You have 4 summaries left

Dwarkesh Podcast

Brett Harrison - FTX US Former President & HFT Veteran Speaks Out

Mon Mar 13 2023

Trading and Market Dynamics

  • Illiquid or difficult to access markets can result in price discrepancies, which firms like Jane Street aim to arbitrage over time.
  • Different trading firms establish niches in specific areas and use varying strategies to make money.
  • There are famous hacks and tricks used in finance, such as preloading TCP/IP messages to reduce latency when trading.
  • The goal of technology in fields like video games and quantitative finance is to produce something that looks good enough and is fast enough for the user, not necessarily the most theoretically perfect simulation.
  • The advantage to society as a whole of getting information faster through algorithms and optimizations is that it leads to smart, sophisticated technologies competing at the margins, which ultimately benefits liquid, tight, efficient markets.
  • It's difficult to imagine designing a market where everyone stops competing at a certain point; on average, people should be competing.
  • The value of milliseconds, microseconds, and nanoseconds in price discovery cannot be judged.
  • Instantaneous price discovery is always preferred.
  • There is a margin beyond which spending resources on getting tight spreads is not worth it.
  • Prop trading does not make up a large percentage of GDP, so the amount spent on technology for high-speed finance is not wasteful yet.
  • Working in high-speed finance helps estimate the probability of rare events but does not suggest solutions to them.
  • Solutions to rare events require collective effort.

Market Making and Matching Engines

  • Market making firms prioritize employee calibration and risk assessment skills.
  • Financial circles were quicker to recognize the risks of COVID and take action due to financial flexibility.
  • HFT firms had to adjust to remote work during the pandemic but remained profitable.
  • Matching engines can handle orders from different locations but may result in different prices than expected.
  • There is potential for a single market and matching engine for interstellar trading, but it would require changes.
  • Adverse selection can occur in markets where orders are sent from far away locations or co-location facilities.
  • Automated market makers on slow blockchains have established financial marketplaces by using an automated function that moves prices at the matching engine, resulting in a predetermined fair price based on prevailing liquidity.
  • Building a NASDAQ for interstellar markets is similar to building Uniswap now on Ethereum in terms of order magnitude and speed.
  • Periodic auctions instead of continuous matching could potentially help mitigate some issues related to adverse selection.

Hiring and Company Culture

  • There is a pipeline of orders from top universities to certain trading companies, but many talented individuals from lesser-known universities may be overlooked.
  • A resume from a graduating senior with no internship experience may not necessarily be a negative selection and could indicate the best person in their region.
  • Companies with strong brands and reputations for difficult interviews and good work environments can attract positive selections of candidates.
  • Jane Street should not replace OCaml with Rust due to existing infrastructure, but if they could start from scratch at zero cost, Rust would be a good choice for low latency distributed infrastructure that needs to never fail.
  • Jane Street pays interns a high salary because they believe in paying full-time employee rates for the time interns are actually there and doing real work.
  • The internship program provides on-the-job training and selects only the best candidates, making it more efficient than waiting for new hires to come on site and train them.
  • There is a free rider problem with internships, but companies like Jane Street work hard to make their experience as valuable as possible.

Personal Experiences and Perspectives

  • The speaker studied computer science and math in college and initially thought a career in programming would be isolating and unenjoyable.
  • They worked as an intern at Jane Street doing both dev and trading work, but ultimately found that they enjoyed the dev work more.
  • The culture of New York/Chicago quant finance is focused on extreme pragmatism and profitability, while Silicon Valley tech culture values creativity and innovation.
  • Cross-pollination between the two cultures could be beneficial for both.
  • Current culture around inconspicuous consumption is a function of millennial consumption habits.
  • People are focusing more on experiences than material objects, which has affected high earning tech and finance culture.
  • The social consequence of young, wealthy individuals is uncertain and will play out over the next decade or two.
  • Optimistically, younger generations amassing wealth may inject more dynamism into the distribution of wealth later on.
  • Being a quant or trader comes with a lot of responsibility, but there are many layers of safeguards in place to prevent major losses.

