The Machine is Always Watching | Steve Sokolowski

Introduction

Another season has arrived, and the litigation web has expanded significantly since the last update. There are now six causes of action across five cases. And, better than a spy novel, Gemini uncovered evidence suggesting that there is more than meets the eye.

This post is separated into three main parts. First, we review how the web has exploded and discuss the facts behind each of the cases. Second, we explore several massive technological advancements—particularly GPT-5 Pro— and explain why the current state of the adversary proceeding is a direct result of these advancements; the machine is always watching. Finally, we discuss the situation with the Genesis Wind-Down Debtors, a new party that has become adverse to us, and the immense and sad wastefulness of how that came to be.

1. September update

The current situation

The litigation web started out as a relatively simple case in Pennsylvania against three defendants—DCG, Silbert, and Moro—for the fraud they participated in against us. It has now expanded to six different causes of action, with a seventh imminent:

The original Pennsylvania case.
The backup Connecticut case, which was stayed on September 12, 2025. The stay was not opposed by the defendants.
An adversary proceeding in the New York bankruptcy court, filed by DCG working together at least initially with the creditors' representatives (the Debtors), seeking to stop the two cases above.
A new action, UBMI v. PROHASHING, filed two weeks after a Pennsylvania brief stated that PROHASHING possesses "terabytes of data," in which PROHASHING denies all allegations and which has defects we have pointed out in our opposition brief to the adversary processing's preliminary injunction motion.
A debt collection lawsuit from Wells Fargo that started the exact same day the adversary proceeding was filed.
A counterclaim in the adversary proceeding filed by us to bind the Debtors to the adversary proceeding’s decision, after the Debtors stated that it was “aligned on the requested relief” from DCG’s adversary proceeding.
(Not filed yet) The Debtors have stated that they are exploring the idea of joining Jefferies Leveraged Credit Products LLC in the adversary proceeding as part of an assignment defense. However, even after we asked them, Jefferies itself has not asserted that it owns our claims.

The Future of the Adversary Proceeding

The adversary proceeding is at an inflection point. It could become a massive case, similar to a construction liability situation. In those cases, one might see that a building collapses and kills a person, so the family sues the contractor, who then files a claim to blame the roofer, who then files to place blame on the roofing tile manufacturer, who then counterclaims against the contractor for improper installation—and eventually the case ends up with hundreds of defendants and lasts for a decade.

Or, the adversary proceeding could wither and die quickly.

Procedurally, the critical difference may be whether the Jefferies bankruptcy claim sale contract enters evidence. To be clear, not only can I defend the contract's text, the contract itself is irrelevant for a number of other more complex reasons. The easiest reason to understand is what we all learn in first grade - contracts usually don't bind people who haven't signed them. Additionally, one cannot enforce a contract on another person's behalf. Every other party in this litigation - DCG, Silbert, Moro, and the Debtors - has been attempting to enforce a contract between Cryptocurrency Management LLC and Jefferies - yet both of those parties aren't trying to enforce it themselves.

But if the Genesis Debtors are successful in expanding the counterclaim to a massive debate about standing using this contract, the models have projected a scenario, of a low but increasing probability, where many of DCG's defenses are adjudicated in the bankruptcy court except for personal jurisdiction (which would then become irrelevant). We don't want the merits of our case to be decided in the bankruptcy court - but we also cannot afford any "do overs" later in Pennsylvania because the right parties weren't bound in the first place. In our view, the Debtors are the ones introducing this expansion risk, and we'll discuss this issue below.

You can probably guess by now that there is a lot more to this adversary proceeding than what meets the eye - most importantly, how it came to be. And the only reason we figured that out is because of the next section - AI usage.

2. AI changed the game

Standard advances in AI

First, we'll review some of the less exciting, but still technologically significant, reasons for AI advancement.

