A Conversation with Robinhood co-founder and CEO Vlad Tenev

The limitation of LLM training on internet data is that it's predictive and pattern-matching rather than reasoning from first principles; current LLMs are 'trained on the whole internet, there's a lot of garbage out there' but Aristotle's synthetic data approach allows it to create new knowledge.

“I think about this in the sense of, you know, the models are trained on the whole internet. There's a lot of garbage out there. Uh they're predictive in the sense of they're really uh you know to the walk me through maybe the prediction versus reasoning pure reasoning from first principles”

Successful e-commerce companies like Amazon become the default first place to buy things because they do a good enough job that even though you could find something cheaper elsewhere, the convenience and experience make them the first choice.

“in some cases if you look at the really successful e-commerce companies like an Amazon what they've been able to do is do such a good job that they've become the default place you go to buy stuff. And sure, maybe at this point if you want a comparison shop and you want to buy a new belt and you really wanted to like save on on the belt, you could probably find it cheaper somewhere else. But they do a good enough job and they serve you well for such a large majority of things that they become the first option.”

Robinhood grew from 700 employees and ~$200 million annual revenue at end of 2019 to thousands of employees and close to $1 billion in revenue by end of 2020, a 3x+ growth, driven by COVID's impact on retail trading.

“we went from I'd say end of 2019 we had 700 people and something like 200 200 200 and change million in annual revenue. To in the next year end of 2020 we had thousands of people and close to a billion in revenue.”

Robinhood recently launched Robinhood Banking with real-time cash delivery (available in New York) and products like child savings accounts, joint accounts, and family portfolio management—differentiating from traditional banks that penalize checking account usage.

“a couple weeks ago we started rolling out Robin Hood banking. And if you look on social media, Robin Hood banking is a hit... real-time cash delivery uh which is available in New York... child savings accounts, joint accounts... a family first experience where you can actually manage the finances of the whole family in one place.”

As Aristotle improves, the length and complexity of provable statements will grow exponentially—from 10-page proofs today to 100-page proofs within a year, to 100,000-page proofs within 3 years—enabling it to tackle the deepest unsolved problems in mathematics (like unified field theory in physics) and create knowledge that no human could manually verify.

“right now, let's say you can easily at low cost produce a proof that's 10 pages long. And actually, we can produce longer ones, but just uh as an example, per per unit cost and time, you can produce a 10-page proof. Well, in a year it'll get to 100 pages. In three years it'll get to 100,000 pages. And you can say, okay, what type of problem has a 100,000 page proof? That's probably some of the deepest problems in in mathematics.”

Success, for Tenev personally, is creating dramatically more value for the world than you create for yourself—his legacy would be enabling broad-based equity ownership so that when he tells his grandchildren about it, he can speak proudly about building a more stable and prosperous society.

“For me personally, I think success is um creating dramatically more value for the world than you create for yourself...I would feel good if the aggregate impact on the rest of the world...is much bigger than what happens to me personally...that's a legacy that I would get excited to tell my grandchildren proudly uh about when they're, you know, huddled around the the fire on Christmas.”

Robinhood's mission to give people 'skin in the game' through broad equity ownership is a hedge against communism and authoritarianism, because when people own a stake in the capitalist economy they become invested in its success and stability, whereas when people are excluded from ownership they become vulnerable to revolutionary sentiment.

“you know I love the idea of owning a part of the American dream in that sense...I think about that in terms of like how do we get more people invested in capitalism and one of the ways is is ownership of assets...if if I can play a part in doing that across privates, publiclix, all post IPO and make it so that more people own uh the great industries of of this country will lead to not just smarter and more well-off individuals, but also a more stable and and prosperous society.”

If there had been an easy way for Tenev's grandparents to invest and protect wealth during Bulgaria's collapse, it would have been much better for the country and the family, suggesting that access to productive assets (stocks, real estate) enables resilience during economic crises.

“I I think that if there was an easy way to invest and to protect your wealth, um it would have been much better for for the country uh at the end of the day because you know the the country got set back.”

