParaFi, Castle Island and Nascent joining.

3/ Ibis (Private Credit)

Businesses in frontier economies are among the hardest borrowers to reach from Wall Street. Ibis is building institutional private credit for them. It came out of stealth on October 28 with a $250M structured credit facility from FalconX, which also led its pre-seed with Anchorage and Innovating Capital. The product is on testnet and access is by request. Three posts in, this is as early as it gets.

4/ Kova (Payments)

A small shop that wants to take stablecoins usually needs a card machine and a processor. Kova turns a phone into the till: the shop taps or scans, the customer pays in digital dollars, and the money settles on Monad in about a second, straight to the shop's own wallet. Since November 2 users also get a free virtual Mastercard funded from their balance. The numbers are real and still tiny: 70 people, 158 transactions, $3,396 moved, 7 shops open and payouts to Nigerian bank accounts. The iPhone app runs on TestFlight.

5/ Nuntio (Team Spending)

Give an AI agent a company card and you find out what it bought later. Nuntio is one shared onchain account for a team, where every person and every AI agent gets its own spending limit. An agent proposes a payment, the contract checks the limit, and anything above it waits for a human to approve. Teams already use it to buy AI compute and Claude with AUSD. Live on Monad with agents in beta, and 50 active users in the last 30 days.

6/ Buoy (Trading Vaults)

Handing money to a trader usually means trusting them with it. Buoy runs non-custodial vaults on Hyperliquid's HyperEVM: a lead trader manages pooled USDC, depositors hold redeemable shares as tokens, and the trader earns a fee only when the vault hits a new profit high. Live on mainnet, with Pear Protocol's trading engine plugged in since October 9. By October 1 its profitable vaults had made depositors $9,000+. No token. One vault's good month says little about the next one.

7/ Heres (Inheritance)

If you die tomorrow, your seed phrase most likely dies with you. Heres is a succession protocol on Solana: you create a capsule that sets who gets which assets and when, and it settles onchain. Its site counts 170 capsules created, 139 already delivered and about 54 SOL secured. Built with MagicBlock, it took second place at the MagicBlock Blitz hackathon and is in Alchemy's Solana fund program.

8/ Pegana (Peg Risk)

Pegs rarely break all at once, the gap widens first. Pegana watches 67 Solana stablecoins, liquid staking tokens and yield-bearing dollars against their market and intrinsic value, and every change of state comes with a hashed receipt anyone can replay. The oracle went open source on October 26, with an API and an MCP server for AI agents. So far it is funded by a Superteam Brazil grant.

9/ Branchpoint (Conditional Markets)

Every big event splits the future in two, and your portfolio has to live in both. Branchpoint splits an asset into conditional versions, one for each possible outcome, so you can trade, lend and borrow inside each branch, then recombine or settle once reality is known. Instead of betting on one world, you price all of them. Testnet is live, the beta is closed and the waitlist is open.

The pattern this week: crypto is finally rebuilding the dull core of finance. A loan without collateral, credit backed by apartment buildings, a till on a phone, a budget an AI agent cannot overspend, a trader who never touches your keys, a will that settles itself and a peg report with receipts. None of this trends on the timeline, and that is the point. These are early-stage projects, several still on testnet, in beta or on TestFlight, so higher upside comes with higher risk. Do your own research, nothing here is financial advice.

Which one are you watching?

Most Elysium headlines are about Uniswap and Pump.fun. The more interesting list is the teams that only exist on Hyperliquid. We mapped 15 of them: each one has posted its own Elysium plans in the last few weeks.

What Elysium is

Elysium is the L2 that Kinetiq built for Hyperliquid. HYPE is the gas, blocks land every 100 ms, and it targets about 100x the gas per second of HyperEVM. Sequencer revenue is split three ways: 50% buys back and burns KNTQ, 25% goes to the apps that bring the traffic, 25% to the treasury.

Mainnet opens on November 9. Until then the network is open to partners only, and Elysium has warned users about unofficial tooling.

DEXs and aggregators

• Nest: HYPE-aligned DEX with liquidity from the first block (announcement)
• Hybra: Hyperliquid's public liquidity layer, deploying on Elysium from day one (announcement)
• HyperFlow: aggregator with $1B+ routed on Hyperliquid, swaps live on day one (announcement)

Launchpads

• Ascend: no platform token, KNTQ buyback share raised to 80% (announcement)
• Signal: creators choose who their coin's fees go to (announcement)
• Oxley: token launches by auction instead of a bonding curve (announcement)

Trading and markets

• Speculate: prediction markets and card packs, live November 9 (announcement)
• Hence: turns a market thesis into a tradable basket (announcement)
• Kalos: trading hub for Elysium, first pick of Ascend's incubator (announcement)

