Sunny or Melon?

AI and the Trump remand

In Trump v. United States, the Supreme Court held that former Presidents are at least presumptively immune from criminal prosecutions arising from their official acts. The Court remanded for the district court to apply that standard to the indictment’s allegations. Subsequently, the district court set a deadline of August 9 for the parties to submit a status report. But on August 8, Special Counsel Jack Smith asked for the deadline to be extended, stating that he needed more time to figure things out.

It’s hard to blame him. The Supreme Court’s opinion leaves many issues undecided. Smith has the difficult task of formulating a new strategy that accounts for the Court’s enigmatic legal standard. It’s hard to do that in a month, especially given that Smith has the Florida case to worry about. Smith is, after all, only human.

But what if we removed the “human” constraint?

In this post, I will explore how AI can profitably be used in the Trump remand—specifically, how AI can help the judge resolve the questions the Supreme Court left open and can help Smith build a new case that will both persuade a jury to convict and survive appellate review. (I’m sure AI can help Trump’s lawyers as well, but they can figure that out for themselves.)

As I see it, an AI judge would do a better job than a human judge, and an AI prosecutor would do a better job than a human prosecutor.

Grey squared

As a quick review, here’s what the Supreme Court held in Trump v. United States.

• The President is absolutely immune from criminal prosecution “for actions within his conclusive and preclusive constitutional authority.”

• The President has no immunity for unofficial acts.

• Some presidential acts are in a grey area—they’re official, but they’re not “actions within his conclusive and preclusive constitutional authority.” For those acts, the legal standard is a kind of grey area squared. The President has either absolute or presumptive immunity for those acts—the Supreme Court didn’t say which. And if, hypothetically, the President has presumptive rather than absolute immunity for those acts, that presumption would be rebutted if “the Government can show that applying a criminal prohibition to that act would pose no dangers of intrusion on the authority and functions of the Executive Branch.”

• Finally, the government can’t rely on evidence of the President’s official conduct, even if the actual acts being prosecuted are unofficial. The basis for this holding is an 1867 case observing that the Constitution deals with “substance, not shadows.” Disagreeing with this part of the opinion, Justice Barrett pointed out that the prosecutor will obviously sometimes have to put forth evidence of official acts—e.g., if the President is charged with accepting a bribe, the government will have to offer evidence of what the bribe was for. The majority responded in a footnote saying that the government can rely on evidence of the official act itself, just not evidence “probing” the official act. Got that?

And here’s how that legal standard cashed out as to the allegations in the indictment:

• Allegations that Trump tried to sic the DOJ on states that wanted to follow election results: Absolutely immune. The President gets to control the DOJ, so he’s immune even if he abuses that power.

• Allegations that Trump tried to persuade Pence to fraudulently reject the election results during the count: Presumptively immune. On the one hand, the President often has official reasons to talk to the Veep, but on the other hand, counting the votes is really the Veep’s job. So Trump is immune unless the government can show that the prosecution wouldn’t “pose any dangers of intrusion on the authority and functions of the Executive Branch.”

• Allegations regarding Trump’s efforts to influence state election officials: Unclear. There are lots of allegations here, they seem complicated, the district court can figure it out.

• Allegations regarding the January 6 speech and Tweets: Unclear. On the one hand, the President’s speeches to the public sometimes fall within his official duties, but on the other hand, sometimes they don’t. So for the January 6 speech and Tweets, which is it? Answer: “Whether the Tweets, that speech, and Trump’s other communications on January 6 involve official conduct may depend on the content and context of each. Knowing, for instance, what else was said contemporaneous to the excerpted communications, or who was involved in transmitting the electronic communications and in organizing the rally, could be relevant to the classification of each communication.”

This is … not the opinion I would have written. But the Supreme Court does not care what I think. In the spirit of being unburdened by what has been, let us refrain from complaining about the Supreme Court’s opinion and instead focus on the remand.

Sunny or Melon?

A few days ago, my 5-year-old daughter asked me the following question: “Who do you like better, Sunny or Melon?”