Robinhood has built data systems that are easily queryable and hygenic (one place, clean interfaces, validated), which incidentally made it much easier to plug in AI models later without vendors having to do custom integration work.

“from a pretty early point, we wanted to make our data easily queryable internally because we wanted to run good analytics to understand what was going on. And so once your data is queriable, it makes it so that you you generally have good systems hygiene. It's, you know, in one place. Uh it's like got clean interfaces. You spend time making sure that it's correct. Um and and I think the work that we did not really thinking about AI made it a little bit easier to plug plug in these models”

Robinhood's AI adoption strategy focuses on two high-ROI functions—customer service and software engineering—where metrics (AI deflection rate for support, percentage of code commits from AI) can be precisely tracked and measured to ensure AI is genuinely increasing productivity rather than merely increasing automation percentages.

“what we told the team from the very beginning is look, there's two areas where we want to start with and be absolute best-in-class in our space. And that's customer service and software engineering. Because if you think about what really moves the needle, those are the big teams that have multiplicative impact...we've done so much innovation there. By and large, if you interact with customer service on Robin Hood, you're you're interacting with our AI agents.”

Good marketing communication starts from first principles and explains why a concept (like prediction markets) is important for society, not by listing features; the CEO must personally be involved or the story doesn't get told.

“If you're someone who has never heard of a prediction market for example, how do you explain what a prediction market is and why why it's innovative? Why it's important for society for these things to exist? We we start there and then I think once you have the foundation, you try to make sure all the products plug into that foundation.”

The second and third-order consequences of private market tokenization are: (1) easier capital raising for entrepreneurs via digital fundraising rather than person-to-person pitching, (2) secondary liquidity for early employees and executives to sell shares without complex paperwork or waiting for IPO, (3) explosion of entrepreneurial activity because founders can spend less time fundraising and more time building.

“I think the uh second order consequence is that for entrepreneurs the other side of the market uh you'll have an easier time raising capital. So rather than going through this opaque process of spending a lot of time meeting investors one-on-one, pitching in person, uh you could imagine a digital fundraising process...secondary liquidity will be much easier...you'll be able to sell easily on a transparent market without the complexity of dealing with paperwork. Um yeah, or or waiting till IPO. So I I think those are the things and I think the third order consequence will be uh an explosion of entrepreneurial activity”

Fixing organizational problems requires undoing previous decisions, which is psychologically painful because it means admitting you were wrong, so people tend to undo things only gradually and incompletely, never capturing the full benefit of the change.

“fixing things everybody says that but we're it sort of like meets difficulty is that means undoing something I've already decided and that means admitting that I was wrong and so people sort of like tend in general not everybody they tend to like slowly sort of like oh I'll undo little bit of it and then I don't get the any of the results.”

The interviewer suggests killing wellness days by making a public show of admitting error on something small (like office snacks or Sichuan catering on Wednesdays) to practice reversing decisions, so that leaders can build the confidence and cultural permission to undo larger bad decisions.

“usually things uh things get easier if you do them multiple times and uh you could practice doing this once and making a big show of it. Like actually maybe you take a small thing that you were wrong about that you wanted to undo and just say I want to tell you about something that I completely screwed up. It was a wrong decision and now we're taking it back.”

Tenev vividly remembers being shocked by the abundance of goods in the US grocery store upon arriving at age 5, specifically the availability of bananas which were a luxury delicacy in Bulgaria that had to be sourced from Cuba, and this contrast shaped his appreciation for market systems and abundance.

“when I came to the US and one thing that shocked me was that there were bananas in the grocery store. Like you go to the grocery store and you'd see the big thing of bananas. And in Bulgaria bananas were like a crazy delicacy. Well, because you had to get them from Cuba.”

Math training serves as cognitive weight-training for business decision-making because no business problem is as complicated as solving advanced mathematics—if you can endure the mental pain of struggling on a hard math problem for 12 hours, ordinary business challenges become manageable by comparison.