NFTs and games

• AtlasHL: NFT marketplace for HyperEVM and Elysium (announcement)
• Elytra: first NFT collection on Elysium, with its own marketplace (announcement)
• Lutra: 5,555 otters that evolve with your Hyperliquid trades (announcement)
• Eloscape: virtual mining protocol on Elysium (announcement)

Data and analytics

• HypeDexer: indexed chain data and archive RPC for builders (announcement)
• Liquid Terminal: free Elysium dashboard with a tx simulator and token checker (announcement)

Also on day one

Uniswap, 0x, KyberSwap, Pump.fun, fomo and Banana Gun are launching on Elysium too. The big Hyperliquid DeFi names (Felix, HyperLend, Hyperbeat and others) have not announced Elysium plans yet, so for now the native ecosystem is mostly young 2026 teams: launchpads, DEXs, NFTs and terminals.

On November 2, Brazil's Mercado Bitcoin froze deposits and withdrawals after a cyberattack with over $40M in suspected losses. It came four days after Binance lost $6.36B, the largest crypto hack ever. Two exchanges, two very different sizes, one question: what happens to an exchange after it gets hacked?

We went through every centralized exchange hacked for $1M or more since 2011. There are 69 of them.

The split

• Survived on their own: 35 (51%)
• Rescued by a buyer: 6 (9%)
• Sold later, at a premium: 4 (6%)
• Died after the hack: 14 (20%)
• Survived the hack, closed later: 10 (14%)

In total, 24 of 69 are gone, about one in three.

Survived on their own (35)

Bybit lost $1.46B in February 2025 and never stopped withdrawals. Partner loans and deposits refilled its reserves within 72 hours. KuCoin lost $281M in 2020, had BTC and ETH withdrawals back in 11 days and covered the last $45.6M from its insurance fund. Bitfinex went the hard way: after its 2016 hack it cut every balance by 36%, issued tokens for the loss and bought all of them back at $1 within 8 months.

Rescued by a buyer (6)

Coincheck lost $530M in 2018, paid clients back from its own capital and was sold to Monex 70 days later for $33.5M, about 6% of the loss. Zaif went to Fisco, and DMM Bitcoin's accounts went to SBI. On median, a buyer showed up 2.6 months after the hack. A rescue is no guarantee: Liquid needed a $120M loan from FTX and died with it, and two smaller rescues collapsed under their buyers.

Sold later, at a premium (4)

A hack does not always mark the top of the risk. Deribit lost $28M in 2022, and Coinbase later bought it for $2.9B, 104x the hack. Bitstamp lost $5.3M in 2015, and Robinhood paid about $224M for it. Upbit, hacked for $49M in 2019, has a pending deal with Naver.

Died after the hack (14)

Mt. Gox, BitGrail, Cryptopia and Lykke are the best known. The dead were mostly small: their median hack was $6.5M, against $26M for exchanges that survived. Of the 25 exchanges that lost $40M or more, only Mt. Gox and BitGrail died right after. Below $40M, 12 of 44 did. Dollar size here is a proxy for exchange size, so read it as a pattern, not a law.

Survived, then closed (10)

BitMart, CoinEx, EXMO, AscendEX and NoOnes all got through their hacks, then announced closures between July and September 2026. Only EXMO partly tied the decision to its hack.

The clock

We checked withdrawal reopening dates in 20 cases. Exchanges that survived had withdrawals back in a median of 5 days, and 11 of 13 reopened within three weeks. Bybit, Crypto.com and Deribit took less than a day. Mt. Gox, Cryptopia and Lykke never reopened. All three died.

Who paid matters as much. In 12 of 21 cases the exchange covered the loss itself from reserves, SAFU or an insurance fund. Four relied on loans or outside money, two cut client balances, and in three cases nobody paid. All three of those died.

Binance: survivor territory

Binance lost $6.36B on October 29, about 3.7% of its $171B in reserves. It replaced 95% of the stolen amount from SAFU, its insurance fund and treasury, and reopened withdrawals 2.4 days after the drain started. On every measure we tracked, that sits inside the survivor range.

Mercado Bitcoin: the number to watch

Mercado Bitcoin, founded in 2013, serves 4.5M clients. The main drain took about 34 minutes across roughly 10 chains, through ordinary signed transfers rather than a smart contract bug. MB says the incident is contained, with no harm to clients and no data leak, and Tether froze 4.5M USDT.

MB applied for its new Brazilian central bank licence on October 27, six days before the hack. The rules already required an incident response plan, and nothing in them says a hack costs a licence.

Deposits and withdrawals are still paused, with no reopening date. History says that date is the clearest early signal of how this ends.

Sources: company statements and filings, Banco Central do Brasil, Hypernative, CoinDesk, The Block, Reuters, CryptoDiffer analysis of 69 exchanges hacked for $1M+ (2011-2026).

