Lighter 创始人 Vlad 与顶级交易员 Robert Chang、Marty 激辩｜TG Podcast

Mr. Vlad, welcome, man. How are you?

Hey, good to be here.

Yeah, dude. I'm a fan, man. I appreciate you coming on. The timing is pretty good on this one as well. I feel like you've been on a little bit of a media tour the last couple weeks here.

Yeah. It's funny because some of the stuff was actually scheduled weeks ago, so the fact that it's happening right after the crazy markets is more of a coincidence, but a good coincidence that we can talk through what's going on. The timing is honestly pretty good.

Yeah, welcome to the show. Do you want to start really quick? I'll give you a second if you want to give a quick intro into who you are for people who are not familiar, and then we could dive into some details.

Sure. Yeah. I'm Vlad, founder and CEO of Lighter, which is a perp DEX on top of Ethereum. I guess I'm not a crypto guy who got into finance; I'm more of a finance guy who took a detour to AI and fintech and then got into crypto. We can talk about all of that if folks are interested, but that's kind of a quick summary.

Hell yeah. I like the way you phrase that. I have a bunch of stuff for you and I want to get into some Lighter-specific stuff and your background a little bit, but I think let's start with what's most recent and relevant right now. Friday was, I think, the craziest day that most people here have ever experienced in crypto. Since Friday, I've had a bunch of traders come on the stream and talk about what they think happened, how they played it, why they think it happened, and what we do from here.

But I haven't had a founder, particularly one with your domain knowledge and expertise running Lighter, running a perp DEX, talk about what happened who was directly involved. Can you sort of set the stage on how and what happened on Friday?

Right. Yeah. One thing that's kind of interesting for me is that it was a little bit of déjà vu because I was in TradFi, first at Citadel and then at another hedge fund, during the 2008 crisis. The most volatile day of 2008 was October 10 as well.

Is that true?

Yeah. The S&P started out the day at 900, went all the way down to 830, and then closed at over 1,000. So that was a little bit of a déjà vu moment for me, but that was 17 years ago.

Whoa. I think, yeah.

On that exact date.

I was 6, 17 years ago on that exact date. What's the significance of October 10? Is it just a coincidence, the October 10 thing?

Well, October is always volatile. It's the most volatile month of the year. A lot of the big crashes happened in October historically. If you look at the biggest crash ever, in 1987, that was October. I think a lot of the big moves in 1929 that led to the Great Depression were in October also. October is the most volatile month in general.

But why the 10th? That's probably a coincidence, but that was definitely déjà vu for me. Ultimately, it all comes down to the same thing: people are overlevered and a lot of liquidations are happening. This kind of move—down and then up—means that people are actually getting liquidated from both sides, the longs and the shorts.

I can tell you what it was looking like from where we sat, because the markets were already pretty choppy all week. We saw that activity was elevated in terms of the TPS that we were observing, and all that, so we knew something was kind of weird in the markets. Our team is distributed, but I'm on the East Coast in the U.S., so I was looking at this around 3 to 4. There was a big sell-off from 3 to 4.

There's some alpha here: this is something that has existed for many years, but usually during these volatile days, the sell-off is accelerated in that last hour. That's actually a Sharpe 2 strategy if you just trade that, but that's beside the point.

Then it kept going in crypto. Obviously, TradFi closed, but crypto kept going. We were looking at it thinking, “Okay, these things happen.” Since we launched Lighter, there have been some days—I think February had some volatile days, and April had some volatile days. We just want to make sure the system works, addressing customer tickets and all that.

Then all of a sudden, we were watching our internal market maker that the community can buy, called LLP. LLP was flat, flat, flat, and then it was up 40 basis points; 5 minutes later, it was down 3%. Our first thought was that this was an attack, because there have been attacks on liquidity providers in the past. We had one back in March that we were able to resolve and made everyone whole. Hyperliquid had that JELLYJELLY attack. There was the Mango Markets attack from 2022, all this stuff.

Our first thought was that there was a coordinated attack on LLP, because all of a sudden it went from flat to up a little bit to down 3% in 5 minutes. We were really heads-down digging into that, making sure there was no attack going on. We saw that the markets were volatile, but again, it's not the first or only volatile day of the year. We were more worried that there was an attack going on.

After spending an hour investigating, we concluded that there was no attack and no market manipulation. It was just market moves—deleveraging and other stuff like that. I'm sorry to disappoint, but I don't have great insight into what caused it, because our responsibility is to our product and our customers. We wanted to make sure that we weren't being hit by an attack at that time.