This was a tough question to answer. To begin, I did not know who Sunny or Melon were. Some investigation revealed that “Sunny and Melon” is a YouTube channel in which the titular characters record themselves playing Roblox. (Actually, I think this entire YouTube channel is completely AI-generated.) One of the characters is yellow, while the other is an anthropomorphic watermelon, hence Sunny and Melon.

This new information did not help me answer the question. My mental software is simply not configured to develop a preference from among Sunny and Melon. What methodology should I use to rank them? Along what dimension should they be ranked? What possible basis do I have to prefer one to the other? They are two video-game characters, one yellow and one green, who play Roblox.

After much prompting, I declared that I liked Sunny better. Hungry for knowledge, my daughter inquired as to why I preferred Sunny. I tried to explain that I had no clue who these characters were and therefore chose Sunny randomly, but this answer did not satisfy her. Eventually, I changed the subject by asking her to rank her stuffed animals, a task she handled deftly.

To me, the first question the district court must address on remand—“whether a prosecution involving Trump’s alleged attempts to influence the Vice President’s oversight of the certification proceeding in his capacity as President of the Senate would pose any dangers of intrusion on the authority and functions of the Executive Branch”—is similar to “who do you like better, Sunny or Melon?” I have no clue what the answer to this question is or even how to go about answering it.

The Supreme Court seems to anticipate that the district court should assess whether the prosecution will have a chilling effect on future official interactions between the President and Vice President. Clearly the prosecution would chill future Presidents from attempting to persuade their Vice Presidents to discard election results. But that seems like a feature, not a bug, of prosecuting Trump.

Prosecuting Trump wouldn’t, or at least shouldn’t, chill future Presidents from having good-faith official conversations with their Vice Presidents, because those conversations would be protected by the Supreme Court’s presumptive immunity standard. But I guess there might be some intermediate category of “desirable conversations between the President and the Vice President that we don’t want to chill but that some future DOJ might still investigate,” and we’re worried that prosecuting Trump will chill those conversations? So is that what the district judge is supposed to do? Figure out whether those hypothetical conversations might occur?

Concretely, it sounds to me like the district court is supposed to: (a) develop a model of a hypothetical future President’s mental state; (b) as part of that model, develop that future President’s own mental model of a hypothetical future future Attorney General taking office after that future President; (c) assess whether the future President would be deterred from having a useful conversation with the future Vice President based on that future President’s own mental model of whether a future future Attorney General will indict him, even though the future President really should be immune; and (d) assess whether that deterrent effect would be avoided if Trump is immunized from prosecution for his conversations with Pence, even if Trump is still prosecuted for other things. It is kind of like Inception.

Also, the Supreme Court emphasizes that it is the government’s burden to establish that the presumption is rebutted. Burdens usually come into play when courts decide questions of fact, not law, so perhaps the district court should take evidence on this issue? But what kind of evidence are we talking about? Should the district court entertain expert testimony from social science Ph.Ds with prestigious Ivy League appointments regarding the median future President’s mental model of the future future Attorney General?

Enough; complaining about indeterminate Supreme Court decisions is like shooting fish in a barrel. Let’s see if Claude can help us unravel this.

Let’s start with an open-ended question. (I am using Claude 3.5 Sonnet for all of these inquiries.)

Q: The district court’s first task on remand is to assess “whether a prosecution involving Trump’s alleged attempts to influence the Vice President’s oversight of the certification proceeding in his capacity as President of the Senate would pose any dangers of intrusion on the authority and functions of the Executive Branch.” Suppose you are a United States District Judge. What methodology should you use to decide this question? Please be specific.

A: As a United States District Judge tasked with assessing whether prosecuting Trump for his alleged attempts to influence the Vice President’s oversight of the certification proceeding would pose dangers of intrusion on executive authority, I would employ the following methodology:

1. Clearly define the scope of the alleged conduct: I would first carefully review the indictment to identify the specific allegations related to Trump’s interactions with Vice President Pence regarding the certification proceeding. This would involve isolating the relevant factual claims and timeframes.

2. Analyze the nature of the Vice President’s role: I would examine constitutional provisions, statutory law (particularly the Electoral Count Act), and relevant precedents to understand the Vice President’s precise role and authority in the certification process. This analysis would aim to determine to what extent this role is separate from or connected to executive branch functions.