On October 29 the Binance hacker drained the exchange's hot wallets at 20:24 UTC and started a race against the freezes. Stolen USDT can be blocked by Tether at any moment, ETH cannot. So between 20:34 and 20:49 UTC the hacker sold 325M USDT in 9 orders through UniswapX and 1inch Fusion and got 122,316 ETH, at an average of $2,657. Binance was trading near $2,300. The last order went through 8 minutes before the first freeze of the night. Speed cost the hacker about $45M on this one swap.

ETH at $3,487, for one block

At 21:09 UTC, ETH in Uniswap's main ETH/USDT pool hit $3,487 while Binance showed $2,266. In the same block, USDT itself traded at $0.798 against USDC on Uniswap. Part of the spike was USDT breaking down, not ETH flying: in USDC pools ETH peaked around $2,780. Either way, the on-chain premium stayed above 10% for about 45 minutes.

Who got the money

• Block builders and validators: ~7,200 ETH (~$16.5M). That is what they collected in tips in the hour from 20:30 UTC. The same hour a day earlier paid 4.55 ETH. Titan built 191 of the 298 blocks, and a single transaction paid beaverbuild 1,070 ETH (~$2.4M).
• Arbitrage bots. The best one borrowed 11,701 ETH on Aave in 18 loans, sold it on-chain near $2,595 and bought it back near $2,221, clearing about $4.4M. Another wallet pulled 4,175 ETH from Kraken and sold it on-chain 72 seconds later for $11.24M, about $1.8M above the Binance price.
• A 1inch resolver that kept the stolen USDT. Instead of buying ETH on the market, it posted 28.6M of the hacker's USDT as collateral on Compound, borrowed 10,041 ETH and handed the hacker 9,566. The position is still open, worth about $6M, and the stolen USDT is still sitting there.
• The biggest filler, with the smallest cut. A resolver registered on 1inch as Kipseli handled 145M USDT, 45% of the volume. It kept about 149 ETH (~$337K) and paid builders 1,221 ETH, eight times more.

Then came the altcoins

After ETH, the hacker dumped dozens of tokens at a discount, and the bots were waiting. jaredfromsubway sandwiched about 25 of the hacker's 29 ONDO sales. 7M UNI went for an average of $5.69 while Binance showed $6.41. On Tron the attacker swapped 234M USDT into TRX at $0.367 against $0.334 on Binance, another ~$21M given away. Across all legs, our on-chain estimate puts the hacker's total overpayment at about $90M.

Why market makers did not take it all

The trade was open to anyone: buy ETH on a centralized exchange, sell it on-chain for 10% more. The catch was having ETH outside Binance, where withdrawals were frozen, and beating bots that hand most of the spread to builders just to land first. Most fillers bought the hacker's ETH on Uniswap in the same transaction, so the spread leaked to pools, arbitrageurs and block builders.

Every dollar the hacker paid for speed ended up with someone who was faster.

Figures are on-chain estimates against Binance prices in the same minute. Builder tips are paid out of the arbitrage spreads, so the groups overlap. Sources: Ethereum and Tron on-chain data, Binance 1-minute prices, TRM Labs.

Between February 2025 and October 2026, attacks publicly tied to TraderTraitor, the North Korean unit inside Lazarus, took about $8.57B from crypto exchanges and bridges. Binance accounts for 74% of that sum. About 33% of the total is frozen, but almost all of it comes from one case.

Bybit, Feb 2025: $1.46B

A Safe{Wallet} developer's computer was hacked and malicious JavaScript was injected into the interface. Bybit's signers saw a normal-looking transaction and approved the drain of a cold wallet. About 3% was frozen. Within a month 86% of the ETH had been converted to BTC, most of it through THORChain. This is the only attack in the list with formal FBI attribution.

BtcTurk, Aug 2025: $48M

Hot wallets were breached, most likely through a stolen key. There is no data on frozen funds. Part of the stolen ETH later helped fund the Kelp attack.

Kelp DAO, Apr 2026: $292M

The attackers hacked two internal RPC nodes and DDoSed the external ones. The LayerZero bridge relied on a single verifier, which confirmed a fake burn. The Arbitrum Security Council froze about $71-75M (~25%). Around $220M has been laundered through THORChain, Wasabi, Tornado and Umbra.

AFX Trade, Jul 2026: $24.15M

A fake recruiter spent two weeks working with the team. A malicious git config gave the attackers the bridge validator keys. The stolen USDC became 12,467 ETH. Nothing has been recovered.

Bitget, Sep 2026: $387.5M

A zero-day in a third-party security product let the attackers forge withdrawal commands in the wallet system and wipe the logs. Every stablecoin was swapped within 41 minutes, so only $0.84M (0.2%) was frozen. $269M went through THORChain.