Could you tell, leading up to this, that OI and maybe some other metrics were getting out of hand and that we were potentially susceptible to a flush like this? Was something like this on your radar or not really?

Well, yeah. I think I wrote a post about an exchange I had with Hib a month ago, where there was a discussion about volume versus open interest. I think Hib was talking about how it's healthy to have those in a good proportion, and his point was that if there's too much volume and not enough open interest, that's probably indicative of an unhealthy market, like wash trading and whatnot.

My point was that's true, but the other extreme isn't good either. If there's too much open interest and not enough volume, that means there are some really big positions just waiting to be unwound and not much liquidity. That's not a good extreme either.

I think on some other platforms, not on Lighter, maybe open interest was a bit too high relative to the liquidity. I think that's probably true. You actually don't want open interest to be too high either.

We can move on from Friday in a minute, but I'm just curious: does it make you feel differently about building a perp product after watching crazy liquidations and people get carried out on a day like Friday? Does it change how you feel about the product you're building and its value or importance to the market?

Right. Well, I feel that our mission is actually more important than ever, because our whole premise with Lighter is to build a system that is verifiable. The matching is verifiable, and the liquidations are verifiable. Every liquidation that happened on Lighter can be cryptographically proven to have been correct and to have followed the rules in our ZK circuits.

Now, I think there are certainly things we could have done differently, and should have done differently. One of those relates to liquidations. We have rules that govern how liquidations work, from partial liquidation to full liquidation to ADL. Those are all in the docs. They're all there. The math is all there, but they're at the bottom of a 100-page document. All of it is in the code; it's all transparent.

But my point is that I think that logic should be communicated more proactively to traders. A lot of people—not just on Lighter, but a lot of people—didn't even know what ADL was before Friday.

Yeah.

And it's not because people were trying to hide it. Every perp product needs some form of ADL. Everyone had that in the docs. It's not like it was a secret thing, but it just wasn't really brought to the forefront.

So I think, as it relates to people taking too much risk and not understanding what that can mean, we should be more proactive about explaining all those rules and how it all works. The whole point of perps is to give customers leverage, but you want to explain more clearly what the risks are and how liquidations can cascade. The industry can do a better job of educating traders on that, and we’re definitely happy to be a part of it.

Yeah, thank you for that. I wasn’t trading on Lighter or following it super closely, but there was all this drama in the wake of Friday that Lighter was down, people couldn’t trade on it, and their positions were stuck. Can you clarify exactly what happened tech-wise with Lighter during all this?

Right. This is an important point. I want to be clear about what happened, and I don’t want it to sound like we’re making excuses, because any kind of outage or system degradation is a problem for customers. I want to make that clear.

That being said, I think people conflated the market crash with the outage. The reality is that the outage happened 5 hours after the market crash. During the market crash, the system worked exactly as it was supposed to. We have 250,000 traders, so did a handful have some glitches? Sure, but that happens. For 99.99% of the traders, the system worked exactly as it was supposed to during the actual flash crash.

The system gave out and had an outage 5 hours later, during a pretty quiet time in the markets. The other thing that happened was what I alluded to before: even before the crash, we saw very high TPS in general. Lighter was functional, but performance was a bit degraded. It was glitchy for a couple of hours before the crash.

When we did the postmortem, it was clear that those two things shouldn’t have happened. We should have upgraded our infrastructure sooner, so neither the degradation before the crash nor the outage 5 hours later would have happened. But neither of those resulted in the kind of losses that would have happened during the actual flash crash if there had been an outage then.

So why was there an outage afterward? What happened?

Our plan was to upgrade the database over the weekend, which seemed reasonable. You want to upgrade during quiet time, from Saturday to Sunday. We had just launched publicly; before that, we had been running a private beta mainnet for 8 months. We had launched the public mainnet a week earlier, so we anticipated a lot more new customers and knew we should upgrade the infrastructure.

We had a planned infrastructure upgrade Saturday night. The flash crash happened 36 hours before that, so the old database was still there at the time. It survived the crash, but eventually, after 5 hours, it gave way. The database got into a bad state.

You’d have to talk to the database team about exactly why databases give out after a period of severe stress. There are a lot of very technical details there, but that’s what happened at the end of the day. All of this could have been avoided if we had just done the upgrade Thursday night and taken the hit of a 10-minute outage then.

Makes sense. Thank you for that. You mentioned LLP. Can you talk a little bit about how LLP is different from HLP, and why did LLP take a hit?

