# Trial Testimony of Nishad Singh, October 16 and 17, 2023: Transcript, part 10 of 13 (pages 1532 to 1555) [Index](index.html) · [PDF, October 16](pdf/2023-10-16-trial-transcript.pdf) · [PDF, October 17](pdf/2023-10-17-trial-transcript.pdf) · [Whole document](trial-testimony-2023-10-16-17.html) · [Web page](trial-testimony-2023-10-16-17-part-10.html) · Part 10 of 13. Parts: [1](trial-testimony-2023-10-16-17-part-1.html) [2](trial-testimony-2023-10-16-17-part-2.html) [3](trial-testimony-2023-10-16-17-part-3.html) [4](trial-testimony-2023-10-16-17-part-4.html) [5](trial-testimony-2023-10-16-17-part-5.html) [6](trial-testimony-2023-10-16-17-part-6.html) [7](trial-testimony-2023-10-16-17-part-7.html) [8](trial-testimony-2023-10-16-17-part-8.html) [9](trial-testimony-2023-10-16-17-part-9.html) 10 [11](trial-testimony-2023-10-16-17-part-11.html) [12](trial-testimony-2023-10-16-17-part-12.html) [13](trial-testimony-2023-10-16-17-part-13.html) · About this text version: at the end of this page --- [1532:5] **Q.** Mr. Singh, take a moment and read this to yourself. The question is whether it refreshes your recollection as to what you told the prosecutors about your haziness in the meeting with them on January 4, 2023. [1532:9] **A.** Sorry. Could you repeat the ask again? [1532:10] **Q.** Sure. This is whether or not it refreshes your recollection as to what you told prosecutors about your surprising amount of haziness in 2020 in June and July of 2022. Yes or no, does it refresh your recollection? [1532:14] **A.** I still don't remember having said it specifically about June or July of 2022, but I could believe that I did. [1532:16] **MR. COHEN:** We can take that down. [1532:17] **THE COURT:** The question was, does it refresh your recollection, not whether you could believe that you did. [1532:19] **THE WITNESS:** Fair point. The transcript or, I guess, the notes themselves don't actually refresh my recollection. I do separately remember -- [1532:22] **THE COURT:** You've answered the question. [1532:23] Could we move on. [1532:24] **MR. COHEN:** Yes, your Honor. [1532:25] **Q.** Let's continue with what happened in June. [1533:1] Do you recall telling us, Mr. Singh, that after the discussions we have already covered there was a discussion about fixing the bug? [1533:4] **A.** Forgive me. I missed I think part of your sentence, but I do remember there being discussions about fixing the bug. [1533:6] **Q.** Just to remind us, what were those conversations about? [1533:7] **MR. ROOS:** Objection. [1533:8] **THE COURT:** I imagine they were about fixing the bug. Isn't that what the point was? [1533:10] **MR. COHEN:** It did occur to me, your Honor. [1533:11] **THE COURT:** Good. [1533:12] **Q.** Who was tasked with fixing the bug? [1533:13] **A.** I wasn't involved in the project, so I don't know who was tasked with it. I do know that Gary and Adam worked on it. [1533:15] **Q.** Was a document created to document what had happened with the bug fix? [1533:17] **A.** Adam wrote one. [1533:18] **Q.** Did you ever hear the term a postmortem document? [1533:19] **A.** Yes. [1533:20] **Q.** What did that refer to in this context? [1533:21] **A.** A document describing an error or like a bug and the sequence of steps taken to address it and lessons to take. [1533:23] **Q.** And did you have occasion to review the postmortem relating to the bug that Adam prepared? [1533:25] **A.** I don't think that I read it in its entirety. I did talk [1534:1] with Adam about it. [1534:2] **Q.** Did you think that the fix had been done properly? [1534:3] **A.** Yes. [1534:4] **Q.** And the postmortem referred to the fix of the $8 billion bug, correct? [1534:6] **A.** And other considerations around it. [1534:7] **Q.** Such as? [1534:8] **A.** There were identifying the exact size of the bug, had some complicated elements. Adam walked through a lot of the considerations that he and Gary had sort of thought through. Beyond talking about the fix for the bug, it talked about the mechanical process that they took to ensure it wouldn't happen again and ensure a safe migration. [1534:14] **Q.** So the technical issues that had been involved in fixing the bug and making sure it wouldn't happen again? [1534:16] **A.** Yes. [1534:17] **Q.** Did you think this was an appropriate thing for Gary and Adam to be doing? [1534:19] **MR. ROOS:** Objection. [1534:20] **THE COURT:** Sustained. [1534:21] **Q.