Founder @brcapital_fund | @super__protocol Fund, Liquidity, HFT, Institutional DeFi and Confidential AI

Decentralized
Why should you #HODL your #Bitcoin and never sell it? I know why. A short thread 👇
995
685
6,272
Before @CoWSwap and Pave Bank, we started the fund 2017 to 2018, when bitcoin:native had just dropped from $20k to $4k. Those were quite the times, check out this episode to learn more about the start of @brcapital_fund.
“Nobody wanted an interesting project if it had no token.” The fund started in that window: idea in 2017, launch in 2018, right after the ICO crash. - Bitcoin had gone from $20k to $4k. Pitching investors then was brutal. - Before the fund, there were attempts at diamonds on-chain and payments infrastructure. No token. No interest. - Mining, early deals, first-round names like Solana — those years were a learning curve, not a finished thesis. What came later filled in the map: - There were almost no DAOs yet. The first Ethereum DAO had already failed. - MetaCartel, The LAO, and the rooms after that brought pipeline and ideas. - They’re still in those communities. The industry was still being shaped in public. BR Labs founder @rezosh on why the fund was built in the worst window, and why that timing still matters. nitter.net/brlabsxyz/status/20994…
4
8
831
If you want to operate in the institutional world today, you cannot do it without intermediaries. We're in a transition phase, and inside it there are exactly two things a desk can do. First: minimize their number, because zero is practically impossible. Second: assess them (and their stack) for security, continuously. Because the fight for security is always a fight against zero-day, and that fight never stops. The vendor who passed the audit last quarter is the vendor whose backend gets compromised this time.
6
685
The actual consequences of pausing withdrawals: You don't need a finance degree to see it. Everything breaks: - Access to capital stops - Hedges you can't rebalance - Open positions you can't close. Counterparties you can't pay The channel you run every financial operation through goes dark, not the balance on the screen, the ability to act on it. That's why the Bybit-vs-Bitget distinction isn't PR nuance. Open withdrawals mean the failure stayed inside the exchange, while paused withdrawals mean it was exported to every desk with inventory there.
2
3
677
If we compare the Bitget situation to Bybit: Bybit lost $1.5B and kept withdrawals open, processed about $4B of them in the first 12 hours, and refilled reserves in 72 hours through emergency loans from Galaxy, FalconX, and Wintermute. vs Bitget lost $351.6M and paused withdrawals, while keeping deposits and trading open. A $464M protection fund sits against the hole; the hole is 76% of the fund. Until Bitget trusts its own request path again, users aren't holding coins; they're holding Bitget credit.
3
737
What makes the Bitget incident interesting is that no one stole the keys. That's the whole story, and almost nobody's reading it. Bitget's own account: - A compromised backend spoofed transaction data - Bitget's authorization process signed the withdrawals itself - Private-key compromise ruled out Hot wallet here just means inventory that a running approval process was still allowed to move. The secret held, but the instruction path in front of it didn't. Bybit, February 2025, was the same class of failure: a forged signing policy, not a stolen key. So the industry's largest hack and this week's both ran through the layer everyone treats as solved: not "who holds the key" but "what can tell the key to sign."