Right. At a high level, there are probably more similarities than differences. The differences came into view in particular this past week.

If you zoom out, LLP has had very strong performance. It’s up 221% year to date, with a Sharpe ratio of 5.5. It’s had drawdowns before. During those volatile periods in February and April, if you’re up high double digits or triple digits, it’s reasonable to have drawdowns in the mid-single digits. It had drawdowns of 5%, 6%, and 7% before, and it was down 5% on Friday.

I guess the question is, on a volatile day like this, you would think that because it collects liquidation fees, it should have made a lot of money, right?

And other similar vehicles, like HLP, did make a lot of money.

Yeah.

Let’s look at it as a zero-sum game. If you look at the total P&L of the day for all the participants, it’s zero. You have traders on the platform, other market makers, what we’ll call bad actors—the folks who might be doing manipulation, like the JELLYJELLY situation—and LLP. The sum of all of that is zero.

We ruled out the bad actors, so they were not a factor on Friday. Most of the other market makers left once they saw the extreme volatility; they just bounced. What that means is that the $20 million LLP lost was $20 million in positive P&L for all of our traders.

Is that a bad outcome? LLP had 15 up days in a row before that. The people who got in early in the year were up triple-digit percentages, and the people who got in around April or May were up high double digits. We think it’s the right trade-off.

We don’t know exactly why it happened that way, because LLP is less aggressive in how it applies the ADL logic. It favors the trader a bit more than the liquidity provider. It does charge a higher liquidation fee. About a month ago, we were getting criticism for why our liquidation fee was so high—1%. For most of the time, that was the industry standard. It started coming down on other platforms.

The trade-off is that having a higher liquidation fee means you have more funds in reserve when things get really bad. If things are pretty bad, you get liquidated and pay the 1%. But if things get really bad, you’re protected from ADL for longer. Eventually, if the market had gone down another level—if Bitcoin had gone all the way down to $60,000—we would have had ADL too. It’s not like we don’t have it; it was just less aggressive.

Got it. Going back really quickly, because I’m interested in the founder’s wartime experience: when you thought LLP was under attack during that first hour, what did you do? How did you identify that? Was it an all-hands situation where the whole team got on? Did you give a wartime speech? What happened?

We had gone through it before, on March 16. We still remember it well. At that point, we had only been live for less than 2 months.

There was an actual attack at that point. We were very enthusiastic about listing new coins and new markets. We listed a bunch of new markets, and LLP provided a lot of liquidity. It was actually providing more liquidity than the spot markets on Binance. Someone was manipulating the spot markets and profiting from the futures on Lighter.

Within 20 minutes or so, they made $500,000, which at the time was a lot. The entire LLP TVL was around $5 million, so it was a 10% hit. We made everyone whole because we should have known better. We shouldn’t have provided that much liquidity to those new markets.

We were looking at where we were losing. There was some weird activity on Solana and some weird activity on BTC. We asked whether there was one participant profiting in an irregular way, or whether there were any patterns around that. We had done some of that analysis back in March, so we were ready to do it again.

It was all hands on deck. Once we concluded that there was no attack, we switched to making sure the system stayed up. As I said, 5 hours later it didn’t stay up, but for a while it did.

A lot of customers were still dealing with the fog of war. We had far fewer liquidations than other platforms, but out of 200,000 traders, that’s still a lot of liquidations. Even if it’s a smaller number, people are still thinking, “I got liquidated. Can you check that this happened correctly and that the liquidation was fair?”

We were digging into those cases. Maybe we had 1,000 liquidations instead of 5,000, but that’s still a lot of liquidations to track. We were focusing on that, and then 5 hours later—our team is distributed, so by then it was around 10 p.m. for the U.S. folks and 3:00 a.m. for the European folks—we had this outage, and everyone was working for four hours straight to bring the system back up.

Sounds like a movie in the background.

So it was kind of three phases, right? First, we thought, “Okay, we’re good. There’s no attack. We’re good.” Then we finished all these customer tickets, and we took another hit through the outage.

We thought we were golden, then we weren’t. Now, okay, we fixed it. We’re good.

Yeah. Thank you for that. We have about 10 minutes before I unleash Robert Chang on you, so I want to get into a couple more things.

I think over the last couple of months, prior to what happened on Friday, the narrative around perps has gotten really hot for the on-chain crowd. A lot of the memecoin guys have started trading perps. All anyone is talking about is Hyperliquid, Lighter, points, and things like this.