Binance, Oct 2026: $6.36B

A DevOps engineer's laptop led to a service container and then to signing access in the hot-wallet system across 10 chains. On Ethereum the attackers used 50 pre-signed EIP-7702 authorizations, after 36 hours of staging and test runs. About $2.7B (42%) is frozen: Tether ~$1.77B, Circle ~$825M, Ripple $40.5M, WLFI $27.8M, Arbitrum ~$30M. Roughly $2.4B in ETH and BTC is out of reach.

What the six have in common

• People, not code. None of the attacks used a smart contract bug. The way in was always a person or a piece of infrastructure.
• Stablecoins are the only thing that gets frozen. Binance's 42% exists because issuers caught about $2.66B in stables in time. Without Binance, about $117M of $2.21B was frozen, roughly 5%.
• THORChain is the exit. Loot from Bybit, Kelp, Bitget and Binance was moved to BTC through it.
• Weeks of preparation. Operator wallets funded through Tornado ahead of time, two weeks of social engineering at AFX, 36 hours of staging at Binance.

How solid is the link

Only Bybit has formal FBI attribution. The rest comes from analysts: TRM Labs, Elliptic and others. So far the link between the hacks rests on on-chain tracing: ZachXBT matched Binance and Bitget funds at one THORChain address. A shared laundering address points to one network, but it is not proof of one team.

Also linked: Phemex (~$85M, January 2025) sits just outside the window. MistTrack traced Binance funds to a wallet the Phemex hacker used before. With Phemex, the total is about $8.66B in 21 months.

Excluded: Drift ($285M, April 2026). TRM ties it to a different North Korean group.

Sums are at the time of each attack. Sources: TRM Labs, ZachXBT, Elliptic, BlockSec, MistTrack.

Nine early-stage projects that hit our radar, real products rather than memecoins. The theme: guardrails. Most crypto horror stories start the same way: a wick liquidates you, a public address shows the world your balance, or money moves that nobody really approved. This week's names are built around exactly those failure points. Options with up to 100x and no liquidations, wallets that give every payment a fresh or encrypted address, AI that can score work but cannot release the money, and agents that watch your positions and wait for your signature. Follower counts are small on purpose, you are seeing these early. With a link to each.

1/ Isometric (Options)

Plenty of leveraged traders are right on direction and still get wiped out by a wick. Isometric sells options that trade like perps: pick a direction, leverage from 5x to 100x and a term of 1, 3, 7 or 30 days. Only the price at settlement counts, so there is no liquidation and no funding. One USDG pool per market takes the other side, and its LPs earn the premiums. Live on Arbitrum's Sepolia testnet since October 20 with a mainnet waitlist, launching on Arbitrum and Robinhood Chain, built by a public team that came from Liquidium.

2/ Treasures (Tokenized Stocks)

The same stock can now exist as several different tokens from different issuers, on different chains, at different prices. Treasures quotes them side by side, six issuers including Ondo, xStocks, Backpack and Coinbase across Solana, Ethereum, Robinhood Chain, Base and Arbitrum, and routes each USDC order to wherever it buys the most shares, for a flat 0.1% fee. It covers 2,000+ tokenized stocks, ETFs and commodities, runs copytrade baskets from Buffett to Citrini, and sells the same routing as an API to apps and AI agents. Eight products are already building on it.

3/ Cloaked (Private Wallets)

Every Ethereum address is a public bank statement. Cloaked is a self-custody wallet where each incoming payment lands at a fresh address, while you still see one balance and share one name, yourname.clkd.eth. Private sends route through Privacy Pools, and you pay gas in the token you are sending. Its browser extension went live on October 21 and connects every app with a fresh address. Live on Ethereum, Base and OP Mainnet, with $1.5M+ moved through it by mid-September.

4/ Alcove (Private Payments)

Stablecoin payments are fast, but every transfer is public. Alcove is a mobile wallet for dollars on Aleo, where USDCx is backed 1:1 by USDC and the sender, receiver and amount are encrypted by default. You can shield USDC in from other chains, and you need neither an ID nor a seed phrase. Live on iPhone and Android, and it supported Circle's Arc from day one.

5/ Signo (DeFi Agents)

DeFi positions do not sleep, and neither does the risk. Signo runs AI agents that watch your positions on Aave, Morpho, Pendle, Uniswap and Aero around the clock, then send a Telegram alert with a ready transaction that you sign yourself. In early access, agents can also act on their own, but only within limits you sign onchain, on Arbitrum and X Layer. It never holds your keys. Signo went live at OKX Dev Day in Singapore on October 20 and took first place in one track and second in another.

6/ SprintOS (Milestone Escrow)

When an AI checks your work, who decides you get paid? On SprintOS a sponsor locks USDC for each milestone on Stellar, the builder submits proof, and an AI scores it against requirements hashed into the contract. The AI cannot move a cent: only the named human reviewer's wallet can release the escrow, even after a perfect score. The first real mainnet payments went through on October 6, and so far it is backed only by Stellar grants.