Our opposition brief to DCG’s preliminary injunction (https://stevesokolowski.com/sokolowski-v-fraud/documents/pi-opposition-brief.pdf) is the first brief we wrote with the assistance of GPT-5 Pro. While we obviously checked every case and spent 300 hours ensuring that we understood the arguments and that they were in line with what we wanted to say, we concluded that an “AI-first” workflow is now feasible. That’s in comparison to May, when we filed our opposition brief to DCG’s motions to dismiss: then, we had to write the brief ourselves and use models to ensure we weren’t making mistakes.

There’s one astonishing advance that changes the game—GPT-5 Pro has a 100% case-correctness rate. Hallucinations regarding cases are gone—a thing of the past. Those lawyers who were sanctioned for making up cases will probably remain in judges’ minds for years, but GPT-5 output is, without question, superior to what a human lawyer can produce. The nuances in cases that the model can understand are unbelievable, and we’re constantly amazed at how it not only never makes mistakes, but it corrects our plain reading of the case text by pointing out something else that changes the context of a paragraph. The only issue we found with the output briefs was "overkill;" we had to actually delete many of the correct cases the model cited.

We would suggest to readers that they compare some of the adversary proceeding briefs from September, which are posted on this site. Contrast our opposition brief to the Debtors' motion to dismiss. Focus on which briefs you believe are more concise, have the more relevant case law, and tell a better story, and decide for yourself.

So as we plan for the future, what we always consider now is "what will the models be able to do in X months?" In particular, we're asking "by the time discovery opens into the first and second schemes, will there be an open source model we can run on our servers that is comparable to GPT-5?" We believe that answer is "yes," and if it is, then we will be able to accomplish the work of 100 paralegals with just the two of us. We are assuming that the discovered data will be confidential, so we're already setting up the servers.

Using models to optimize strategy

Now, we get to the main focus of this post. This is a blog about how AI is used to pursue those who stole our life savings and to hold accountable those who are attempting to stand in our way of doing so.

As early as July 1, the day DCG added a footnote in its Pennsylvania brief about starting an adversary proceeding that had never been discussed before, we immediately suspected that something was going on that was not being fully and publicly disclosed. Once the adversary proceeding was actually filed, our suspicions were confirmed and we started writing a motion to dismiss (which we later discarded) using Gemini Pro 2.5. Without us asking, Gemini Pro 2.5 immediately pointed out that there was language in multiple briefs in another case containing strange phrases - most notably, "property of Genesis Global's estate." DCG was arguing that our claims were "derivative," which means that the estate owned them. Vincent Falco, the other defendant in the adversary proceeding, wrote in his own complaint, in the fifth paragraph, that the bitcoins he was suing over had become "property of Genesis Global's estate."

Shortly after the adversary proceeding was filed, the Debtors moved to dismiss the case against us. Their argument was that only they had the right to sue who they called "creditors," not DCG; they just weren't going to do it at this time.

We used multiple models to determine what the optimal strategy was given this new information. After days of simulations into how the adversary proceeding would unfold, we determined that neutrality was overwhelmingly the best strategy, leading to a "Statement of Neutral Position."

But then the other parties did something extraordinary - they started blaming each other for starting the adversary proceeding in their own briefs. The Debtors, for some reason that to this day is baffling to us, had decided in early July that they were "aligned on the requested relief" with DCG.

The Litigation Oversight Committee was using money that would have been distributed to creditors to fight Falco, who appeared at that point to be an innocent victim, and us. They were spending this money to protect DCG from accountability for the DCG's original fraudulent scheme. A reasonable equivalency is that DCG's legal bills in defending against our case were being reduced with the help of creditors' money. The creditors are also generating a massive record that will harm them against DCG in Delaware.

It's obvious why DCG wanted to enjoin these suits - they don't want to pay out money. The Debtors helping DCG while at the same time claiming that DCG owed them billions of dollars in Delaware seemed to us like it was cruel. We later termed this the "second scheme," because though the extent of this new scheme is not clear and it certainly doesn't involve a promissory note like the first scheme, it is a continuing pattern that involves the same two entities participating in a scheme to take or stop us from recovering our money.