Hyperliquid is obviously a dominant force on the DEX side right now. Why would somebody moving forward trade on Lighter versus other options such as Hyperliquid?

I think our value proposition has been the same since we started building three years ago. I don’t think people necessarily know that. We started working on Lighter in late 2022. In fact, when we first started working on it, FTX was still around. It was right around this time of year, in October 2022.

The thesis was always: let’s build an exchange that has all the security and verifiability you can get from blockchain, but is also as user-friendly, fast, and cheap as, if not better than, centralized exchanges. That was always the thesis, and that’s the value proposition today.

It’s cheaper. If you’re a retail trader, it’s zero fees. If you’re not a retail trader—if you’re an algorithmic or API trader—it’s still lower fees than competitors.

It’s faster. A lot of people have done measurements showing that it’s a few hundred milliseconds faster, which really makes a difference, whether you’re using the front end or the API.

It’s backed by Ethereum security. With exchanges that are running their own L1, if that L1 goes down, that’s it. You can’t get your assets out. On the other hand, if Lighter’s outage had lasted longer and we couldn’t fix it, everyone could use the escape hatch and get all their assets out from Ethereum.

Last but not least is the verifiability piece. Everything happens with a cryptographic proof: all the matches and all the liquidations. We stand by every single matched order and every single liquidation that happens. That’s what we were doing with the customer tickets asking, “Why was I liquidated?” We could explain exactly why and show the proof.

Those are the value propositions. If you look at our growth, I think they have been resonating with at least some customers. We’re always open to feedback, new feature requests, and all of that. The team is ready to build more.

To follow up there, I have a question from Gwart. I’ll read his exact question:

“Could you distinguish between payment for order flow and what Lighter is doing by charging for API access? And, to follow up, what do you think Lighter could charge in the long run versus what HFTs will charge in TradFi for payment for order flow?”

I think there are a lot of low-information actors on X. When we talk about the Robinhood model, the point isn’t literally Robinhood or payment for order flow. It just means the bigger-picture model of being free for retail and charging professional traders or institutions. That’s where the analogy ends.

With Lighter, it’s on-chain, so there are no side deals with anybody. It doesn’t matter if you’re the biggest market maker in the world or two guys in a basement in France running HFT strategies. It’s the same fee structure, and it’s transparent and on-chain.

It’s not like that. We’re big fans of Robinhood, and I think a lot of the criticism they’ve gotten has been unfounded. But even if you believe that criticism, none of it applies here because everything is on-chain and transparent, with the same fee structure for everybody.

That was sick. Shout-out to Gwart, low-information actor.

Thank you for that, Mr. Vlad. If you’re down, let’s get Chang in here. I think that’s most of what I had. We could go for an hour, but I said, “Fuck it. Let’s get Chang in here.” What do you think?

Sure, let’s do it.

Okay. Chang is going to join. I don’t really know what he has for you, but I’m going to moderate. We’ll keep it PG. Then I think Marty is going to join at some point. Let me get the overlay right.

Mr. Chang, what’s up?

Yo, what’s up?

What’s up, man?

How are you?

Morning. Good morning.

Welcome, Mr. Robert Chang, friend of the stream. I don’t think you really need to give an intro, but you can if you want.

Yeah, there’s no intro. I guess we just start.

Yeah. Sick. I don’t really have much prepared for you, so I’ll just play the middle here if it gets hot. Chang, you can start, and then we can go back and forth between you and Vlad.

I appreciate you stepping up to the plate and suggesting that we do this. I’m excited.

Yeah. I think one of my largest questions for Vlad isn’t really something against Lighter, but why do you think the Hyperliquid team allowing that short from one wallet to become 25% of the entire open interest on BTC was the main problem that caused everything?

If so, what are your thoughts on that, and how can something like this be prevented in the future? A really large wallet could always repeat the same strategy, right?

As far as their rules, that’s more of a question for Jeff. But as I’ve said before, people talk about how too high a ratio of volume to open interest is unhealthy. A too-high ratio of open interest to volume is also unhealthy.

If you have a system with tens of billions of dollars in open interest and not much liquidity to absorb that open interest when people want to get out, that’s not healthy either.

At the end of the day, you need to grow both sides of the marketplace. It’s a flywheel: you get the whales, you get the small fish, then you get the smaller sharks, then you get bigger fish, then you get bigger sharks, then you get the whales, and you get even bigger.

You need to grow the ecosystem like that. You can’t just say, “We have a couple of small market makers and a liquidity provider of last resort, so let’s get the biggest whales in the world in there.” That can be unhealthy for the system. That’s what I have to say about that.