1
4
836
Solana is one of our holdings that we were proud of, one of the only blockchains to survive its own "death." As for institutions, people misunderstand ETH vs SOL, they are designed to cater to two different users. ETH = nobody’s layer SOL = one fast ecosystem (consumers, volume, heavy compute) nitter.net/brlabsxyz/status/21031…
Solana: A Payment Infrastructure with a Familiar Interface Solana’s institutional payments story is becoming easier to evaluate because the announcements now describe specific financial operations. Banks are settling card-network obligations, payment companies are issuing digital dollars, and existing distribution networks are connecting those dollars to customers. Each development answers a different question about whether a public blockchain can support recurring financial activity. In December 2025, Visa announced that Cross River Bank and Lead Bank had begun settling with it in USDC over Solana. That changes how participating institutions discharge obligations to the network. Customers can continue using familiar cards while part of the settlement process runs on blockchain infrastructure. Seven-day settlement gives treasury teams another way to manage the gap between transactions happening continuously and funding arrangements that follow narrower operating windows. The economic opportunity is the ability to move and reuse liquidity when obligations arise. Where participants previously needed to maintain balances ahead of a weekend or wait for the next funding window, more frequent settlement can change how much cash must sit in each location. The actual saving depends on collateral policies, access to stablecoins, redemption arrangements and the availability of liquidity at the receiving end. A fast transfer only solves the part of the process that can actually use it. Visa’s September update reported a $20b annualized stablecoin settlement run rate, up from the $7b figure it published in April when the pilot supported nine blockchains. That figure belongs to the overall program; it is not Solana-specific volume. Mastercard’s June announcement also included Solana among multiple supported networks. Early partners were named at the time; this month SoFi began settling a live U.S. card program in SoFiUSD on that network, with most of the token’s supply on solana:So11111111111111111111111111111111111111112. Institutional demand is expanding through a choice of infrastructure providers. Western Union adds another dimension. It launched USDPT on Solana in May, issued by Anchorage Digital Bank, and introduced Stablecard with Rain in August. The chain is becoming connected to an existing money-transfer and spending business, giving digital balances a distribution path beyond crypto-native wallets. For BR Labs, the useful next questions are operational: which counterparties settle repeatedly, how much volume persists without incentives, what happens during congestion, and whether funds arrive in a form the recipient can immediately use. We would also look for evidence that settlement changes reduce reconciliation work or improve liquidity utilization. These measures will say more about Solana’s role in payments than another list of institutional logos. Infrastructure becomes economically significant when counterparties keep relying on it to meet obligations, including on the days when trading activity and token prices offer little encouragement.
2
7
794
One thing that AI makes increasingly clear with things like Grok Bot, Muse, even GPT and Claude, is that the best UI or best UX is no UX. It’s talking to a chatbot or talking to your agent, with it doing all the work and running the applications in the background.
7
638
Why AI needs crypto. Think of it this way, I cannot send someone $0.05 via conventional methods. It’s impossible. SWIFT costs like $15 minimum, gets up to $ 40-45, or even $70 per international payment. Plus, nobody will give accounts to agents and AI. So it’s crypto per call, per job, international, 24/7, permissionless.
1
1
10
635
To be clear, BTC-backed loans as working collateral for small businesses are a real use case. It's also the minority version. The majority version is the same as every other cycle's.
1
7
790
My faith in the space didn’t weaken. The conviction didn’t simplify, it got denser. Imagine that it’s like the same box, just more consolidated, heavier, tighter. Before it was spread across like many vectors, many bets, now it’s concentrated.
2
6
639
The borrow-USDC-buy-BTC-repledge loop is nothing new. Looping is leverage, leverage is leverage. These are the canons of how markets work. Better collateral mechanics, a fixed rate, a nicer app, none of it rewrites the canon. BTC-backed loans still change the mechanics, specifically who holds the leverage. For the first time, people without a desk, a risk model, or a view on the cycle get a clean way to lever Bitcoin inside the app they already use. I read that the way I read it in 2021: retail just got access to leverage that will take them out in the first real drawdown. Some will spend carefully. Most who loop won't. The product is fine. The button is dangerous.
5
722
BTC-backed loans let you borrow USDC -> buy more BTC -> re-pledge. This raises the risk by exactly the amount you added, and every liquidation cascade in history ran on the same physics.
3
5
748
When borrowing strategically, the rate isn't what matters: timing, risk assessment, and, above all, knowing where we are in the four-year cycle. Borrowing against a volatile asset near the top of a cycle and borrowing near the bottom are different trades that happen to use the same product. The product doesn't know which one you're doing. You have to.
6
13
734
Our trading desk almost never borrows, but when we do, it's surgical and targeted. One example is how we funded our new prediction-markets desk with a loan against our Solana holdings rather than selling any of it. That's the entire use case for us: turn a position into working capital for a specific deployment, keep the position. Not spend. Not loop.
7
749
Coinbase launched fixed-rate USDC loans against Bitcoin through Morpho Midnight. Easy to read this as "millions can now borrow without selling," but they already could. @coinbase's floating-rate book has been live since January 2025 and sits at $1.4B outstanding against roughly $3B collateral. What launched is a LOCKED RATE and a LOCKED REPAYMENT DATE, with maturities at the end of this month or next. Two ways to lose the coins now instead of one: price and calendar.