It was clear that we needed to force a final decision on the "direct" issue. If we didn't force the decision, we could end up with the Debtors coming back after years of work and trying to get all of the money. o3 suggested a counterclaim as the best way to do that.

The facts

Before we get started with how AI pieced together what was really going on, here's a snapshot of the events that occurred. A subset was input into Gemini, which then led to additional paths to investigate, which were then input into Gemini again, and eventually we ended up with this chain that we believe is important.

Key events

December 6, 2024 — We requested to meet with the Debtors/Estate and had a friendly conversation.
March 4, 2025 - DCG accuses us of "criminal bankruptcy fraud," threatening sanctions and citing felony statutes.
May 6, 2025 — Vincent Falco files a consumer lawsuit against the same defendants we did. In his complaint, he describes the bitcoin at issue as "property of Genesis Global's estate."
May 13, 2025 — The Genesis Debtors sue DCG in Delaware and the complaint is sealed until June 11.
May 17, 2025 — We discuss aspects of our case with Genesis representative Amelia Alvarez; she cheers us on.
June 13, 2025 — UBMI sues PROHASHING for $2 million "upon information and belief," two weeks after our opposition brief states that PROHASHING has "terabytes of data."
July 1, 2025 — The Debtors E-Mail DCG that, although they disagree with some points in DCG’s draft, they are “aligned on the requested relief” and would likely file a “me too” submission supporting an injunction. This is on the public record and proved that the Estate was working with its supposed enemy in a second scheme to stop victims from holding DCG accountable. They exchanged "several more E-Mails" over a week on a "common interest basis."
July 7, 2025 — DCG and the Estate finish submitting briefs in their Delaware case.
July 7, 2025 — In the Falco case, the judge issues a sua sponte order. The order from that judge states that there had been an "elongated briefing schedule" in that case - lasting nearly five months - for a simple motion to dismiss. Typical federal rules would conclude briefing in just one or two months. The briefing schedule would have ended after the adversary proceeding's preliminary injunction hearing, so the judge would not have ruled on the Falco case's dismissal until after a ruling in the adversary proceeding. This unexpected order instructs Falco to amend his complaint by July 25. This order was missed by us until Deep Research surfaced it when we asked it to infer what the parties were arguing on the Falco docket.
July 8, 2025 — DCG files the bankruptcy adversary proceeding in New York and a motion for a preliminary injunction that, if granted, would halt our Pennsylvania and Connecticut consumer cases. Three seconds before the case was filed, the Debtors change their mind and stop supporting DCG, stating their belief that DCG was misguided and was using the adversary proceeding as a means to influence their Delaware case.
July 8, 2025 (same day) — The Wells Fargo case is filed in the Centre County Court of Common Pleas (Pennsylvania). This debt originated from the Genesis failure and this suit was served ten hours before the adversary proceeding.
July 14, 2025 — The Debtors formally change their position and file a motion to dismiss the adversary proceeding, arguing that they (not DCG) should be the ones who get to choose whether our case proceeds. They did not state that our case should be allowed to proceed. We initially viewed this as the Estate being on our side for this particular issue; o3 did not and that gave us pause.
July 16, 2025 — Chris tries to contact Vincent Falco's attorney and she does not respond, which we found strange.
July 25, 2025 — Falco dismisses his case rather than amending his complaint as the judge ordered on July 7.
July 28 and July 30, 2025 — The Debtors meet with us. Jennifer Selendy, partner at Selendy Gay, which represents the Debtors, states "to be honest, [some of] the claims are different from those of the estate" and says that our Pennsylvania case is a direct claim, implying it is not owned by them. But after the meeting, Vijay Boyapati contacts us and says they are always willing to coordinate on claims that belong to the estate, "as [they] did several times with Falco."
August 12-13, 2025 — The Debtors publish a brief, in which they refer to us as "creditors" tens of times and includes those E-Mails about "alignment" with DCG. This causes us to take a "neutral" position in their fight with DCG. The crafting of the "Statement of Neutrality" was arrived at after a day of o3 running repeated Monte Carlo simulations based on all the briefs in its project data to determine the highest winrate in the overall litigation web (not just the adversary proceeding).