I see. Do you think this whole decentralized-perps thing—people pretend Hyperliquid is decentralized—but in your opinion, on the spectrum, how decentralized is Hyperliquid compared to Lighter?

Do you think this is a problem in the future? Obviously, there are trade-offs, but it kind of feels like Jeff is the reason why Hyperliquid is Hyperliquid. Everyone is fighting Lighter because it didn’t work for a while.

It feels like this is a problem where an exchange pretends to be decentralized, but it isn’t very decentralized. What are your thoughts there?

I think, again, you’re talking about the outage, and that was a big problem. But people should be aware that it happened five hours after the flash crash. It was technical maintenance that should have been done earlier but got done later.

To your broader point about decentralization and how it applies to other platforms, we actually started building before Hyperliquid. We started building around this time of year in 2022. We were aware of Chameleon Trading and the backstory there.

I think they’ve done a great job overall. Given the architecture choices they made, it’s worked very well. Choosing an L1 instead of an Ethereum L2 wasn’t the choice we made. We thought about it too.

I remember when we were first debating what we should build and how. We had this long meeting in July 2022, and one of our engineering leads said, “Hey guys, why don’t we just use Cosmos? Forget all this Ethereum stuff.”

The others on the team had a pretty strong view that we really liked building on Ethereum and inheriting Ethereum security and composability. That’s going to be important.

So anyway, had we gone the other way, we probably would have had a similar stack to dYdX v4 or Hyperliquid, which both started out with a Cosmos chain. There are trade-offs. I think that, from a—like we mentioned—we have Ethereum security, verifiability, low latency, and low cost. The verifiability part is something that the L1 architecture has as well, so that’s not necessarily unique.

I think the other guys who tried to build on Ethereum don’t have that part, but what’s unique about us is that we have verifiability on top of Ethereum. The security part is very important, but we also gain composability. There are something like $150 billion, or maybe less after the crash—perhaps $120 billion—of assets on top of Ethereum, and all of those could be used as collateral on Lighter. Whereas if you build your own L1, you have to rebuild that ecosystem on your L1.

I think they’ve done a great job at that. They have $8 billion in TVL or something like that, which is great, but $8 billion is still a lot less than $150 billion or $120 billion. There are trade-offs here. I think Vitalik made the point that the more secure the L1, the more performant the L2 on top of it can be.

For us, the latency is faster. I think most people who have tried Lighter and other products can feel that. But there are downsides to the architecture. For somebody who cares about sovereignty, there is a centralized sequencer. The sequencer has to play by the rules; that’s what the whole ZK architecture ensures. If the sequencer messes up, gets corrupted, or gets hacked, it can’t post its proof to Ethereum, and no money can change hands between the L1 and the L2.

You could still argue that you’re relying on a centralized sequencer. Philosophically, some people might not like that. There are downsides to our architecture too, but we think the priorities we care about—low cost, low latency, verifiability, security, and composability—are areas where we do pretty well across the board.

Why do you think composability is important? Are there any examples of when composability actually makes sense?

For example, if you have spot and perpetuals that share a sequencer, you can do things like basis trading atomically. That’s composability. Here’s another example: if you have pools like LLP or other pools, you can tokenize them, and they can live as ERC-20 tokens on Ethereum. Someone could borrow against an LLP position without even going into Lighter.

That’s pretty cool. You can essentially run on-chain finance in much more interesting ways when you have composability.

I agree. But don’t you think the downside is that Ethereum didn’t really perform well under stress on that day? What gives users assurance that Ethereum isn’t going to perform terribly again? It seems like, by being an L2 on Ethereum, there are those kinds of downsides.

Are you talking about high gas fees?

Gas fees mainly. Gas fees are the issue.

I think gas fees are supply and demand. That’s how it’s supposed to work. That’s why you have things like insurance funds. If you have an on-chain strategy, you would factor that in. Ethereum didn’t have an outage; it actually—

Compared to a lot of other chains, Ethereum has been very robust. I think the last time it had an outage was in 2015, so that’s 10 years of stable performance. The gas fees can be handled through financial engineering, with insurance pools that pay for when that happens. I actually think that’s a strength, not a weakness, because it’s all composable.

With gas fees on a different L1 that has less history, you can’t really underwrite that risk. I think it’s better to do that on Ethereum, which has many years of history, because you can underwrite those risks.