“I think math is a good way to train your brain into uh doing hard things, right? And it's almost like if you can deadlift 500 lb, then picking up your crawling baby uh from the floor is very very easy, right? So no business problem is as complicated as solving a really really hard math problem. So if if if you if you get used to the pain and the suffering and like the mental uh stress uh of beating your head against the wall for 12 hours on one single math problem...Um they think that's really really good training. It's like going to the gym for uh for business problems.”

Robinhood breaks AI-powered customer support into three phases: Phase 1 (help center in chatbot) is what most companies do; Phase 2 (AI reading customer account data) is used by ~10% of companies; Phase 3 (AI taking write actions like refunds) requires deep backend integration and is used by very few companies, and Robinhood operates at Phase 3.

“Phase one is uh a company puts the help center into the AI chatbot...Most companies doing AI for customer support are in this phase...Phase two is all right, you're getting a little bit more sophisticated. You're not just taking the data from your help center...the AI can actually go into the database...you can pull that data and you can use it to provide better contextual support, but it's still read only...rough numbers, let's say 90% of companies are in phase one. 90% of the ones that aren't in phase one are probably in phase two. Then you have phase three, which is okay, now I can actually do non-readonly actions...That's where Robin Hood's in.”

Robinhood's core operating values are: (1) High Performance—hiring elite talent and rewarding impact over headcount, (2) Safety Always—never compromising on security and compliance despite pressure to move fast, (3) Lean and Discipline—constantly asking how to do more with less, and (4) emphasis on early-career talent to maintain connection to younger generations and avoid becoming a generational company.

“Our values are high performance...we make it clear to not just employees, but anyone applying for a job that this isn't meant to be like a cushy, chill job...Safety always...you can't use that as an excuse to cut corners or compromise on uh the security of customers and and their money...Lean and discipline”

The 'Make IPOs Great Again' roundtable in New York (recent, 2 days before this interview) identified systematic improvements to the IPO process across all stages, recognizing that the IPO process has become over-ossified and suffers from bad branding, creating an opportunity to make going public easier and more attractive.

“There was actually a round table that I was part of here in New York uh two days ago. It was called make IPOs great again. We had the red hats and everything um at the New York Stock Exchange. And I think that there's improvements. the IPO process has gotten so oified and the branding of it is bad at all stages”

Robinhood's partnership with Sage Home Loans offers a 75 basis point discount off the national average mortgage rate because Robinhood has superior customer financial data (complete transaction history, account information) and can use AI to generate better predictive models of repayment risk than traditional underwriting.

“we have a mortgage uh partnership with Sage Home Loans that gives you 75 basis points off of uh the national average uh because you have better information uh about people too, right? you have history, complete history information, and you can use that with AI, I would imagine, to actually probably generate better predictive outcomes than humans.”

Robin Hood's COVID hiring bulked up the company and added fat (increased headcount and operating costs unsustainably), but the 2022 contraction allowed a 'massive leaning out' similar to a bodybuilder bulking then cutting, resulting in a healthier, more disciplined company with lower structural costs.

“if you want to get really really strong, um, there's one way where you just like gradually build up muscle mass and keep your fat uh, low over long periods of time. But what what Robin Hood did was we bulked up gigantically and also gained a lot of fat in the process and then we did like a massive leaning out.”

Product events (like the December 16 'Yes No' event on prediction markets and AI) serve as a forcing function for clear communication because communicating to 20 million people requires distilling products to their essence, preventing the trap of jargon and excessive detail that founders naturally fall into.

“I think the events themselves are a good forcing function because if you're communicating a new product to 20 million people, it really forces you to distill it into the essence of what 20 million people can understand”

Bulgaria's economic decline started around Tenev's birth year (1987) and continued for decades, partly because of poor policy but also because people lost confidence in the future and became pessimistic about raising children, investing in the economy, and building businesses, suggesting that economic psychology and optimism are as important as policy in determining outcomes.

“the year I was born, 19 1987, maybe the time around that was was the heyday of the country and then for a long time it was just a gradual decline...so much of it starts from the will of the people and and and how optimistic they are about the future being able to raise children uh the economy”

Robinhood's north star is maximizing direct equity ownership from retail across the world; this goal drives decisions about expanding into public stocks, private markets, younger investors, and geographic expansion.