** Did you have a view about whether or not this memorandum should have been prepared? [1534:23] **MR. ROOS:** Objection. [1534:24] **THE COURT:** Sustained. [1534:25] **Q.** Let's move on, Mr. Singh. [1535:1] You mentioned yesterday that there came a time when there was an attempt to split out the balance of the fiat@ account from Alameda and FTX. [1535:4] Do you recall that, sir? [1535:5] **A.** I do. [1535:6] **Q.** And when in relation to the bug fix was this? [1535:7] **A.** I believe it was after. [1535:8] **Q.** Was it after the memo we just discussed? [1535:9] **A.** I believe so, but I am not 100 percent certain. [1535:10] **Q.** Can you describe for us what steps were taken to split out the balances. [1535:12] **MR. ROOS:** Foundation. [1535:13] **THE COURT:** Sustained. [1535:14] **Q.** Were you involved with this at all, Mr. Singh? [1535:15] **A.** No. [1535:16] **Q.** Did you come to learn of it? [1535:17] **A.** I came to learn that the exercise had been done. [1535:18] **Q.** Did you learn the outcome of the exercise? [1535:19] **A.** Some parts of it, yes. [1535:20] **Q.** Tell us what you learned. [1535:21] **A.** I learned that there had been an account, a new subaccount called FTX fiat old made under the info@AlamedaResearch.com main account, and that it now reflected the amounts owed by Alameda to FTX and that fiat@FTX.com now reflected the amounts that FTX owed to FTX. [1536:1] **Q.** Are you done with your answer? [1536:2] **A.** I am. [1536:3] **Q.** Did you have an understanding of the purpose of the FTX underscore fiat underscore old account? [1536:5] **A.** I inferred it. [1536:6] **Q.** What did you infer? [1536:7] **MR. ROOS:** Objection. [1536:8] **THE COURT:** Ground. [1536:9] **MR. ROOS:** Foundation, relevance. [1536:10] **THE COURT:** What's the relevance? [1536:11] **MR. COHEN:** The relevance is we heard about this sequence yesterday in a way designed to suggest it was done in secret and designed to keep this account from everyone. [1536:14] **MR. ROOS:** Withdrawn. [1536:15] **MR. COHEN:** I'm entitled to explore it. [1536:16] **THE COURT:** The objection is withdrawn. [1536:17] **MR. COHEN:** Can we have the question back, please. [1536:18] **THE COURT:** Yes, of course. [1536:19] (Record read) [1536:20] **A.** I inferred that it was made and separated from fiat@FTX.com to track the correct amount that Alameda owed FTX in customer fiat deposits. [1536:23] **Q.** To track it, correct? [1536:24] **A.** Correct. [1536:25] **Q.** Then there came a time when that account was moved again, [1537:1] correct? [1537:2] **A.** Yes. [1537:3] **Q.** And I believe you told us about -- I am going to pronounce this incorrectly I'm sure -- the Seoyun account. [1537:5] Do you recall that, sir? [1537:6] **A.** I do. [1537:7] **Q.** Do you have any involvement with that? [1537:8] **A.** I moved the subaccount to this Seoyun account. [1537:9] **Q.** You were the person who moved the account from FTX old to Seoyun? [1537:11] **A.** I either moved it or I asked Andrea to move it. Either way, I directed it. [1537:13] **Q.** For record purposes Andrea is Andrea Lincoln? [1537:14] **A.** Correct. [1537:15] **Q.** Who worked with Adam and Gary, correct? [1537:16] **A.** Right. [1537:17] **Q.** What was your purpose for moving it? [1537:18] **A.** I was told to. [1537:19] **Q.** Who told you? [1537:20] **A.** Sam told me to move it and a trader told me where I should move it to, such that it was still tracked. [1537:22] **Q.** I believe you referred yesterday to -- the trader was Terence Choo, is that correct? [1537:24] **A.** I think it was Terence Choo, but there is some chance it was Caroline Ellison. [1538:1] **Q.** Both Caroline and Terence worked for Alameda, correct? [1538:2] **A.** Correct. [1538:3] **Q.** In the trading group at Alameda. [1538:4] **A.** That's my understanding. [1538:5] **Q.** They advised you where to move the account, correct? [1538:6] **A.** They did. [1538:7] **Q.** Did you know why -- withdrawn. [1538:8] Did the topic of why this account, as opposed to another one, came up? [1538:10] **A.** In that I specified the parameters that I viewed as important. I specified that it shouldn't be under one of the accounts that was in consideration for how much Alameda was charged line-of-interest fees on. And I asked what is another account that you guys are tracking that isn't in that set. [1538:15] **Q.** And they provided you with the account? [1538:16] **A.