I see. I guess moving forward, you guys will continue to take the stance of being a layer 2 on Ethereum, and that’s not something you would diverge from.

The main thing I’m wondering about is that people want to trade on Lighter, but they want to own the Hyperliquid token, right? If you own the Hyperliquid token, you make money when people get destroyed and HLP prints. But if you touch LLP, which some people think of as an insurance product, once you launch a token—or, not saying there will be one, but if you do launch a token—what’s the growth story for users to hold it if there’s no—

To hold it, right?

I’m not saying Hyperliquid is better.

I think that’s just the whole point. What really matters is creating value for customers and creating value for token holders as well. This whole idea of an L1 premium is pretty dumb. It was probably something a VC said at a cocktail party 3 years ago, and everyone picked it up. The concept has no merit.

If you’re creating customer value, you have a good token. If you’re not, you don’t. To me, it’s that simple.

Okay. But VCs also created the L2 premium narrative, right?

Whoever said that was wrong too. Those aren’t games that are interesting to us. We don’t care about the meta; we care about creating value.

You build based on first principles: what’s the best architecture for the customer? Then you build that and create value for the different parts of the marketplace, whether it’s traders or LPs. LLP has a stronger Sharpe ratio and better performance than anything else out there, so I think that’s good.

Value will accrue to token holders as well. Every constituency will do very well as long as the system is creating value across the board and growing. The best way to grow is to create value for customers, have efficient technology, build a good community, and keep iterating quickly.

The choice of architecture should be based on what creates value for customers, not on what someone pontificating about the meta says.

Right. The thing with LLP is that the description on the website says it will run market-making strategies and handle liquidations. Are you going to explain that further so LLP depositors understand that it’s an insurance layer?

It’s doing both. That description isn’t wrong. What we need to explain further is exactly what that means. What does it mean to handle liquidations? There are partial liquidations, full liquidations, and ADL. What are those risks, and what happens if the market moves dramatically?

Let’s say Bitcoin went all the way from $120,000 to $60,000. What would happen? I think we should break it all down. Theoretically, there is a risk of total loss. I’m talking about moves that have never happened in history, but theoretically, that can happen, and we should explain those risks more.

That’s true for anybody. If you buy BTC spot, there’s also a risk of total loss. What’s in the documents now isn’t wrong; it just isn’t thorough enough in explaining all the different scenarios.

Right. Based on the way it was handled, you could say it was somewhat deceptive to the first set of participants who experienced that event.

I don’t know if deceptive is the word I would choose, but there wasn’t enough clarity about how it worked. That’s why we compensated LLP holders somewhat. The system did what it was supposed to do, but there were enough people who bought into it beforehand—especially during the frenzy in the markets—who probably didn’t know much about what they were getting into.

We feel bad for those people, and we did compensate them somewhat, even though the system did what it was supposed to do.

Yeah.

And again, if they keep holding, our expectation is that it’ll continue to be a Sharpe ratio of 5-plus with a good APR. For what it’s worth, the last 3 days were up again, right? It was up 15 days in a row before that and up the next 3. If this event is a once-in-10-year thing, then you can do the math on that.

Yeah. Another question, maybe just on the liquidation engine side: in the future, how would the liquidation engine become more performant, or why did it seem like it was really slow on that day?

Again, the slowness had nothing to do with the liquidation logic. Like I mentioned, there was degraded performance for a couple of hours, and then later on there was a database issue. That was more of an infrastructure scaling issue. We grew too quickly and didn’t upgrade the infrastructure soon enough. Again, not a good thing. We take responsibility, but it has nothing to do with the liquidation logic.

As a point of fact, we’ve had very volatile markets before. We had them in February, April, and July, and the liquidation engine worked very well through all of that. We’ve done millions of liquidations, and there’s a cryptographic proof that every single one is correct.

I see. Yeah. Thanks for answering all the questions. I don’t really have anything else.

Good questions. Thanks for coming up.

I do want to ask: if Hyperliquid is doing Trade XYZ and giving users this other growth story, where people can trade tokenized equities—or not tokenized, but trade equities and commodities—is this something that’s going to be possible in Lighter as well?

Yeah. We’re looking at real-world assets very closely as well. Stay tuned for that. We’ll have something on that in the coming weeks.

Okay. Do you think there are legal or regulatory risks to letting users get exposure to these assets without KYC? What are your thoughts?

These assets meaning the digital assets or stocks?

Stocks.

That’s a good question. I think it depends on whether it’s spot or perps. We have a couple of advisors, and this is where I think, again, back to the narrative thing, people ask, “Why do you guys have VCs?” These are the kinds of questions the VCs we have are really helpful with.