7/ Yieldy (Stablecoin Yield)

Stablecoin yield looks safe right up until the protocol behind it breaks. Yieldy grades each opportunity on 54 signals across 9 risk areas, builds a strategy for your goal, and its Copilot rebalances or pulls you out when the risk rises. Non-custodial, starting with USDC on Ethereum. It took 1st place at Ownership Roadshow #5 on October 14. Early access is not open yet, so for now it is a waitlist.

8/ Lifted (Stablecoin Looping)

Looping stablecoins can multiply yield, until borrow rates spike and eat it. Lifted is a USDC vault on Solana that loops across RWA-backed and onchain stablecoins on Kamino, chases the best spread between them, rebalances only when it is worth the cost and is built to protect against negative rate spikes. Live since October 20, from a team of Superteam Kazakhstan builders. At launch it showed 13% APY over the previous 30 days. Looping is still leverage, so the risk is higher than a plain deposit.

9/ WhereDidTheyBridge (Onchain Forensics)

Follow money onchain and the trail usually breaks at a bridge. Paste a wallet or a transaction into WhereDidTheyBridge and you get every bridge it used in one timeline, across deBridge, Across, Wormhole, CCTP, Relay, Mayan and more on 100+ chains. Since October 18 it follows NEAR Intents swaps to where the money actually landed, and an MCP server lets Claude or Cursor query it. Free without sign-up, no token, built by an independent developer.

The pattern this week: guardrails are turning into the product. Options that cannot be liquidated, wallets that do not leak your balance, AI that can score work but not pay it out, agents that wait for your signature, and yield that comes with an exit plan. In most of crypto these were footnotes in the docs. Here each one is the whole pitch. These are early-stage projects, several still on testnet or a waitlist, so higher upside comes with higher risk. Do your own research, nothing here is financial advice.

Which one are you watching?

Between January 1 and October 28 we counted 30 hardware wallet incidents across 11 vendors. Known losses add up to about $530M. Not one dollar of it left through a cracked secure chip. The money went out through buggy code, tampered retail units, leaked customer lists and owners who were talked out of their seed.

Where the $530M went

- People tricked: ~$295M (56%)

- Code bugs: ~$117M (22%)

- Tampered retail units: ~$102M (19%)

- A vendor's software wallet: ~$16M (3%)

- Cracked secure chips: $0

Several of these are third-party estimates. Ledger has not confirmed a total for its implant case, and D'CENT has published no figure at all.

1. Code bugs: weak seeds shipped years ago

The biggest technical failure of the year was Coldcard. A firmware build setting introduced in March 2021 made the device skip its hardware random number generator. Seeds came from a software generator fed by the chip ID and timers, which left roughly 40 bits of entropy on Mk2/Mk3 and about 72 on Mk4/Q/Mk5. That is guessable. From July 30 attackers brute-forced seeds remotely, without ever touching a device, and swept 1,789 BTC from 8,865 addresses: $114.7M at time-of-theft prices, per Galaxy. Coinkite patched the firmware but offers no compensation, and its postmortem is still unpublished. Coinkite's own take: "we have to assume that someone used AI to review previous versions of our firmware."

A firmware update does not fix a seed that was already weak. Affected users have to create a new seed and move their funds.

Ledger's Zilliqa app had a similar flaw since 2019. The top 64 bits of every signing nonce were fixed at zero, so a private key could be recovered from 4 or more signatures. Between March and July, 683M ZIL (about $1.9M) was taken from 51 accounts, and 6,772 accounts were exposed. Ledger merged a patch on July 27, which does not protect keys that already signed.

Other bugs were found and fixed before anyone lost money:

- Ledger Monero app (fixed March 20): a compromised computer could pull Monero secret keys without on-screen confirmation.

- BitBox02 (July 8 and August 17 updates): an out-of-bounds write, display spoofing, memory corruption that allowed code from a malicious host on unset Multi-edition devices, and a Silent Payments bug that could lock funds to the wrong address.

- Ledger Ethereum app (August): a race condition let a malicious host swap a transaction during signing. Ledger found it internally and fixed it before TestMachine went public on August 22. Safe version: 1.22.3.

D'CENT is a separate case. On September 15 and 16 funds were drained from its App Wallet, the software mode, for addresses that had signed with app versions below 8.1.0. On-chain trackers count about 12.4M XRP from 7,000+ wallets, roughly $16M. D'CENT has disclosed neither a cause nor a figure. Its hardware wallets are not affected unless the seed was typed into the app.

2. The retail chain: implants, swapped chips and a stolen factory key

On October 9 wallets tied to Ledgers sold by Southeast Asian reseller CryptoBilis were emptied on 6 chains in 47 minutes. The devices were genuine but had been opened and fitted with an LTE implant (microcontroller, modem, eSIM, antenna) that read the recovery phrase off the screen and sent it out. The secure element was untouched, so Ledger's Genuine Check passed. Bitquery estimates ~$102M from 335 wallets, and Tether froze $10M in USDT. Ledger calls the number of affected devices "limited", has not confirmed a total and offers no recall or reimbursement. It halted sales in Indonesia, Singapore, Thailand, the Philippines, Vietnam, Malaysia and Hong Kong, and at MediaMarkt Saturn. OneKey cut ties with CryptoBilis and offers free replacements to its buyers.