August 15, 2025 — A status conference is held. We file an answer with counterclaims against the Estate to force a final decision, based on the recognition that a dismissal of the adversary proceeding would be insufficient; the Debtors might just come after us later. Benjamin Kaminetsky, DCG's attorney, makes a bizarre statement that we may be "violently aligned" with them. We both wrote this statement in our notes. This is an example of a remark that a human would have seemed odd and brushed off, but Gemini repeatedly came back to this as meaning something more.
August 17, 2025 — The remark that broke everything wide open (but which we didn't realize was important until Gemini said so) is made. Boyapati states that the Falco complaint, which was "obviously derivative," had been "copied and pasted" from their own Delaware complaint. The models picked up immediately that there was a timing mismatch - the Falco complaint had been filed a week before the Debtors' complaint. This single comment in the notes is what led the models to begin inferring more and more and searching the Internet for more evidence.
August 17- 18, 2025 — We hold several meetings with the Debtors, after which we stated to them that we believed they had been "not credible." During these meetings, litigation chair Boyapati said, among other things: he and Selendy Gay had had miscommunications, that he never intended to enjoin our cases but that the judge might enjoin them, and that the use of the word "creditors" in their briefs was a misunderstanding. Jennifer Selendy, the Debtors' attorney, agreed that she and Boyapati either "hadn't read" or "hadn't really understood" our case at the time DCG filed the case to enjoin it. They stated they believed the proceeding had been focused on Falco. These statements directly contradicted the public record in multiple ways - for example, the first paragraph of the adversary proceeding refers to us, and their own brief states that they were aware of our case "since [its] inception." Boyapati had "liked" our posts on X in March.
August 18, 2025 — Most importantly, during that meeting, the Debtors' lawyer Selendy requested a bankruptcy claim assignment document, stating that if we gave it to her, she would then consider some sort of stipulation that would label the claims direct or put an end to the proceeding (the exact wording or method wasn't discussed.) The reason she told us she needed the document for contradicted what she had written in their briefs, which led us to decline. This critical decision was also AI-assisted; we asked them for the probability that DCG would use the contract to start a seminar on New York contract law, but the models actually found the specific statement in the specific brief that was a contradiction.
August 18, 2025 — During this meeting, the Debtors' representatives also claimed that DCG would not suffer any adverse effects by losing the case on a "direct" ruling. This led us to review DCG's briefs, and we discovered that DCG had not referenced a single Pennsylvania case in them.
Late August — The Debtors participate in an E-Mail chain requesting that the hearing be postponed so that they could subpoena the bankruptcy claim sale contract they asked for on August 18. After a week of back and forth, the judge declined the postpone the hearing for this discovery.
September 6, 2025 — We post our opposition brief with this timetable (https://stevesokolowski.com/sokolowski-v-fraud/documents/pi-opposition-brief.pdf)
September 8-11, 2025 — The Debtors decide to send out subpoenas for the bankruptcy claim sale contract anyway. We object to them. They ask for a witness to testify at the hearing. We ask them for basic information about who the witness is; they didn't respond to even tell us what his job title was. We contacted Jefferies and had several discussions over E-Mail, phone, and Zoom. Jefferies ultimately decided not to produce the bankruptcy sale contract.
September 11, 2025 — We requested a discovery conference about how the subponeas had been issued anyway. While no conference was held, the preliminary injunction hearing was indefinitely adjourned.

This timeline has been presented publicly in the record. Both sides - DCG and the Debtors - have had three weeks to respond, and have each submitted one brief. Neither has submitted a counter declaration disputing the sworn statements we made regarding this timeline above.