They get out of our way when we’re just building product, but for these kinds of questions, this is the kind of stuff where understanding that the SEC looks at stocks, while stock futures are under the CFTC, matters. How are they thinking about DeFi? If it’s a stock future, that’s actually regulated by both the SEC and the CFTC. There are a lot of nuances there, and we’re working through those before listing those assets.

I see. I’m really excited to hear about what’s coming next. I think you handled everything really well. Generally, everyone on Crypto Twitter right now is a huge fan of yours. Thank you.

I appreciate that.

Yeah, I don’t know. We’re bringing Marty up, Thread Guy, right?

Yeah, we’re bringing Marty up. Chang, this is lit, bro. I’m excited for your tweet rant to follow this, but thanks for coming on, dude.

By the way, I do have some stuff to get to in about 10 or 15 minutes.

All right, we’ll bring him out right now. We’ll rip it right now. Chang, you’re the truth, bro. Thanks for coming on, dude. We’ll get him right now, Vlad.

We’ll cap him at 10 or 15 minutes.

Okay, cool.

How do you think that was? That was fun.

Well, I was surprised. Some of the tough questions were not so much about us, but about our competitors. That was interesting.

What’s up, Marty? Let me just save his image real quick because of your PFPs.

I got plus one Vlad there. Vlad.

Plus one Vlad.

Good.

All right, cool. Let me get back in here. Marty, welcome. It’s good to have you on here. I think you just listened to the Chang one, so you can kind of rip if you want to get into some stuff, and we can go from there. Thanks for coming on.

Yeah. Thanks for having me. I appreciate everybody making the time. I just want to make it clear: I’m not a hater. I’m not Lighter to zero, but as a user, as somebody who’s gone through the API docs, I just want to share our initial experience.

Friday was essentially bad for everybody. Some exchanges were worse than others. A little bit of backstory: I think, Vlad, we’ve been in some chats before. We’ve been talking with the Lighter team since January. We were looking to start doing API in April or May, when it had come out in January. It wasn’t available, and then as soon as we launched API, we launched the pool.

We got some pushback initially from the team, which I think is the main problem I have with the Lighter team: their communication. It’s not necessarily about getting pushback, but things that seem fairly simple. We were initially asking if we could do an API on a pool, and somebody said yes, somebody said no, and people in the Gen 1 chat said they were already doing it.

My main issue is that there are no API docs and we’re getting mixed answers from the team.

Sorry, were you talking about a few months ago or now?

Last week, like 2 weeks ago. When we did the public mainnet launch—exactly 2 weeks ago, October 1—the API docs were updated and published on the main website.

There might still be gaps there, but they were definitely revamped quite a bit. Did you see those revamped API docs, or are you talking about the ones from months ago?

Yeah, we were able to go through them, and I think some users even today have issues. If you go in that Gen 1 chat, the pools chat, there were some other issues trying to simply get an API up on the pools.

What do you mean by an API on the pools? Do you mean for running a custom pool?

Yeah.

With an API?

Yep.

Okay. To be fair, that’s a pretty new concept for us. I think the idea initially was that we have LLP, and then it was, “Let’s do some copy-trading-type stuff for some of our star traders,” who have had very nice track records. Those are discretionary, retail-type traders.

The idea of doing pools for programmatic traders is almost like running an on-chain quant hedge fund. I’m excited about that idea. Having done TradFi before, I think that’s really cool—democratizing finance. Philosophically, we’re on the same page that something like that needs to exist, and it would be really cool if it were prevalent on Lighter.

Practically speaking, we’ve just gotten into that. We finished onboarding about 600 API traders in the months starting from, I guess, April through September. You were one of the people who looked into the API. Between April and September, we onboarded about 600 API traders, or small HFT shops, some not so small.

We’re only now starting to think about how to do that in the context of the pool. Maybe just bear with us a little bit. The intersection of custom pools and API trading has only become a priority for the team over the past couple of weeks.

No, I’m all good. Everybody’s in the testing phase. I think the main issue is that nobody is on the same page. We have the public Gen 1 chat, an API chat, and another general chat with you guys. Somebody says you can, somebody says you can’t, and then 2 minutes later somebody says, “Yeah, you can do it, but it’s not advised.”

Come on, let’s simply all get on the same page. Can we do it or can we not? We ended up fixing it ourselves and making an API key on the pool, and lo and behold, it worked, right?