On October 13 hardware hacker Joe Grand showed a Trezor One bought on TikTok Shop and a Model T from Facebook Marketplace whose microcontroller had been replaced to generate pre-set seeds. No losses were reported. Trezor will end all reseller authorizations on January 1, 2027.

Then came Keystone. Between October 20 and 26 an attacker got into its website through a third-party monitoring tool and took buyer data plus the WebAuth key: the key written into devices at the factory to prove they are genuine. With it, counterfeits could pass Keystone's authentication check. Keystone will destroy 18,835 unsold units, has stopped shipping web orders and moved to preorders with a new key. No funds have been lost. Devices delivered before October 20 are fine. If you received one after that date, contact Keystone support before using it, even if it passes the check.

The lesson from all three: a genuine check proves the chip, not the whole box.

3. Leaks and hijacked vendor channels

Vendors and their suppliers leaked buyer data six times this year, almost always through a third party:

- Ledger via payment partner Global-e (January 5): names, addresses, emails, phones and order details. Count not disclosed.

- Trezor via shipper ShipMonk (August 13, widened September 4): 80,689 customers, including ~67,000 US records from 2019 to 2021 that Trezor had been told were deleted.

- SafePal (August 16): a flaw in an order-tracking plugin exposed 39,798 customers.

- Trezor and BitBox via email provider Brevo (September 9 and 10): 347,149 Trezor contacts were exported and sent a fake "STM32 Entropy Vulnerability" alert. 2,500 people clicked in the ~20 minutes before takedown.

- Keystone (October): 25,964 buyer records plus 12,873 unpaid checkouts, part of it data that a broken erasure job should have deleted.

- Foundation (October 25 and 26): attackers with compromised Cloudflare API keys sent 23,000+ fake "recall" emails from official domains. ~5,600 mailing addresses and ~5,200 phone numbers were potentially exposed.

That is ~165K records with names and shipping details, not counting Ledger's. These lists feed the paper letters and calls described below.

Phishing also came from vendors' real channels:

- NGRAVE (October 17): attackers hijacked the push-notification provider of its LIQUID app and sent a fake "Pixel 9, Moscow" login alert linking to a seed-phishing page. At least $2.56M was stolen from ~70 addresses. The CEO: "we have absolutely and utterly failed."

- SecuX (October 16 to 19): the banner inside the SecuX App was modified to link to fake SecuX sites.

- Coldcard (October 11): its official X account published a phishing "migration" post.

"Check the sender" no longer protects you.

4. People and fake apps

The largest single theft of 2026 was not a device failure. On January 10 one hardware wallet owner lost $282M+ in LTC and BTC (2.05M LTC and 1,459 BTC) to social engineering, per ZachXBT. The funds were swapped to Monero and bridged via THORChain.

The rest of the list:

- February: paper letters to Trezor owners demanded a fake "Authentication Check" by QR code.

- April 7 to 13: a fake "Ledger Live" app in Apple's App Store asked for recovery phrases and took $9.5M from 50+ victims. Apple removed it.

- May: letters signed in the name of Ledger CTO Charles Guillemet about "Quantum Resistance", again with a QR code to a seed-phishing site.

- June: a group posing as wallet and exchange support drained $1.2M from a Trezor owner after a spoofed BitcoinIRA email (ZachXBT).

- July: Kaspersky found OkoBot malware ("SeedHunter") that overlays a fake recovery screen on Ledger Live and Trezor Suite. Hundreds of victims in 25+ countries.

- August: a phishing site on sites.google.com was the top sponsored Google result for "Trezor wallet". Rapid7's "Operation ASTERIX" paired phone scams with fake Trezor Suite, Ledger Live and Exodus apps.

- October 16: fake "Ledger firmware update" emails rode the CryptoBilis panic.

In none of these did the hardware fail. The user typed the seed into something else.

5. The chip held

Two lab attacks on secure chips worked this year, and both cost users $0. Ledger's Donjon team used laser fault injection to bypass the firmware-signature check of the TROPIC01 chip in the Trezor Safe 7, and to reset the password on Tangem cards. Both need physical access and a lab. The Tangem setup costs around $250,000. Neither can be patched with firmware, and neither has led to a reported theft.

In court

- Kim v. Ledger SAS (S.D.N.Y., August 27): a proposed class action over the 2023 Connect Kit incident and later impersonation scams, seeking at least $500M.

- Naderi v. Coinkite (Ontario, September 11): a proposed class action over the Coldcard seed bug, alleging more than $140M in losses.