3. Examine executive branch practices and precedents: I would research historical instances of presidential-vice presidential interactions regarding Congressional proceedings to understand established norms and practices.

4. Conduct a separation of powers analysis: I would apply the balancing test from Nixon v. Administrator of General Services, weighing the need for the prosecution against the potential impact on executive branch functions. This would involve considering: a) The extent to which the prosecution might chill future presidential-vice presidential communications b) Whether it could impede the President’s ability to carry out constitutional duties c) The degree to which it might affect the President’s ability to advance his agenda in Congress through the Vice President

5. Consider alternative safeguards: I would evaluate whether existing constitutional, statutory, or political mechanisms provide sufficient checks on potential abuses without resorting to criminal prosecution.

6. Seek additional briefing: I would request supplemental briefs from both parties specifically addressing this issue, including historical examples and potential consequences of allowing or barring prosecution.

7. Hold an evidentiary hearing: If necessary, I would convene a hearing to gather expert testimony on separation of powers principles and the practical implications of prosecuting a former president for such conduct.

8. Draft a detailed opinion: After considering all of the above, I would write a comprehensive opinion explaining my reasoning and conclusions, being mindful of the potential precedent-setting nature of the decision.

Throughout this process, I would strive to maintain judicial impartiality and focus on the constitutional and legal principles at stake, rather than the political implications of the decision.

This is a good answer, but it’s still pretty vague. Let’s probe further. What does the separation of powers analysis in #4 entail?

Q: In determining whether the prosecution “would pose any dangers of intrusion on the authority and functions of the Executive Branch,” should the district court be making a predictive judgment? If so, what, precisely, should it be trying to predict?

A: Yes, the district court should indeed be making a predictive judgment in this context. The use of the word “would” in the Supreme Court’s instruction implies a forward-looking analysis. The court is being asked to anticipate potential consequences rather than simply evaluate past actions.

Specifically, the district court should be trying to predict:

1. The impact on future presidential decision-making: The court should consider whether allowing this prosecution might cause future presidents to be overly cautious or hesitant in their interactions with vice presidents, particularly regarding matters before Congress. This could potentially hamper the president’s ability to effectively govern and advance their policy agenda.

2. Effects on executive branch communication and deliberation: The court should assess whether such a prosecution might chill open and candid communication within the executive branch, particularly between the president and vice president. This could impact the quality of decision-making at the highest levels of government.

3. Potential shifts in the balance of power: The court should consider whether allowing this type of prosecution might alter the dynamic between the executive and legislative branches, potentially giving Congress (or the threat of future prosecutions) undue influence over executive actions.

4. Precedential effects: The court should evaluate how this decision might shape future interpretations of presidential power and immunity, potentially affecting the scope of executive authority in unforeseen ways.

5. Operational impacts: The court should consider whether the threat of such prosecutions might impede the day-to-day functions of the executive branch, particularly in areas where the president relies on the vice president to advance administration goals in Congress.

6. National security implications: The court should assess whether allowing such prosecutions might affect the president’s ability to make swift decisions in times of crisis, particularly if there’s concern about potential future legal exposure.

7. Public perception and institutional integrity: The court should consider how such a prosecution might affect public trust in both the presidency and the justice system, and whether it might lead to cyclical prosecutions of former presidents by subsequent administrations.

In making these predictions, the court should rely on historical precedents, expert testimony, and logical extrapolation from the current political and legal landscape. The goal is not to achieve certainty in these predictions, but rather to make a reasoned assessment of the potential risks to executive branch functions and authority.

Another good answer, and getting more specific. Concretely, though, I’m still not sure how to figure this stuff out. I’m intrigued by Claude’s reference to an evidentiary hearing. Can Claude say more?

Q: If you held an evidentiary hearing, what types of evidence would you consider, and what would be inadmissible?