** Yes. [1538:17] **MR. COHEN:** Your Honor, this might be a good time for our morning break. [1538:19] **THE COURT:** OK. 11:15, folks. [1538:20] (Recess) [1538:21] **THE COURT:** The record will reflect that the defendant and the jurors all are present, as they have been throughout. [1538:23] The witness is reminded he is still under oath. [1538:24] Mr. Cohen, you may proceed when you are ready. [1538:25] **MR. COHEN:** Thank you, your Honor. [1539:1] **Q.** New topic, Mr. Singh. [1539:2] I want to go back. Counsel covered with you some questions about certain of the code base features. [1539:4] Do you recall your testimony about that yesterday? [1539:5] **A.** Forgive me. I don't know what you mean by counsel. [1539:6] **Q.** I mean Mr. Roos. I'm sorry. [1539:7] **A.** I recall talking about code changes. [1539:8] **Q.** Let me start with allow negative. My first question is a when question. When did you first learn about the allow-negative feature? [1539:11] **A.** July 2019. [1539:12] **Q.** How did you learn about that? [1539:13] **A.** I was told to write it. I did. [1539:14] **Q.** So you were the person who put it together? [1539:15] **A.** In some sense, I wrote the code, but I was sort of given pretty clear and explicit instructions from others. [1539:17] **Q.** When you wrote it in 2019, what was the purpose of the code? [1539:19] **A.** It was to facilitate FTX admins moving FTT from designated accounts or making trades in FTT from designated accounts and to modernize an existing set of features that would allow accounting-oriented accounts to go negative. [1539:23] **Q.** Was this at all in connection with market-making functions? [1539:24] **A.** I am not sure. [1539:25] **Q.** Did there come a time that you came to believe that the [1540:1] allow-negative feature was used in connection with market-making functions? [1540:3] **A.** Yes. [1540:4] **Q.** When was that, sir? [1540:5] **A.** I don't know precisely. [1540:6] **Q.** Approximately is fine. [1540:7] **A.** 2020. [1540:8] **Q.** How did you learn of that? [1540:9] **A.** I remember a conversation I had with Gary Wang in which I was thinking about making some changes to some code related to OTC trades, and I noticed that Alameda was the only provider -- [1540:12] **Q.** Could I interrupt you for a moment, sir. Can you tell the jury what OTC trades are. [1540:14] **A.** Over the counter. I don't think that does a great job describing it. If I may. [1540:16] **Q.** Does it mean not on the exchange? [1540:17] **A.** No. [1540:18] **Q.** Tell us what you are thinking. [1540:19] **A.** It means not on an order book. So a customer could basically request -- just say I want to buy one Bitcoin. Tell me how many dollars that takes. They get back an answer. They have some time before they have to accept -- before they can -- before that sort of quote expires. They can say I accept and it happens. This is different from submitting an order on an order book. [1541:1] **Q.** Please continue. [1541:2] **MR. ROOS:** Objection. Calls for a narrative. [1541:3] **THE COURT:** Sustained. Ask a more specific question, please. [1541:5] **MR. COHEN:** Sure. [1541:6] **Q.** You were describing for us how you came to the understanding of a connection, if any, between allow negative and market making. [1541:9] **A.** As a part. [1541:10] **MR. ROOS:** Objection. I don't think there was a question. He was just describing what the witness was testifying to. [1541:13] **THE COURT:** Yes. I agree. [1541:14] **Q.** Could you tell us the basis for coming to the view, if you did, that there was a connection between the allow-negative code and market making? [1541:17] **A.** I am describing only one instance. There may have been others. Is that OK? [1541:19] **Q.** Yup. [1541:20] **A.** I was thinking about making some change, I can't recall what, to the code related to these OTC trades. I noticed that the balances of the provider, Alameda, weren't checked when accepting an OTC trade, meaning that if Alameda didn't have funds that would otherwise be necessary for making the trade, it could go through anyway. [1542:1] I asked Gary about this. This is a reflection of allow negative. And he described to me the type of situation where this seems like the right thing for the exchange to do. [1542:4] **Q.** Which was what? [1542:5] **A.