What to do

- Buy only from the vendor, not from resellers or marketplaces.

- Generate the seed on the device. If your Coldcard seed was made on affected firmware, create a new one and move your funds.

- Never type a seed into a website, app, letter, QR page or support chat. No vendor will ever ask for it, and there are no "recalls".

- Treat any urgent security alert as phishing, even if it comes from an official email, app or X account.

- If you bought from Trezor, SafePal, Keystone, Ledger or Foundation, assume your name and address are already out there.

Sources: Coinkite, Galaxy, Zilliqa, Ledger, Bitquery, Keystone, Trezor, NGRAVE, D'CENT, Ledger Donjon, ZachXBT, Rapid7, Kaspersky, court filings, CryptoDiffer research.

Microsoft, Alphabet, Goldman Sachs and 19 other US companies have formally objected to their shares trading as tokens on OKXICE, the planned on-chain stock venue from NYSE owner ICE and OKX. That is 19 of the 62 companies on the venue's launch list, plus three that weren't even on it.

How the veto works. In September the SEC opened a five-year "Innovation Exemption" that lets tokenized versions of US-listed stocks trade on-chain, 24/7. The tokens must be backed 1:1 by real shares and pass through dividends and votes. If the token is issued by a third party rather than the company itself, the venue has to notify the company and wait 30 days. A written objection in that window keeps the token off that venue. It doesn't stop other venues.

Who said no. OKXICE published its launch list of 63 tickers on October 4. By October 23 its public notice listed 22 objectors: Cerebras, AppLovin, Applied Digital, Rocket Lab, CoreWeave, Verizon, Applied Optoelectronics, Broadcom, UnitedHealth, Goldman Sachs, Microsoft, Qualcomm, Caterpillar, McDonald's, Western Digital, Alphabet, Arm, Wells Fargo, Moderna, Cisco, Intuit and Marvell. Sixteen are Nasdaq-listed, six trade on the NYSE. Ten are chip or AI-infrastructure companies.

Why. The objection notices give no reasons, but the logic is easy to piece together.

- No upside. The company raises no money and earns nothing. Fees go to the venue and the tokenizer, while any price gap or blowup lands on the company's name. One analyst summed it up: the revenue goes to the tokenizers, the risk goes to the companies. Applied Optoelectronics, the only objector to comment on the record, said it sees no "clear benefit to AOI or our shareholders that justifies these uncertainties."

- Weaker plumbing. These venues sit outside the national market system: no consolidated tape, no real-time best-price display, trade reports can lag up to 10 minutes. Nasdaq opted its own stock out for exactly these reasons.

- No more quiet hours. Companies usually drop earnings and big news while markets are closed. A 24/7 token market removes that pause.

- Messy shareholder mechanics. Who holds the shares behind the token, how votes and dividends reach holders, what the transfer agent sees: none of it is standard yet.

- Saying no is free. Legal teams get 30 days to judge a brand-new market. Silence counts as consent, an objection costs nothing.

- The venue's own fine print. OKXICE's notice says it isn't SEC-registered, isn't subject to Reg NMS, and can upgrade its smart contracts with no timelock or advance notice.

Who stayed quiet. 43 listed companies haven't objected, including Nvidia, Apple, Tesla, Amazon, JPMorgan and Walmart. Every crypto-linked name on the list (Coinbase, Robinhood, Circle, Strategy, Securitize, BitGo, SoFi) is in that group, and Securitize and BitGo have said they won't object. The 30-day windows for the first batch close in early November, so the count can still grow.

What it means. This is a vote against one venue's third-party model, not against tokenization itself. Nasdaq, which opted out, is building its own tokenized-stock framework. Trading on OKXICE is also capped at 0.25% of each large-cap stock's average daily volume. But the message from corporate America is clear: if you want to put our stock on-chain, ask us first, and show us what's in it for us.

Sources: OKXICE public notices (Oct 4, 19, 23), SEC TSV Exemption order, CoinDesk, Fortune, Decrypt, Markets Media, Covington.

October was busy, so we kept only products you can use today, shipped this month by teams that were already established. Eight made the cut, and seven of them live onchain. The pattern is hard to miss. Bets got shorter, simpler and stranger: 1,000x calls that a 0.1% move can end, 15-minute BTC options, odds on the next DeFi exploit. Wall Street assets kept moving onchain, from Nvidia shares with dividends to pre-IPO OpenAI tokens. Underneath, DeFi rebuilt its plumbing: a market for the gap between funding rates, and two of the biggest DEXs merging into one. And OKX built a new front door, a dollar account for emerging markets. Each product gets a mini-review and one honest catch.

