The Citadel Alum Reshaping The World Of Trading

Summary

• The core product claim: Lighter built custom ZK circuits specifically for finance rather than adapting general-purpose ones, and the compute needed to prove them “is actually only 1% of what you need for a general purpose computer.” Vlad’s framing of the prize: finance is 30-40% of the economy, so “imagine you find a way to run 40% of the economy with 1% of the cost.” He claims Lighter is “stronger on all of those dimensions” — low cost, low latency, verifiable, secure — “than pretty much any of our competitors,” while crediting earlier entrants with a “step function change relative to what came before them.”
• The CeFi-to-DeFi call: the top five to ten decentralized venues “are stronger than centralized exchanges for most customers, and that’s where I think most of the shift will happen — from CeFi to DeFi.”
• Value accrual is unambiguous by design: Lighter is a US C corp with no dual foundation, the pre-token round was committed as “the last equity round we’ll ever raise,” and only 1% of the cap table took the opportunity to get out (the round was 5x oversubscribed). “All value is accruing to the token” — the host summarizes that equity gets no revenues or profit, and Vlad responds “Okay” — and the desired end state, conditional on the Clarity Act passing and the SEC continuing to innovate, is that the token would also be equity.
• The Madoff anecdote is the thesis in miniature: in Vlad’s first week at Citadel in 2004 he flagged Madoff’s suspiciously high ratio (likely Sharpe) and Ken Griffin replied, “Oh, no, that’s a Ponzi. Don’t worry about that one” — five years before it was uncovered. Madoff survived an SEC check by handing over a fake DTCC number nobody ever called; on chain, “all those trades would be on chain, anybody could verify it.”
• Roadmap: real options on chain in Q3 of 26, sharing one balance sheet and risk-collateral model with perps and spot inside Lighter — and the stated goal is that users trade those options through Robinhood, whose partnership Vlad attributes to tech diligence, Ethereum interoperability, and forward-deployed engineers.
• A surprise worth noting: ZK could deliver dark pools and trade privacy, but “I’ve actually been surprised that there’s not as much market demand for that as I would have thought” — Lighter uses ZK for scaling because privacy “hasn’t been much of a customer request.”
• Founder mechanics: the Lunch Club-to-Lighter pivot retained 80% of the engineering team by running “an internal YC” — three projects in parallel and an internal demo day — plus healthy runway and a problem space (cryptography, scaling, quant finance) the team actually preferred.

Deep dive

1. Olympiad kid to Harvard at 16 — a 94 when the class average was 37

• Vlad was programming at eight — writing QBasic and Pascal bots to practice card games like Durak so he could beat his family — but says competitions, discovered at 12, are what unlocked him: “I didn’t really know I was [competitive] until I discovered these competitions… you have a very clear feedback loop and you can practice and get better.” They were also his first community, competitive but much more collaborative.
• His Harvard concentration logic is a portfolio decision worth hearing as told: programming skill had “already kind of hit diminishing returns,” so college was for learning what to build. Physics had received substantial funding but wasn’t necessarily growing very fast, biology meant lab work he disliked, and economics had “a lot of exciting stuff going on” — option pricing, portfolio theory — where his math and programming applied.
• The origin of his first job: a microeconomics midterm score of 94 that he assumed was a near-miss A (“in high school it’s just based on the absolute value”) turned out to sit against a class average of 37 — one of the highest standard deviations the course had ever seen. The professor recruited him into research on the spot. He finished Harvard in two and a half years.
• His own framing of the network he fell into: “like Forest Gump if he had been a math person” — Informatics Olympiad friend Adam D’Angelo (Zuckerberg’s high school roommate, first Facebook CTO, now on what was likely OpenAI’s board), two physics-camp peers who co-founded Anthropic, and somebody from his high school who was likely Vlad Tenev of Robinhood.

2. Ken Griffin closed him at 18 — and called Madoff a Ponzi in 2004

• On the recruiting myth: “I wouldn’t say he personally recruited me, but he personally closed me.” Griffin’s pitch was disarmingly direct — the option strategies Vlad had been trading on his own in college “did work for many years, but I just arbitraged them away, so that stuff doesn’t work anymore. You should instead come work with us.” This was right after markets went electronic, as Citadel built its first trading desks.
• The episode’s best story: first week on the job, reviewing competitor track records, Vlad asked why they weren’t reverse-engineering Madoff given his outlier ratio. Ken’s answer — “Oh, no, that’s a Ponzi. Don’t worry about that one” — came five years before it was uncovered. The SEC did check Madoff, but he gave them a fake DTCC number “and they never made the call. Had they just picked up the phone, it would have been uncovered.”
• That’s Vlad’s whole case for on-chain finance: Madoff never traded at all, and detection “literally depended on somebody picking up the phone.” On chain, “if you’re up 10% because you bought Microsoft at 30 and sold at 40, all those trades would be on chain — anybody could verify it.”
• The hosts’ pushback — people don’t want their trades public — draws a candid concession: ZK proofs can prove trades without revealing them, dark pools are buildable, “but I’ve actually been surprised that there’s not as much market demand for that as I would have thought.” Lighter uses ZK for scaling; privacy “hasn’t been much of a customer request,” and firms trading on chain today “still have a lot of alpha in that environment.”