** I don't know if he gave this precise example, but one comes to mind. [1542:7] **THE COURT:** Could we stick to what actually happened instead of what you think now. [1542:9] **THE WITNESS:** Fair enough. [1542:10] **A.** He gave an example. I can't remember the exact example. It seemed valid to me. [1542:12] **Q.** Let me ask you this, Mr. Singh. As a result of the allow-negative feature, was Alameda allowed to sell tokens it didn't have at the time? [1542:15] **A.** In some ways. But I have since discovered that that was -- allow negative was actually not the section of code proximately responsible for that. [1542:18] **Q.** But at the time, meaning 2019 and 2020, that's what you believed? [1542:20] **THE COURT:** Sorry. What is what he believed? [1542:21] **MR. COHEN:** That the allow negative allowed Alameda to sell coins it did not yet own. [1542:23] **A.** I don't know that I thought about it in 2019 and 2020. I expect that I was told that or thought about it in 2021 and certainly at least in one instance in 2022. [1543:1] **Q.** In 2019, 2020, when you were working on the code, did you have a view about whether it permitted Alameda to purchase newly issued tokens? [1543:4] **MR. ROOS:** Objection. [1543:5] **THE COURT:** Sustained. [1543:6] **Q.** Did you ever have a view about -- let me back up. [1543:7] Did you participate, Mr. Singh -- let me back up. [1543:8] Do you have an understanding of what it is to issue a token? [1543:10] **A.** I have some understanding. [1543:11] **Q.** Why don't you tell us that. [1543:12] **A.** I am not sure I know that I am talking about what you are. Can you specify a little more what you mean? [1543:14] **Q.** Sure. Do exchanges sometimes issue tokens? [1543:15] **A.** Can you specify even further? [1543:16] **Q.** For example, Binance issued BNB. Are you familiar with that? [1543:18] **A.** Yes. [1543:19] **Q.** And if Binance was issuing BNB or some similar token and a customer on Alameda wanted to buy at the issuance, would they be able to? [1543:22] **MR. ROOS:** Objection. Foundation. [1543:23] **THE COURT:** Yes. I think you have to lay a foundation for that. [1543:25] **Q.** Based on your experience as the author of the [1544:1] allow-negative feature in the code, are you familiar with whether those features had anything to do with Alameda's ability to buy newly issued tokens? [1544:4] **A.** At the time I wrote the feature, I don't think I had anything relating to buying tokens. [1544:6] **Q.** I meant selling tokens. [1544:7] **A.** Even that. That was not the purpose that I was told it was made for. [1544:9] **Q.** We talked yesterday about stablecoins. [1544:10] Do you recall that, sir? [1544:11] **A.** Yes. [1544:12] **Q.** In your experience as the author of the allow-negative feature, did it have any impact on Alameda's ability to purchase stablecoins? [1544:15] **A.** No. [1544:16] **Q.** Did it have any impact on stablecoins at all? [1544:17] **MR. ROOS:** Objection. Vague. [1544:18] **THE COURT:** Overruled. [1544:19] **A.** The feature changed nothing about Alameda's interaction with stablecoins. [1544:21] **Q.** Did it have any effect at all? [1544:22] **MR. ROOS:** Objection. [1544:23] **THE COURT:** Asked and answered. [1544:24] **Q.** Let me ask you this, Mr. Singh. It is going to take me a moment to pronounce this. Have you ever heard of a term called [1545:1] auto de-leveraging event? [1545:2] **A.** Yes. [1545:3] **Q.** What's that? [1545:4] **A.** It may take me a moment to explain. Is that OK? [1545:5] **Q.** That's fine. [1545:6] **A.** In the context of FTX, some liquidations were performed off market, which is to say an account that didn't have sufficient collateral to support its positions would enter liquidation. And, if it was getting close enough to being under water, those positions would be closed out or traded against designated backstop liquidity providers, as opposed to selling those positions in the market so it can be done more quickly. [1545:13] Those backstop liquidity providers could continue to absorb these trades, as in act as the counterparty to these liquidations, so long as they met a few conditions. [1545:16] If those conditions weren't met for any backstop liquidity providers for a given liquidation instance, then FTX enacted auto de-leveraging. This is not code I wrote, so I'm not super familiar, but my understanding of what this is is that it's essentially like picking other users that have not explicitly opted in being backstop liquidity providers to perform the same role for the sake of that liquidation. [1546:1] BY MR. COHEN: [1546:2] **Q.