Extraordinary pattern matching

As you can see, after every meeting that isn't protected by privilege or an evidentiary rule, and which we can't record, we spend up to an hour writing detailed notes. For a typical meeting, the notes stretch across two pages of 12 point single spaced text. We have numerous notes like this dating back for some time. We aren't making the full notes of these meetings public on our own initiative, although we did offer to give them to the judge and have no issue making them public if he orders.

After our final meeting on August 18, we came to believe that we had a different interpretation of events than the Debtors did. But there was a key issue that we were having problems with. There were a lot of things that just didn't add up in what everyone was saying; the public briefs in the adversary proceeding said one thing, and there was a second narrative that was completely inconsistent across everything else.

So we told Vijay we would get back to him later, and spent 36 out of the next 72 hours with constant churning away using Gemini Pro 2.5's million-token context window. We input over 60 documents - all the relevant documents from the Pennsylvania, Connecticut, UBMI, Wells Fargo, Falco, adversary proceeding, and other cases, our notes, the CM LLC operating agreement, various evidence, their Delaware complaint, and more. Each question would take as long as 3 or 4 minutes to answer. We would often run the same prompt 3 times in a row to see what would come out. We also added 40 documents to OpenAI's tools and condensed multiple documents into one to fit within their 40 document limit. We probably cost these companies $100 or $200 in electricity.

We didn't need very long. In the first half hour, a critical comment that we had missed was picked up - Boyapati had stated that the Falco complaint was "copied and pasted" from the Debtors' Delaware complaint, and the timing mismatch. We won't speculate specifically on with Boyapati meant when he made that statement, but even if we had the wrong impression, the comment "broke open the case" to start exploring how everything fit together.

Once the context window would get full, we would select the most important insights from the previous runs, and then attempt to connect further from whatever that was to the next step. A Deep Research run about Falco's case revealed the timing of Judge Cote's sua sponte order in relation to the other events, and Gemini also pointed out the "elongated briefing schedule" text and explained why that could be important - something we hadn't understood initially. Other important connections started to fall into place - the number of days required for a cross-country demand letter in the Wells Fargo case to be delivered, statements made in the Delaware briefs, contradictions between the meeting notes and public statements, and more.

Three days later, we had pretty much figured out exactly what was going on. We then wrote "The Scheme" (https://stevesokolowski.com/songs/the-scheme/), which was inspired by the breakthrough.

Unfortunately, we can't share the results of our research publicly, because there is always a chance we could be wrong. We are only able to point out the patterns the models picked up on, without the connecting inferences, and post the evidence, which is what we did in our opposition brief. At the very least, we now know exactly who and what we need to subpoena during discovery in the Pennsylvania case to fill in the blanks and prove the missing links.

Planning for a previous paradigm

While we haven't won yet, what has happened so far would simply not have been possible before large-context models were released. And that's exactly the reason why the "agreement" between DCG and the Debtors fell apart. They, like many of the attorneys we talk to, are operating in a world before superintelligent AI existed (i.e. before May.) In fact, however narrow or widespread the "second scheme" ends up being, it undoubtedly started* before Gemini 2.5 and GPT-5 were released and would have been known to be able to piece it together.

In that pre-Gemini world, an associate would be required to read thousands of pages of briefs to figure out which were relevant. Footnotes in an unrelated case would be missed. It wouldn't be clear what exactly even to read, and by the time something important is read, it might be forgotten. Notes taken about one brief would be useless because the reader might not understand what is important enough to write it down.

We noticed that lawyers like to bury inconvienent facts in footnotes - like the comment in DCG's reply brief in the Pennsylvania case arguing that the denial of the NYAG case's MTD was based on different elements. To LLMs, footnotes are the same size as everything else.

LLMs can chew through massive briefs in two minutes and pick out suspicious sections instantly, cross-referencing them with our own notes or the summarized findings of a previous context window. GPT-5 Pro will search 100+ websites in a single query in addition to reviewing the 40 large documents in its vector database.

