Code as Counterparty
Crypto did not eliminate intermediation. It changed who intermediates, how they get paid, and who absorbs losses when the system breaks.
The first question any trader learns to ask is: who is on the other side?
On a funding desk at Citadel, that question was settled before a trade ever happened. ISDA agreements, margining frameworks, credit support annexes, all negotiated by humans over months before a single dollar moved. DTCC cleared the trades. Prime brokers custodied the collateral. If something broke, there were phone numbers to call and people who could use judgment: waive a margin call, extend a deadline, absorb a loss.
The first time I executed a trade in crypto, nobody was on the other end. Not a person, not a desk, not a counterparty with a name. The other side was a smart contract, executing logic written by a developer I would never meet. The margin was managed by a liquidation engine, not a margin clerk. The settlement was atomic, not T+1. And if something went wrong, there was no one who could decide to make an exception.
That is not a flaw. That is the design. And it produces three consequences: everything moves faster, including panic; human discretion disappears, replaced by mechanical execution; and when something breaks, there is no balance sheet standing behind the system to absorb the loss.
The views expressed are solely my own and do not represent the views of Ava Labs, the Avalanche Foundation, or any affiliate of either organization.
This newsletter is for informational purposes only and does not constitute financial, investment, legal, or tax advice.
Nothing herein should be construed as an offer or solicitation to buy, sell, or hold any digital asset or financial instrument.
Speed: Everything Clears Faster, Including Panic
On October 10, 2025, President Trump announced 100% tariffs on Chinese goods. Bitcoin fell from $122,000 to $105,000 in hours. In equity markets, a move of that magnitude triggers circuit breakers designed to give humans time to think. Crypto has no equivalent. No committee meets. No bell rings. $19.13 billion in leveraged positions were liquidated in 24 hours, 1.62 million accounts wiped clean. At peak intensity, $6.93 billion was liquidated in 40 minutes: an 86x acceleration, executed entirely by code.
But speed cuts both ways. On the night of February 28, 2026, when the US and Israel struck Iranian nuclear facilities and every traditional market on the planet was closed, crypto did not wait. Hyperliquid's oil perpetual surged 5% and recorded $1.99 billion in daily volume, with no traditional market open for reference pricing. Crypto was the only functioning price-discovery mechanism on Earth. October 10 showed the cost of machine speed; the Iran weekend showed the benefit.
Discretion: When the System Cannot Make Exceptions
At Citadel, when a bank counterparty faced a margin shortfall, there was a process: a phone call, a conversation, sometimes a waiver. That judgment introduced delay and inconsistency, but it also acted as a shock absorber. A margin clerk who waits until morning is adding friction. That friction, in aggregate, is what prevents cascades.
Crypto's liquidation engines have no such capacity. Funding rates on perpetual swaps, 78% of derivatives volume in 2025, adjust mechanically. The engine monitors margin ratios and, when thresholds break, liquidates automatically, instantly, and without appeal. Crypto even introduces auto-deleveraging: when losses exceed the insurance fund, exchanges forcibly close profitable positions on the other side to plug the deficit. Your hedge worked, and the exchange closed your winning trade to cover someone else's loss.
In my view, this is the single most important structural difference between these markets. Not volatility. Not leverage. It is the absence of discretionary shock absorption. When markets need exceptions, the machine cannot provide one. During October 10, Binance's risk system rejected exit orders for over 100 minutes; Hyperliquid, fully on-chain, maintained 100% uptime with zero bad debt. Different code, different outcomes. But in neither case was a human making a judgment call.
Accountability: Who Absorbs the Loss?
When a bank fails, the chain of loss absorption is long and clear: shareholders, bondholders, FDIC, central bank. Every traditional intermediary is a shock absorber with capital committed to the market. The new intermediaries capture fees too, but often without the same obligation to stand in during stress.
When the JELLY exploit hit Hyperliquid in March 2025, validators voted to delist the token and reset the oracle price. It worked, but it was a human intervention on a system marketed as decentralized, and the vault's TVL plunged from $540 million to $150 million as liquidity providers, who had no obligation to stay, withdrew. That is the accountability gap. Permissionless liquidity is rented, not committed. And rented liquidity disappears under stress: order-book depth fell more than 90% during October and stayed 40% below pre-crash levels months later.
What Must Remain Human
The thesis I carried from Citadel to crypto was simple: better plumbing. Faster settlement, lower cost, broader access. I still believe it. But October 10, the JELLY exploit, and the liquidity that never fully came back reveal what that thesis underestimated. Intermediaries do not just add cost. They add discretion, accountability, and committed capital.
The real question is not "code or humans." It is: which functions of intermediation can be safely automated, and which will markets insist remain human? Settlement can be code. Custody can be code. Price discovery can be code. But the capacity to pause, to absorb, to exercise judgment when the model breaks, that may be the function that resists automation longest.
Next: Why Crypto Still Feels Early.
Drop me a line. What do you think? I'd love to hear it: akram@span.blog.