** Let me see if I can just break this down. [1546:3] You said that this situation arose when an account was in danger or at risk of being closed out. Can you explain what you mean by that. [1546:6] **A.** Accounts were liquidated, under some circumstances; some liquidations could result in ADL events. [1546:8] **Q.** Okay. And by closed out, is this what we were talking about——well, let me rephrase. [1546:10] This means a customer has a position in an account and a certain amount of collateral and the value of the position starts to fall sufficiently that the——that that——actually, I forgot to ask you one more thing. [1546:14] Have you ever heard the term "risk engine"? [1546:15] **A.** Yeah. [1546:16] **Q.** Okay. Did FTX have a risk engine? [1546:17] **A.** Yeah. [1546:18] **Q.** How did that——what was that? [1546:19] **A.** A lot of things that went into it, I suppose. Do you mind clarifying some more. [1546:21] **Q.** Did the risk engine have anything to do with closing out positions? [1546:23] **A.** Right. Liquidations were a part of the risk engine. [1546:24] **Q.** Okay. And so if the risk engine, which was basically computer run, noticed that an account was going below its [1547:1] collateral limit, it would step in and liquidate that account, correct? [1547:3] **A.** That's almost right. If it noticed that it had insufficient collateral to support its positions. There was not a collateral limit at play. [1547:6] **Q.** So it would go into the customer's account and sell the positions to get it back in balance, correct? [1547:8] **A.** That's not exactly how I'd put it. [1547:9] **Q.** How would you put it, sir? [1547:10] **A.** That the liquidation engine would sell, or buy, close out their positions, because the customer could be long or short. [1547:12] **Q.** Okay. And then you referenced something called a backstop liquidity provider. What was that? [1547:14] **A.** Designated accounts were used as the counterparties for some of the trades that were required to close out these positions. [1547:17] **Q.** And was Alameda a backstop liquidity provider? [1547:18] **A.** It was. [1547:19] **Q.** Were other entities backstop liquidity providers? [1547:20] **A.** Over time there were many. [1547:21] **Q.** So if the engine was closing out an individual customer's account and there weren't sufficient assets, the backstop liquidity provider would step in, correct? [1547:24] **A.** It didn't depend on if there were insufficient assets. The customer would be getting liquidated because they had [1548:1] insufficient assets. [1548:2] **Q.** And if that were the case, then the backstop liquidity provider would step in. [1548:4] **A.** Right, under some further conditions. [1548:5] **Q.** Okay. And what would happen if the backstop liquidity providers didn't themselves have sufficient collateral? [1548:7] **A.** If none of them did——this would be one criterion for eligibility of a backstop to match against this liquidation. If there were no such eligible backstop liquidity providers, then the system would perform ADLs, or auto-deleveraging fills. [1548:11] **Q.** What did that mean? [1548:12] **A.** It meant——this area I'm a little uncertain about because I didn't write the code, but I believe it's that it selected other customers to perform the same role that the backstop liquidity providers would have, and serve as counterparties to the liquidation trade. [1548:17] **Q.** Okay. So if customer A had an account that was going into liquidity——liquidation and that account had insufficient assets, the next stage would be the backstop liquidity providers, correct? [1548:21] **A.** Sorry. Could you repeat that. [1548:22] **Q.** Sure. If the——I want to see if we can make this a little bit more concrete for the jury. [1548:24] So if customer A had an account that was being liquidated and had insufficient assets, I believe you told us [1549:1] the next step would be to take it the level of the backstop liquidity providers. [1549:3] **A.** It's getting liquidated because it has insufficient assets. [1549:4] **Q.** Correct. [1549:5] **A.** So in getting liquidated, it may go and get——it may enter a mode in which it will be liquidated against backstop liquidity providers. [1549:8] **Q.** Correct. And then I think you told us that if they, the backstop liquidity providers, had insufficient assets, we'd have an auto-deleveraging event? [1549:11] **A.