The firms involved in this mess likely embarked on their behavior thinking that nobody could connect the dots. And they were right - no human could. This pattern matching has significant implications for the future - not just in the law - but in using publicly available data to make it dramatically harder to engage in any type of hidden scheme.

Where the models were wrong (and what we learned)

The models have, so far, accurately predicted the behavior of all of our adversaries thoughout the litigation web, in terms of court filings. They have not made a single error in their prediction of how any party would respond to any document we filed based on the wording we used.

They have made only two errors. First, they predicted that the judge would hold the discovery conference and quash the subpoenas until the conference; he did not do that. We don't criticize the court; we point this out, though, because the models said there was a 95% chance he would, and after we told them what happened, the models have been unable to produce insight as to why they were wrong. In the end, the result was the same - Jefferies declined to produce the contract.

The second error is that the models predicted that the defendants would initiate settlement discussions rather than lose the adversary proceeding - specifically, right before the hearing (which was adjourned.) By contrast, the complete opposite is true - while we aren't contacting them about a settlement, the Debtors have not responded to any E-Mail about anything at all we sent them in three weeks, even for basic questions like what the job title of the witness they were calling was. So, while we aren't looking for a settlement, nobody would talk to us anyway if we were, and they clearly don't want one.

The models have consistently overpredicted what we would call "reasonableness" of everyone involved. We suspect that the models' training data likely consists of public court filings, which obviously paint a picture of formal, straightforward litigation. This training data would never include the outright contradictions that we have been presented with by the Debtors in their private meetings and the constant and blatant misuse of the system, like sending the wrong type of subpoenas. It also wouldn't be able to trained on missing data that doesn't exist - for example, the lack of a response from Falco's attorney - and learn that such things, as we've discovered, are common behind the scenes.

3. The sad state of affairs

A simple overview and a misunderstanding

The current state of affairs can be summarized accurately in one sentence: three years to the day after DCG participated in a massive fraudulent scheme with the CEO of Genesis that ruined our lives, the new management of Genesis "aligned" with DCG yet again to try to ruin our lives yet again by enjoining our case to recover for the first scheme. We don't yet know the full extent of the alignment and how many law firms and individuals are involved, and even if we somehow got it completely and utterly wrong and this is a bizarre sequence of coincidences, but it's clear at a minimum that DCG is engaged in a continuing pattern of conduct that we intend to use to show the defendants in Pennsylvania should owe full treble damages. Our position is that the conduct, if proven, would be so egregious that what would be the largest verdict in Pennsylvania consumer protection history would then be not only appropriate, but mandated.

We are also exploring the expansion of the litigation web to include different claims involving a number of parties. We mentioned this in our opposition brief, but we can't discuss this now unless we obtain more evidence. We may borrow money against our stock holdings - with the Fed lowering rates this is now possible - to bring on a firm to assist with the UTPCPL case. The UTPCPL case allows us to recover those attorney's fees to pay back the loan, while we would then focus on the new claims that wouldn't provide for fees.

Despite all this, it still seems that it isn't getting through to all the numerous parties who are or may eventually be proven through discovery to be involved in the second scheme that we simply can't "give up." If it isn't clear already from our previous posts, we have two outcomes here: we win, or we go bankrupt and lose everything. Often, in litigation, the goal is to come to a settlement. But it is clear from our sworn declarations that many of the people with whom we are dealing are not, unfortunately, negotiating in good faith. Good faith negotiations involve telling the truth. We have repeatedly attended meetings with some of these parties in the hopes of being their allies or coming to an agreement. There's no possible way to negotiate anything if all parties aren't telling the complete truth; we honestly don't know how the legal system even functions given the deficit between what the rules say and how many lawyers actually act.