FTX and Alameda Research

  • Sam Bankman-Fried founded FTX with aspirations for the US business to grow the spot exchange and find a regulated path for bringing offshore products onshore.
  • FTX aimed to provide a predictable trading experience for customers through efficient collateral management, auto liquidation, and matching engine behavior.
  • Sam was conflict-evident and avoided direct confrontation with employees, leading to significant issues being ignored.
  • Nisad was the number two person at FTX and made many day-to-day decisions in Sam's absence.
  • Sam and Gary built a cross-collateralization system for trading that uses one wallet with all assets, appropriately haircut based on volatility and liquidity.
  • The success of the product came from Sam's initial intuitions as a trader on exchanges.
  • The company was deliberately falsifying information that went into audited financials.
  • It was difficult for anyone within the company, especially in the US side, to know what was going on.
  • Caroline Ellison became CEO of Alameda after leaving Jane Street where she may have been one of the speaker's bootcamp students.
  • The relationship between Alameda and FTX was separate, with Sam Trebucco no longer involved in Alameda.
  • Trading firms and investment banks often have physical separation between buy side and sell side.
  • Sam, the CEO of Alameda Research, was late in paying mid-year bonuses to employees, causing concern about potential liquidity issues.
  • Other red flags about the organization included an inability to hire more people, tightly controlled dev team access, and lack of separate C-level staff on the US side.
  • The speaker knew Sam when he was younger and saw a significant change in his mental and physical state while working at FTX.
  • People had different theories about what could be causing the deterioration, including stress, lack of sleep, diet, medication, or personality disorder.
  • The speaker gave Sam an ultimatum letter with three requests: more communication between them, hiring a separate management staff for the US side, and growing the tech team.
  • They had fewer than 10 one-on-one conversations during the speaker's time at FTX.
  • The speaker was constructively fired from their previous job after disagreements with management.
  • They started their own company, Architect, after leaving FTX and began fundraising in October.
  • The first inkling of potential trouble with FTX came from Twitter rumors about Binance buying FTX.
  • The speaker was busy with their own company's investment round when FTX announced they were being bought by Binance, causing confusion and shock.
  • Concerns about management and organizational issues were raised multiple times but not addressed.
  • Sam was successful at raising money and building a positive image for the company, but he wasn't leading day-to-day operations.
  • The credibility of FTX was based on experienced high-level executives that SPF had brought on.
  • It's unclear whose job it was to see the fraud coming, as it was expertly crafted to be hidden from those who could have done something about it.
  • The question of what to do when a major fraud occurs and is expertly hidden from those who could have prevented it is difficult to answer.
  • Companies should have a designated person for employees to bring complaints about the workplace and company strategy.
  • Employees should observe how their complaints are handled by the company, whether they take it seriously or retaliate against them.
  • If an employee experiences retaliation, they should start making plans to leave the job.
  • Regulators see crypto as a viable asset class that needs to come within regulatory envelopes appropriate for crypto.
  • FTX was seen as one of the companies willing to work collaboratively with regulators.
  • The investigators working on the ethics case moved quickly and had unsealed indictments within two months, which helped quell misinformation from spreading.
  • The talented people leaving FTX are in a position more like actual mafia people, but many of them came from impressive backgrounds prior to FTX and are expected to continue building something great.
  • Architect aims to build commodity software that traders can deploy out of the box to access multiple venues.
  • Architect's focus is to simplify the complexity of different exchanges and provide a good API for institutions and individuals.
  • The speaker believes that crypto is a viable asset class that needs better infrastructure for trading, regardless of its potential future use cases.
  • There is currently a large supply of job seekers in the tech and finance industries due to layoffs and fear around financial services.
  • The importance of efficient trading for commodities depends on the underlying use case, with some crypto tokens having more value than others.
1