** There were multiple conditions that could lead to a backstop being ineligible. One of them is that they had insufficient assets. If that——if those conditions are met for all backstop liquidity providers, then the system would enact ADLs. [1549:16] **Q.** And that would mean the in——the account would be covered by the assets of other customers on the exchange. [1549:18] **A.** Not quite. [1549:19] **Q.** Okay. Tell us quite. [1549:20] **A.** There is not an exchange of value. There's a very small exchange of value in a liquidation trade. It's not that——it's not that the liquidating account is underwater and therefore needs to get topped up by other customers. It's that——it's that they have positions on that are risky that need to be handed off to other customers. Those other customers, be them [1550:1] backstop liquidity providers or those selected by ADL, receive the positions at slightly better than market value, being that in that moment they actually make some money. It's not that they're giving up their collateral for it, but they are taking on risk that they did not themselves put on. [1550:6] **Q.** Right. So using our example now, customer B or customer C or so forth, in your words, are having positions handed off to them. [1550:9] **A.** Yes. [1550:10] **Q.** And they're going to get them at a favorable price, but they themselves are now at risk, correct? [1550:12] **A.** Right. But it may turn favorable quickly. [1550:13] **Q.** Let me call your attention to July of 2020. Did FTX experience an auto-deleveraging event? [1550:15] **A.** I recall it——I recall one around early August, so it's possible that this one was in July. [1550:17] **Q.** Okay. Why don't you tell us what happened. [1550:18] **MR. ROOS:** Objection, foundation. [1550:19] **THE COURT:** Sustained. [1550:20] **Q.** Okay. What do you recall about the auto-deleveraging event that took place in July or August of 2020? [1550:22] **A.** I recall that it happened that there were ADL fills for what I think was the first time in FTX's existence, meaning that the——forgive me, I'm using the abbreviation, or the acronym——the ADL system had kicked in, which meant that there [1551:1] were no——there was a point at which there was a liquidation and there were no eligible backstop providers. [1551:3] **Q.** So it went to the next level, to the customer level. [1551:4] **A.** Right. [1551:5] **Q.** Correct. Were you aware if FTX had any response to this? [1551:6] **A.** Yes. [1551:7] **Q.** What was it, sir? [1551:8] **A.** I don't remember everything about it, but I can describe some things that I talked about with Sam and Gary—— [1551:10] **Q.** Just describe what you remember. [1551:11] **A.** It was undesirable for customers to take on these positions without opting in, basically being ADL recipients, all else equal. All else equal, this probably——this sort of universe of FTX customers and FTX itself would have preferred that it go to the designated backstop providers. Gary dug into why it was the case that there were no eligible providers. Gary determined that it was because, among other things, Alameda was not eligible and that Alameda was not eligible because its free collateral was zero and that the reason that its free collateral was zero was not because it had a bunch of positions but because it had tons and tons of open orders out——open orders for the purpose of providing liquidity on the hundreds of markets or so. Those exhaust collateral. [1551:24] **Q.** You said that having this auto-deleveraging event was undesirable. What did you mean by that? [1552:1] **A.** It's my opinion. I don't know if I can sort of claim that there's a higher level truth to it being undesirable or not. [1552:3] **Q.** Well, you just said it. I'm just trying to get at what you meant by it. [1552:5] **A.** Yeah, forgive me. I was clarifying. My opinion, and the opinion espoused by Sam and Gary at the time, was that having random customers take on the——these positions and have them handed off to them is worse than having designated customers take them on. [1552:10] **Q.** Okay. To your knowledge, Mr. Singh, was anything done in response to the auto-deleveraging event? [1552:12] **A.** Yes. [1552:13] **Q.** What was that? [1552:14] **A.