Once UBMI and the adversary proceeding were both filed, we were faced with two problems. First, the UBMI case will be so costly to defend, even though it is weak, that it will cause PROHASHING to file for Chapter 7 bankruptcy. If that happens, we are in for years of cleaning up the mess and responding to requests for data and so on. Second, if there is a "derivative" or "assignment" ruling in the adversary proceeding, then we are still in for years of defending ourselves to provide as little testimony and information as legally permitted to the people who are trying to prove that we relied on Genesis's statements, because we could make a mistake exposing ourselves to liability for no gain. So we can't just "quit, get a job, and move on." This is what we decided to devote our life to; there's no backup plan.

Therefore, you can see that the only "way out" is to win. If we win the main case, then we've overcome the derivative/assignment rulings, and can afford to spend the money defending ourselves against UBMI and/or the PROHASHING estate.

The immense harm to the Genesis creditors

We truly do not understand why the situation has unfolded the way it has. The Debtors are accusing DCG of having defrauded them of billions of dollars. At the exact same time they are suing each other, they have—at a minimum—been on the same side for some time, two of the most expensive law firms in the world, to crush us. Although the Debtors are supposed to be working for creditors’ benefit, they have essentially set aside that goal and shared information to help DCG in its adversary proceeding (and we don't even know the extent of what will show up in discovery). And their objective doesn’t really make any sense.

They already said in their personal discussions that the UTPCPL claim was "different from those of the estate." Let’s just review for a minute the consequences of what the Debtors have done over the past few months:

They have spent hundreds of thousands of dollars fighting this adversary proceeding that could otherwise have been distributed to creditors.
It doesn’t seem like they thought through the actual consequence of a “derivative” ruling before aligning themselves with DCG, and either didn't read the complaint or lied about not reading it.
The Debtors have made many statements that cannot be revoked which have significantly reduced the expected value for creditors in their Delaware case regardless of what the end ruling is in the adversary proceeding.
By making statements to us that are contradictory and not credible, and now by refusing to respond to our messages, any ability to work with them to uncover evidence of DCG's first scheme has been lost.
They may yet succeed in turning the adversary proceeding into a massive discovery battle over contract interpretation law that could become a proxy battle that delays creditor distributions for years.

We are not lawyers, but we would suggest that there is a colorable argument that the Debtors have already, through their own actions, reduced distributions to creditors by many millions of dollars.

Conclusion

The current state of the litigation web is a direct consequence of the development of AI. Were it not for models' large context windows and their ability to cross-reference enormous amounts of data, the suspicious dates and times on the timeline would never have become clear, and the preliminary injunction hearing would never have been indefinitely adjourned.

Of course, a model can't make up information out of thin air. We are about 90% certain that we know exactly what the motivations of the parties were and what the full extent of the circumstances leading to the adversary proceeding involved.

We can't prove it yet, but it's going to come up in discovery somewhere. We intend to seek discovery into DCG and the Debtors' alignment, the Falco and UBMI complaints, the subpoena crisis of September 8, and more. If the Estate pushes for bankruptcy claim sale discovery in the adversary proceeding, then this other evidence will come up there. If that proceeding ends with a "direct" ruling, it will come out to prove damages in the Pennsylvania case. If not there, we'll get to the bottom of it with a new case or motion. And even if we lose all of that and go bankrupt, discovery in the Delaware litigation between the Debtors and DCG will give everyone a good idea of what really happened. We'll share what isn't confidential or sealed as the process unfolds.

We have also archived a document where we have written out what the models inferred the scheme was. While we can't make unfounded allegations now, we will publish the document as proof the models were right later, or state that they were wrong. And, whatever the truth, we now MUST get to the bottom of it, because to do otherwise would be a signal that it is acceptable to abuse the legal system against us in the future.

On January 2, 2025, the first day, a lot of lawyers laughed at us online, suggesting that AI tools would be a terrible partner. As it turns out, this second coordination between DCG and the Debtors would have been entirely undetected were it not for Google Gemini. A year ago, a human lawyer would likely have missed it entirely.

It's time for people to stop laughing, because neither the second scheme, even in its best-case scenario, nor the models, are a joke.