1
9
726
BRRRolver ranked 1st on Ethereum this week by @CoWSwap consistency-reward share. 13.4%. That’s the metric that pays you for staying near the best price, not only for winning the auction. We’re investors, but we’re also builders, and BRRRolver is one of the things we're building in public. nitter.net/brlabsxyz/status/21027…
BRRRolver ranked 1st by @CoWSwap consistency-reward share on Ethereum for the week ending 22 Sep. CoW does not let traders choose a solver, they sign an intent, the protocol batches those intents, and solvers compete on surplus. Consistency rewards then pay the solvers that stayed close to the best bid on market orders and settled on time. On the Ethereum consistency-rewards distribution: - BRRRolver took 13.4% of the week - Next were Fractal 11.4%, Sector 11.2%, Quasi 9.4%, and Kipseli 7.7% - The figure is not volume or batch wins, it is how consistently the protocol paid us for staying near the best surplus for users that week Consistency-reward share is the percentage of that weekly budget a solver earned, and how consistently it stayed in the market for users. dune.com/queries/7939178/119…
3
13
684
Desks care about implementation shortfall, fill reliability under stress, and who can see the order before settlement. This piece is more technical, but it is also how we operate. Not just BRRRolver, which we run on @CoWSwap, we already sit on that side of the trade: liquid book, solvers, MEV models. nitter.net/brlabsxyz/status/21023…
Institutional on-chain execution is an implementation-shortfall problem. For a desk selling size, the benchmark is the market price when the trading decision was made. Execution quality depends on the final proceeds after costs and the opportunity cost of any unfilled amount. Spread, market impact, gas, MEV leakage, failed execution, and delays can all affect that outcome. These effects overlap, so attribution matters. That is why “MEV protected” tells an institutional desk too little about how its order will actually execute. Several distinct mechanisms sit behind that label: 🔸️ Private order flow limits exposure to the public mempool. Privacy alone does not add liquidity, and participants receiving the order may still see it before settlement. 🔸️ Batch auctions can match compatible buying and selling interest, reducing the residual volume that needs to reach AMMs. 🔸️ Solver competition can bring together inventory, private market makers, multi-pool routes, and temporary financing. 🔸️ Just-in-time liquidity can add depth around a particular order. That liquidity may disappear immediately after the trade. Intents give the desk a way to define the required outcome: the assets, amount, minimum proceeds, and expiry. Solvers then compete to construct an execution that satisfies those signed constraints. Settlement checks enforce the order’s conditions; the order can still expire unfilled. Routing remains part of the job. A solver also has to assess whether a proposed settlement can be financed, remain economically viable, and reach the chain while gas prices, inventory, and pool states are changing. An attractive route at quote time may no longer be executable by the time it is included. The auction design matters, too. - @CoWSwap Protocol uses combinatorial batch auctions, allowing solvers to bid on individual orders or groups of orders. - @1inch Fusion uses a Dutch auction, with resolvers competing to fill as the offered rate changes. Both delegate execution to specialists, but create different incentives around price, timing, and order matching. At BR Labs, these are the engineering questions behind BRRRolver: combining route discovery with auction timing, gas costs, temporary liquidity, and the probability of a successful fill as market conditions change. Flash liquidity expands the set of executable routes by financing intermediate steps within a single transaction. It can reduce the need for prefunded inventory, while adding repayment and fee requirements that the settlement must satisfy. For institutional desks, the useful evidence is measurable execution performance: implementation shortfall against the decision price, fill rates and time to fill under stress, reliability across changing liquidity conditions, and who can access order information before settlement. Those measures show how much of the available liquidity a desk can actually use, at what cost, and with what reliability. markets.financialcontent.com…
3
20
786
New AI models are dropping left and right and I’m reminded of just how much this contrasts with our industry: - AI is closed: Even the frontier labs can’t always say how it works. - Blockchain and crypto are open: The code is readable, and every rule is on-chain.
1
7
749
One sharp builder who works with AI today beats three average ones. That’s not an edge anymore, that’s the baseline. And it’s not one model, the stack keeps moving: OpenClaw, Hermes, Claude, GPT, now tools/system one models like Jev.
7
805
In 5 years, the LPs will be agents. The retail won’t be sitting in the pool manually. Full disclosure: We already sit on that side of the trade: liquid book, solvers, MEV models.
6
786