** So Gary inferred that it was because——a cause was that Alameda was putting out a ton of orders and therefore had low collateral, had zero free collateral, but free for the——for the purposes of placing more orders or accepting trades, not free in the sense if they were to be liquidated, because there's an assumption that orders can be canceled upon liquidation, which would have just them getting a lot of collateral back. [1552:21] **THE COURT:** You used the term pre collateral or free collateral. Would you explain that. [1552:23] **THE WITNESS:** Free collateral. [1552:24] **THE COURT:** First of all, was it pre collateral or free collateral? [1553:1] **THE WITNESS:** Free collateral. Free meaning available. [1553:3] **THE COURT:** Okay. [1553:4] **A.** Customers——to clarify that, customers in FTX deposited collateral. Collateral would be exhausted or, you know, made not free by putting on positions or by having open orders out on markets that, you know, were liable if traded against to create positions. [1553:9] The latter thing, open orders, were primarily responsible, as Gary told me, for Alameda having zero free collateral at the time of the liquidation that caused the ADLs. [1553:12] I believe you asked me what happened. [1553:13] **Q.** I did. And were any steps taken with respect to the code base upon that determination? [1553:15] **A.** Yes. [1553:16] **Q.** What was that? [1553:17] **A.** I think at first I added some alerts in the section of the code that would basically emit if Alameda had or others had zero free collateral, and if that was actually the cause, such that in the next instance if this happened, we had some more debugging visibility. We did, and it showed zero free collateral. It was talked about——but I don't remember if it actually happened——increasing Alameda's line of credit, because increasing it increases their free collateral. Lines of credit directly increased collateral and therefore free collateral. [1554:1] **Q.** And——I'm sorry. Finish your answer if you're not done. [1554:2] **A.** Ultimately, Sam asked me why would it depend at all on free collateral if Alameda is just going to exhaust it all in open orders, and we know that the open orders aren't themselves risky, but not consuming collateral in the same way that having to position does, then why don't we, for Alameda, just remove the condition that we attend to its free collateral when determining if it's an eligible backstop provider for a liquidation. [1554:10] **Q.** And was that condition removed? [1554:11] **A.** Yes. [1554:12] **Q.** That meant in an auto-deleveraging situation Alameda could again step in and be another source of capital. [1554:14] **A.** It meant——correct. Specifically, the difference is that if the sole cause of Alameda being an ineligible backstop provider was that it didn't have enough free collateral, that was no longer considered. [1554:18] **Q.** And in your view, Mr. Singh, would that be helpful to customers? [1554:20] **A.** Yes. [1554:21] **Q.** Why? [1554:22] **A.** In my view at the time, it would be helpful to customers. [1554:23] **Q.** Why? [1554:24] **A.** Because it would prevent the undesirable ADL as opposed to backstop fills. [1555:1] **Q.** Did you ever hear the term "code commits"? [1555:2] **A.** Yes. [1555:3] **Q.** What does that mean to you? [1555:4] **A.** A code commit is a package change of code submitted to a code base. [1555:6] **Q.** Did you ever author code commits? [1555:7] **A.** All the time. [1555:8] **Q.** Okay. And were they——were these like notes you made for other developers to see? [1555:10] **A.** There were elements of the code commits that were. [1555:11] **MR. COHEN:** If we could call up for the witness——witness only——DX 1102. [1555:13] Pursuant to Defense Exhibit S-3002, it's been stipulated that this document—— [1555:15] **THE COURT:** I'm sorry. DX 3002? [1555:16] **MR. COHEN:** Yes, that's the stipulation as to the authenticity of this document. [1555:18] BY MR. COHEN: [1555:19] **Q.** Mr. Singh, can you take a moment and take a look at what's been marked as Defense Exhibit 1102. [1555:21] **A.