Right. I mean, I think—

The ambiguity wasn’t that—

I’ll dig into the details. I think you’re probably talking about how there’s some communication on Discord and someone on Telegram.

I’m only on Telegram and Discord. You’re lucky, Thread Guy. I had to create a Discord.

Yeah, that was the other thing. The team is like, “Oh, you have a bug reported on Discord.” I’m like, “Well, I don’t have admin access. I have to wait 2 hours to go through the bot.” There should be an easier way for users to communicate a bug to the team.

I put it in a few Telegram chats, and somebody was like, “Put it in Discord.” Like, do you want to hear it or not? I’m not going to Discord. I think that’s my main overall issue with team communication and everybody being on the same page.

If they’re like, “Hey, no, you can’t do API trading on a pool. You can’t even create an API key. It’s not available on the front end. It’s not available in the docs,” sure. But we could create an API key. Somebody said we could. It’s a little flip-floppy and ping-pongy.

Yeah. I think, zooming out, there’s definitely a lot of work that the team has gone through onboarding lots of different market makers and API traders. I think there are some groups where the communication slipped through the cracks, so we apologize for that.

On the substance of it, I think the reason there was a lack of clarity about that specific point is not so much that it’s technically impossible. It’s more that, from a user-risk-management point of view, it’s one thing to show public pools managed by a discretionary trader who’s trading a couple of times a day. There’s not that much risk if that person goes rogue because they’re only allowed to trade on Lighter, and they can’t take the assets out.

If they’re trading with low frequency, there’s not that much damage they can really do. Let’s say that trader is not a good actor. If they’re just pushing buttons to buy or sell a few times a day, there’s not that much they can do. But with API traders, you could write a system that’s harmful to participants in the pool.

I think it’s not about whether you can create an API key. It’s more about making sure that whoever is participating has guardrails in place. We don’t want to show pools at the top that might be seen as risky. Anyway, that’s just a little bit of context.

Oh, that’s a better point about the communication.

Yes, maybe. If you would have just said that—not you specifically, but the team—“Hey, we’re trying to fish out bad actors. We don’t want to push API to retail people who are depositing, for safeguarding,” then, yeah, sure, no problem.

But I think it was just the lack of clarity. That’s my overall issue with the Lighter team: the lack of transparency and communication.

Yeah. The customer-support team is really focused on the retail trader. A lot of the API traders actually onboarded without any help from us. But, again, for context, we have 2 people working part-time who are spending part of their time onboarding API traders. We’ve onboarded around 600 groups in 6 months, so some fell through the cracks. We do apologize for that, but hopefully this clarifies things a bit.

So, the best customer-support call of all time.

Yeah. That’s why I called in today. It was like a 1-800 number. It wasn’t to say we’re going to zero, right?

And the other thing with the pool is, if it is a bad actor, they do have to hold X amount of the supply in the pool. So they’re also losing their own money, right? They could make money somewhere else if they were able to be a negative actor, but they’re also losing their own money, too.

I think another thing I could bring up is that we froze our pool due to a loss on PAXG when it whipped to $3,500. We were able to withdraw all of the creator funds, but we’re still able to trade on the pool with 0% ownership. If you look on mobile, the pool doesn’t show, but if you look on the UI, the pool still shows.

Just odd, right? If the pool’s frozen, it shouldn’t be able to trade. And you’re able to trade with zero ownership?

Yeah. Which one is yours?

I think it’s called Up Down.

Up Down?

It’s down like 85% on a liquidation on PAXG. Fun times.

Yeah. To be fair, we actually haven’t officially released them. If you look at our announcements, the user public pools have been kind of an Easter egg. We actually haven’t announced them yet. This stuff is very much experimental.

But, again, if there was a lack of communication, we’ll do better on that.

Yeah, if a pool’s frozen, it should just be withdraw-only. That seems simple.

Yeah. Makes sense. Cool. Well, I do have to jump.

If I can’t turn the pool back on, then what’s the point, right? I can still trade.

Yeah, I guess we’re getting a bit into the weeds, but we’ll follow up with you on the details of that. I’m sure we’ll be able to work out that specific issue here. But, yeah, thanks so much for having me. I do have to jump.

Vlad, it was a pleasure. Marty, thank you as well. I think 1,800 Lighter is a decent number, but, no, in all seriousness, thanks for coming on. Vlad, it was a pleasure. Thank you for ripping 2 debates, and have a great rest of your day, both of you guys.

All right. Thanks so much, fellas.

All right.