LLMs didn't make anyone a quant. They removed the excuse for not being one. I tested this myself recently. No coding background to speak of, no quant experience in any formal sense — just curiosity and a Claude subscription. Over a few evenings I ended up with something that worked: a data pipeline, a backtest, an execution layer wired into a CEX, basic risk monitoring. It wasn't pretty and it wasn't going to compete with anyone serious, but it ran. And it produced something that at least resembled signal. Eighteen months ago that would have taken me a team and six months. Now it took a weekend and a chat window. The math was never the bottleneck for most people trying to go pro. The scaffolding was. And the scaffolding is now free.
This is a real shift, and worth taking seriously rather than dismissing. The barrier to entry into quantitative crypto trading has fallen further in the last eighteen months than in the previous ten years combined. Whatever happens next, it happens on top of that fact.
The honest question isn't whether anyone can be a quant. It's what kind of quant the world is about to be flooded with, and what that does to a market.
Crypto is the first asset class that will fully feel the answer. Lower friction than equities, more retail-accessible venues than FX, open data that anyone can pull without a Bloomberg terminal, and zero gatekeeping at the strategy level. Every condition that historically slowed the diffusion of quant techniques into a market is absent here. If LLMs are going to compress alpha anywhere first, it's crypto.
The compression is going to be brutal
The naive read is that this produces a flowering of alpha. A thousand new strategies blooming, more efficient price discovery, healthier markets. That isn't what happens.
What happens is the opposite. The easy strategies (funding rate carries, basic basis trades, obvious mean-reversion plays on liquid pairs, cross-exchange arbitrages on the top ten venues) get found, deployed, and compressed to zero faster than in any previous cycle. The half-life of a textbook strategy in crypto has been falling for years. LLMs are about to take it down another order of magnitude.
This isn't a controversial prediction. It's just what happens when you reduce the cost of implementation by 100x in a market that was already converging. The strategies that defined returns in 2021 to 2024 are going to look like the equity stat-arb of the early 2000s by 2027. Real, once, gone.
What survives, and who keeps it
Efficient doesn't mean edgeless. It means the surviving edges are different, and they cluster around things LLMs cannot give you.
Two of them belong to the largest firms on earth and have for decades. Capital is one. Strategies that need real balance sheet (basis at scale, market making, structured trades that require warehouse capacity) don't get democratised by an LLM. You can prototype them in an afternoon and still not run them, because the constraint was never the model. It was the book behind it. Latency is the other. The execution loop is still bounded by physics and infrastructure. LLMs don't operate at microsecond timescales, and the parts of trading that do are still won by colocation, network paths, and matching engine relationships.
These are the moats of Jane Street, Citadel, Jump, HRT. It's not an accident that those firms are now printing tens of billions in net profit per year. They are sitting on the two inefficiencies that LLMs make more valuable, not less, because the rest of the market is converging toward them and away from everything else. The democratisation of quant skill is, in a real sense, a transfer of value from median quants to the firms with capital and infrastructure.
There's a third surviving edge: risk architecture. The lesson of October 10 wasn't about strategy. It was about plumbing. What survived weren't the cleverest books. They were the ones whose collateral lived outside the venue's risk perimeter, whose risk engines could see cross-venue exposure in real time, whose leverage was sized against the portfolio rather than against a matching engine. None of that is something an LLM gives you.
And then there is the one nobody talks about, because it's unglamorous and expensive: liquidity fragmentation across crypto venues.
The fragmentation problem is crypto-specific
This isn't a settlement cycle problem. Crypto venues mostly settle the same way. It's a capital deployment problem, and it's unique to this market in a way most quants underestimate until they try to scale.
Liquidity in crypto is split across dozens of CEXs and a long tail of DEXs. The inefficiencies (the real ones, the ones that haven't been compressed yet) increasingly live on the venues most firms won't touch. Second-tier CEXs with thin order books and real mispricings. Mid-cap perp venues running their own oracle. DEX pools where price moves before CEXs catch up. The alpha is genuinely there. The problem is that capturing it requires prefunded capital on a no-name exchange, and almost nobody is willing or able to put serious size there.
The result is a market where the strategies that work the best are the ones nobody can deploy. An LLM will tell you the trade. It cannot tell you how to fund the venue, how to manage the custody risk, how to repatriate the PnL, or what to do when the alpha moves to a different exchange three days later, which it will. Crypto alpha is unusually mobile. A strategy that works on Venue A this week may have decayed and migrated to Venue C by next Friday. Without a credit line or balance sheet capacity to follow it, you're trading yesterday's edge.
This is the inefficiency that doesn't compress. It is structural, not informational. LLMs cannot solve it because the constraint isn't intelligence. It's the working capital required to maintain simultaneous presence across fragmented venues, the operational architecture to do it without losing on prefunding drag and custody risk, and the credit capacity to amplify a working strategy before its edge migrates away.
The firms that solve this don't out-think the market. They out-capitalise it, out-infrastructure it, and out-architect it.
Bequant sits on both sides of this. We're a principal trading firm and a prime services provider, and the two were built together. The prime infrastructure exists because we needed it for our own book first: a single margin pool across venues, off-exchange settlement, a cross-venue risk engine, lending capacity to amplify a strategy before its edge migrates. Principal trading at scale in crypto isn't possible without it.
The way quants work with us is by coming on board as managers. They bring the strategy. We bring the capital, the infrastructure, and the credit capacity to deploy it across fragmented venues at size. It solves the two problems a competitive quant can't solve alone in this market.
LLMs are about to make a lot of people functional quants. They are not going to make anyone a competitive one. The competitive part was never the part the model could do.
#InstitutionalCrypto #QuantTrading #PrimeBrokerage #AIinFinance #DigitalAssets