** I see it. And forgive me. This isn't what you asked, but—— [1555:23] **THE COURT:** Just a minute. [1555:24] **THE WITNESS:** Yes. --- **About this text version** This is a text version of the court reporter’s transcript of the testimony of Nishad Singh at the trial of Samuel Bankman-Fried, *United States v. Samuel Bankman-Fried*, 22 Cr. 673 (LAK), before Judge Lewis A. Kaplan and a jury in the United States District Court for the Southern District of New York. The government called Singh as a witness on October 16, 2023, and he was excused on October 17, 2023. His testimony runs from page 1298 to page 1623 of the trial transcript, which was prepared by Southern District Reporters, P.C. **Note by this site, not part of the document:** Direct examination by the government begins at trial transcript page 1298 (ECF No. 366), cross-examination by Bankman-Fried’s counsel at page 1492 (ECF No. 368), and redirect at page 1618. Singh’s account of when and how he learned of the hole in customer funds, his choice to stay and become complicit in the conspiracy, and his actions in the short period before the collapse is mainly at pages 1388 to 1417 and 1456 to 1473. - This is part 10 of 13 of the text version, covering pages 1532 to 1555. The document is divided so that software which reads only the beginning of a long page can still read all of it. Other parts: [part 1](trial-testimony-2023-10-16-17-part-1.html), [part 2](trial-testimony-2023-10-16-17-part-2.html), [part 3](trial-testimony-2023-10-16-17-part-3.html), [part 4](trial-testimony-2023-10-16-17-part-4.html), [part 5](trial-testimony-2023-10-16-17-part-5.html), [part 6](trial-testimony-2023-10-16-17-part-6.html), [part 7](trial-testimony-2023-10-16-17-part-7.html), [part 8](trial-testimony-2023-10-16-17-part-8.html), [part 9](trial-testimony-2023-10-16-17-part-9.html), [part 11](trial-testimony-2023-10-16-17-part-11.html), [part 12](trial-testimony-2023-10-16-17-part-12.html), [part 13](trial-testimony-2023-10-16-17-part-13.html). The whole document on one page: [trial-testimony-2023-10-16-17.html](trial-testimony-2023-10-16-17.html). PDFs of the court reporter’s transcript for the two days, each holding the whole trial day except the pages sealed by the court, are held here: [October 16](pdf/2023-10-16-trial-transcript.pdf) and [October 17](pdf/2023-10-17-trial-transcript.pdf). The copies filed on the court’s docket are on CourtListener: [ECF No. 366](https://storage.courtlistener.com/recap/gov.uscourts.nysd.590940/gov.uscourts.nysd.590940.366.0.pdf) and [ECF No. 368](https://storage.courtlistener.com/recap/gov.uscourts.nysd.590940/gov.uscourts.nysd.590940.368.0.pdf). This part as plain text: [trial-testimony-2023-10-16-17-part-10.txt](trial-testimony-2023-10-16-17-part-10.txt). The other documents from the case: [index](index.html). - It was made from a text extraction of the transcript for the two trial days and then compared by machine, line by line, with the PDFs of the court reporter’s transcript that are held here: each of the 7,473 lines of the testimony holds the same words at the same page and line number as in the PDF. The held PDFs do not bear the filing stamp of the court’s docket; the copies filed on the docket are ECF Nos. 366 and 368. If this text differs from the official transcript, the official transcript controls. - A marker such as `[1604:12]` gives the transcript page and line at which the text after it begins. There is a marker at the start of every paragraph and wherever a new page begins, so `[1604:1]` is the top of page 1604. Court filings cite the same page numbers: “Trial Tr. at 1604” means page 1604. - The original is typed 25 numbered lines to a page. Here each paragraph runs on as ordinary text. The line numbers, the reporter’s running header and footer, and the double spaces of the typescript are not reproduced. The line number in each marker is the number printed beside that line in the PDF. 18 pages are not full. The reporter’s note “(Continued on next page)”, which stands at the foot of 17 pages, is left out. - It covers the testimony from the moment the witness was called until he was excused, including the conferences at the bench and the discussions between the judge and the lawyers that took place in between. The places where the transcript for these two days is not reproduced, and the pages that are sealed, are noted in italics. - A paragraph begins here where the typescript labels a question, an answer or a speaker, or indents a line. The heading of each examination and the caption that names the witness are each kept as one paragraph. - The speaker labels, including Q. and A., are set in bold for readability; the original does not do that. The two headings that give the date were added. Dashes are reproduced as they appear, which differs between the court reporters who took turns: some typed `--` and some typed long dashes. - Nothing has been corrected.