1/ OKX Money (Stablecoin App)

OKX Money is a standalone app, launched October 6, that turns more than 50 local currencies into digital dollars: USDG, USDC or USDT. It is rolling out in parts of Latin America, Africa, South Asia and the Middle East, with a virtual or physical Mastercard that charges zero FX fees abroad. Eligible USDG balances earn up to 10% APY with no lockup, paid weekly, and the top rate depends on your 30-day balance, 30-day card spend or OKX VIP level. The catch: OKX has not said what funds the yield, and the app is not offered in the US or EU.

2/ Jupiter YOLO (Leverage Bets)

Pick up or down on SOL, BTC or ETH at 500 to 1,000x leverage, starting from $10. There is no order book: every trade goes against a House pool that collects losing stakes and pays out winners, and the only trading fee is 2% of winning profit. Live in beta since October 15, YOLO generated about $42K in fees in its first week (DefiLlama, Oct 21). The catch: at 1,000x a 0.1% move against you wipes the stake, the code is unaudited, and if the House runs dry, profits wait in a queue with no payout date.

3/ Lighter Binary Options (Binary Options)

Since October 19, Lighter runs 15-minute up/down markets on BTC and ETH. A share costs between $0.01 and $0.99, which reads as the market's odds, and pays 1 USDC if you are right and nothing if you are not. Positions are paid in full in USDC, so there is no leverage, no funding and no liquidation, and every window settles on Chainlink prices. The catch: Lighter has not published any volume yet, it is closed to US users, and at even odds the taker fee means you have to win a little more than half the time just to break even.

4/ Pump.fun PreStocks (Pre-IPO Tokens)

Since October 19, Pump.fun lists PreStocks tokens for eight private companies: OpenAI, Anthropic, Anduril, Neuralink, Figure AI, Kalshi, Polymarket and Oura. You can trade them directly or launch a memecoin priced in one of them, and coins paired with the OpenAI and Anthropic tokens were trading within a day. The tokens are issued by PreStocks against SPV exposure and are closed to US persons. The catch: holders get no ownership, votes or dividends, OpenAI and Anthropic have warned that share transfers to SPVs without their approval can be void, and liquidity is thin: the deepest Anthropic pool held about $2.2M and the biggest memecoin pairs under $30K (GeckoTerminal, Oct 21).

5/ Securitize Stocks (Tokenized Stocks)

Since October 8, Securitize sells 12 US stocks on Solana, including Apple, Nvidia, Tesla and SpaceX, each backed 1:1 by a real share. Holders get dividends and shareholder votes, buy with USDC, and Jump Trading provides liquidity. Unlike most tokenized stocks, these are open to eligible investors in the US and EU after identity checks, through Securitize's SEC-registered broker-dealer. The catch: the tokens only move between approved wallets, trading is 24/5 for now with 24/7 still "coming soon", and Securitize has not published any early trading numbers.

6/ DefiLlama x Trueo (Prediction Markets)

DefiLlama now hosts Trueo prediction markets that settle on DefiLlama's own data. Live examples: will Hyperliquid's October revenue top $60M, will Robinhood Chain DEX volume pass $40B in October, will Aave suffer bad debt or an exploit this month. Trueo's TVL rose to about $1.5M from $1.07M a day earlier (DefiLlama, Oct 21). The catch: the markets are tiny, and on exploit markets the people most likely to know first are the ones who found the bug.

7/ Pendle Boros Spread Markets (Funding Rates)

The same BTC perp pays different funding on different exchanges, and Boros now lets you trade that gap as one position. The first two spread markets pit Hyperliquid against Gate on BTC and ETH, maturing November 27, so arbitrageurs can lock in a fixed spread and others can bet on where it goes. Together they carry about $9.4M in notional open interest (Boros API, Oct 21). The catch: this is a tool for pros, Boros as a whole holds only about $7.5M (DefiLlama, Oct 21), and funding gaps can flip fast.

8/ Aero (DEX)

Aerodrome on Base and Velodrome on Optimism merged into one protocol called Aero, live since 00:00 UTC on October 22. One AERO token now runs across seven chains, Base, Ethereum, OP Mainnet, Arbitrum, Ink, Arc and Robinhood Chain, and swaps route across all of them. Weekly ve(3,3) voting is replaced by Predictive Allocation: sAERO holders direct rewards to pools on any Aero chain and earn exchange revenue from the pools they pick. The two DEXs held about $416M combined going into the switch (DefiLlama, Oct 21). The catch: nothing converts automatically, so holders have to move their AERO and VELO through the upgrade portal themselves, at 1 new AERO per AERO and about 0.044 per VELO.

The pattern this month: products got simpler on the surface and riskier underneath. A $10 up/down button, a 15-minute coin flip, a memecoin priced in OpenAI. Several of the newest bets and pre-IPO tokens are closed to US users, while the one product built inside US securities rules, Securitize Stocks, is open to them.

DYOR: nothing here is financial advice. Betas, unaudited code and thin liquidity are flagged above for a reason.

Which one would you actually use?