3. The Silicon Valley detour, and lessons that used to be contrarian

• He left Wall Street because quant trading, however interesting, meant “we’re not really building new products and new markets” — while friends like D’Angelo were. The path ran Quora (head of ML), then heading engineering at Addepar under Joe Lonsdale, then co-founding Lunch Club with fellow Olympiad alum Scott Wu.
• Two ideas he watched go from crazy to obvious: Lonsdale’s forward-deployed engineers as the tip of the customer-success spear, and the Robinhood team’s zero-fee retail product — “at the time that was considered crazy.” His meta-point on advice: the good ideas are “being contrarian and right — and then it’s no longer contrarian,” yet “there are probably 10,000 people learning that framework every single day because they’re maybe 18 years old.”
• On the elite-circle question: “It’s one big club, but it’s an open club… more merit-based. If you’re really great at customer success, at operations — you’re welcome to the club.”

4. The pivot playbook: an internal YC kept 80% of engineering

• After board approval to pivot, Lunch Club didn’t decree a direction — it ran “an internal YC”: three projects in parallel, an internal demo day, pitch competitions, so “even the least experienced person on the team” could have set the company’s course. Add healthy runway (“not down to the last three months of funding”) and the result was 80% engineering retention through a total product change.
• The third retention factor is the telling one: the team had joined Lunch Club for the people, but Lighter’s problem space — cryptography, scaling systems, quant finance — was what they’d actually wanted to work on.
• His pushback on “nobody cares about tech, it’s all distribution”: distribution and UX depend on tech — Telegram Wallet and Robinhood “absolutely did diligence on the tech,” and users who never learn how you shaved 100 milliseconds off execution “absolutely care about that happening.” The Bitcoin analogy: few users understand it, “but they know that it does work, and they know that if it broke somebody would find an error.”

5. Lighter’s edge: 1% of the compute for 40% of the economy

• The engineering frame: an exchange is a three- or four-dimensional optimization — low cost, low latency, verifiable, secure — where gains usually trade off, but “sometimes you find these really cool ideas where all of the metrics improve… code that’s actually cheaper and faster and more secure. That’s the stuff that really gets engineers going.”
• The breakthrough was realizing they could build custom ZK circuits for finance instead of adapting general-purpose ones — cutting proving compute to 1% of a general-purpose computer’s. Since finance is 30-40% of the economy: “imagine you find a way to run 40% of the economy with 1% of the cost.”
• Asked directly about Hyperliquid, he claims Lighter is “stronger on all of those dimensions than pretty much any of our competitors” while giving earlier entrants credit for “a step function change.” The bigger call: he thinks top-five-to-ten decentralized venues are stronger than centralized exchanges for most customers, so he thinks most of the shift will happen from CeFi to DeFi.
• Ethereum is load-bearing: “the most secure ledger for DeFi… the on-chain equivalent of a clearing house,” with “no hiccups at all” in ten years — migrating off is “not in the top 10 or the top 100 things” on the roadmap. Next up, on-chain options in Q3 of 26 on the same balance sheet and risk-collateral model as perps and spot, ideally tradeable through Robinhood — a problem he calls simultaneously technical, marketing, and partnerships.

6. One token, no ambiguity: all value accrues to the token

• Lighter’s structure is a US C corp from day one, with no dual foundation, built on the thesis that “regulations are slow to come around but they will come around — and that’s been happening.” The pre-token round was committed as “the last equity round we’ll ever raise”; despite 5x oversubscription, only 1% of the cap table took the opportunity to get out. The host asks whether equity gets no revenues or profit; Vlad responds “Okay.” “All value is accruing to the token,” and the team is all-in on it.
• His end state, hedged as conditional on the Clarity Act passing and the SEC “continuing to innovate”: equities themselves get tokenized, so Lighter’s one token “would also be equity… then there will be no ambiguity about these things. From our perspective we’re already there.”
• The closing note on his Twitter presence: he describes the feed as “Q&A” between meetings — “pointing out the absurdity of this industry. The responses are even more absurd than the thing I was trying to make fun of in the first place.”