“our northstar is really just maximizing equity ownership uh direct equity ownership from retail across the world, right?”

Robinhood uses a simple traffic-light performance review system (green, yellow, red) in weekly leadership meetings where red goals receive scrutiny and collective problem-solving, creating psychological safety for acknowledging setbacks by treating them as information rather than failure.

“I like a very simple mechanism for uh sharing progress on goals. It's either green, yellow or red. If it's green, we don't really have to talk about it. If it's red, I think it deserves some scrutiny...I have a gavvel and I I hit the gavvel on the [gavel sound]...it's really nice to lighten up some extremely serious things so that people actually enjoy uh talking openly about, you know, goals that things that aren't going well”

Robinhood Ventures launched to give retail investors access to late-stage private companies through a closed-end fund currently in SEC quiet period, with the goal of democratizing private market access the way Robin Hood did for equities and options trading.

“We launched this thing in the US called Robin Hood Ventures which aims to solve this problem and we have our first closed end fund which we filed to go public which is in the quiet period right now. So I can't talk in much detail about the mechanics of that but I think the the biggest opportunity and the biggest problem is in private markets”

The next big thing for increasing stakes in capitalism is real estate ownership, which is currently too expensive and illiquid for most ordinary people, but could be tokenized or fractionalized to make it accessible.

“I mean, we talked about real estate in the past. Um, and I want to, you know, I'm a I'm a proponent of a diversified portfolio. So I I don't want to say one asset is better than another asset. Um I think that if if we make it easier to I I I think the general approach that we have is look at what wealthy folks have access to, what are the tools they have to protect their wealth and and grow their wealth.”

Multiple specialized AI models will likely dominate in different domains rather than a single monolithic model, because advantage flows to whoever has the best domain-specific training data—Harmonic has mathematicians using Aristotle (math data advantage), Anthropic is strong in coding, Google has highest-quality web data and unlimited compute, which explains why Gemini recently surpassed ChatGPT.

“I think it depends on uh I I think there's going to be multiple models and it's really going to depend on the data that is used to feed them. So for example, one thing that's great about Aristotle which is harmonics model is you've got mathematicians that are using it to ask very complex math questions. So if you if you've got all the mathematicians using the model then you basically have an advantage with mathematics data”

Robinhood doesn't want to be a mortgage originator or hold mortgages; instead, it wants to act as a network that helps customers find the best rates by allowing different banks to compete.

“I don't think we necessarily want to get into the mortgage underwriting business or holding mortgages, but we we want to get we want to help people with all of their financial needs. And I think for that one, being a network and allowing different banks to compete over who can offer you the the best possible rate as a consumer is is probably the road we're going to keep going on.”

Robinhood makes it easy to remove low performers by quickly assessing fit (sometimes within 3 weeks) rather than prolonging tenure, and most companies fail to do this because process complexity prevents swift action on obvious mismatches.

“We want to make it I mean, sometimes it's pretty obvious that uh you know, for for whatever reason, um we made a hiring mistake usually that that someone's not a fit. And at that point, we want to make it easy as easy as possible. So yeah, if it's three weeks, doesn't make sense for us or for the person to to to continue...I think what most companies suffer from is process getting in the way and actually making it very very difficult for people to uh to to get rid of low performers.”

If banks suddenly started paying 5% interest on checking accounts, their earnings and profitability would collapse because checking float is such a large revenue source; this creates a safety-and-soundness problem that regulators and the bank's board would resist.

“How do you adapt if you have trillions of dollars in checking that you're used to paying to to earning 5% on and you suddenly are talking about giving the majority of that back to the customer and and taking no profit. Um, you know, it would... your earnings go down, profitability goes down. if you're public, you know, that that could have an impact on the stock price and suddenly it becomes a safety and soundness concern.”

Pressure from competitors like Robin Hood pushes incumbent credit card companies to modernize, which is good for them and good for consumers, even though the incumbents are being pressured to adapt.

“if we can put pressure on the the credit card companies to start modernizing, that'll be good for them and for the consumer. So um but you don't get 3% like in a transaction fee.”