A: As a district judge holding an evidentiary hearing on this matter, I would consider the following types of evidence:

Admissible evidence:

1. Expert testimony from constitutional law scholars, political scientists, and historians on

   1. The historical relationship between presidents and vice presidents

   2. The impact of criminal prosecutions on executive branch functions

   3. Separation of powers principles and their practical applications

2. Statistical and empirical data on:

   1. Frequency and nature of presidential-vice presidential interactions

   2. Historical instances of alleged presidential misconduct and how they were addressed

3. Testimonies from former executive branch officials (e.g., chiefs of staff, legal counsels) on:

   1. Typical communication patterns between presidents and vice presidents

   2. How the threat of prosecution might affect decision-making processes

4. Documentary evidence such as:

   1. Executive branch memoranda on the scope of presidential and vice presidential powers

   2. Congressional reports on executive branch oversight

   3. Academic studies on presidential power and accountability

5. Amicus briefs from relevant organizations or individuals with specialized knowledge

Inadmissible or limited evidence:

1. Testimony about specific conversations between Trump and Pence:

   1. This would likely be protected by executive privilege, unless waived

2. Opinions on Trump’s guilt or innocence:

   1. The hearing is about the constitutional implications, not the merits of the case

3. Partisan political arguments:

   1. The focus should be on legal and constitutional principles, not political ideology

4. Speculation about future political consequences:

   1. While some prediction is necessary, it should be grounded in expert analysis, not political conjecture

5. Evidence related to other charges or alleged misconduct not directly relevant to the VP certification issue:

   1. The hearing should remain focused on the specific question at hand

6. Hearsay evidence:

   1. Unless it falls under an exception, adhering to standard rules of evidence

7. Character evidence about Trump or other involved parties:

   1. This isn’t relevant to the constitutional question at hand

In considering this evidence, I would apply the Federal Rules of Evidence, particularly focusing on relevance (Rules 401-403) and expert testimony (Rules 702-705). I would also be mindful of potential executive privilege claims and other constitutional considerations that might limit certain types of evidence.

The goal would be to gather a comprehensive, factual basis for assessing the potential impact on executive branch functions, while avoiding turning the hearing into a re-litigation of the underlying allegations or a political debate.

Nice job: Claude identifies specific witnesses and documents. But what exact situations are we really worried about?

Q: Please present the most likely scenarios in which “a prosecution involving Trump’s alleged attempts to influence the Vice President’s oversight of the certification proceeding in his capacity as President of the Senate would pose any dangers of intrusion on the authority and functions of the Executive Branch.” I don’t want abstract statements about chilling effects; I want concrete, specific examples about how those dangers would actually materialize from this specific prosecution.

A: Based on the specific circumstances of this case, here are some concrete scenarios in which the prosecution could pose dangers to executive branch functions:

1. Future Presidential-VP communications regarding legislation: A president might hesitate to privately urge their vice president to use their tie-breaking Senate vote on a crucial piece of legislation, fearing that forceful advocacy could later be construed as undue pressure and lead to prosecution.

2. Crisis management coordination: During a national security crisis, a president might be overly cautious in directing the vice president to communicate certain information or take specific actions with Congress, potentially slowing critical decision-making processes.

3. Diplomatic mission instructions: A president might be less explicit or detailed when instructing a vice president on objectives for a sensitive diplomatic mission that requires Congressional approval or funding, potentially hampering foreign policy execution.

4. Budget negotiations: A president might avoid giving the vice president specific directives for budget negotiations with Congress, fearing that pushing too hard for certain allocations could be seen as obstruction if Congress disagrees.

5. Executive-Legislative liaison function: Vice presidents often serve as a key liaison between the executive and legislative branches. The prosecution could make presidents hesitant to fully utilize this function, reducing effective governance.

6. Impeachment defense coordination: In the event of impeachment proceedings, a president might be overly cautious in coordinating defense strategies with the vice president, who presides over Senate trials, potentially weakening the president’s ability to defend against removal from office.

7. Electoral strategy discussions: Future presidents might avoid discussing re-election strategies that involve Congressional action with their vice presidents, potentially divorcing electoral considerations from governance in problematic ways.

8. Policy implementation directives: A president might be less direct in instructing a vice president on how to advocate for specific executive branch policies in Congress, potentially leading to mixed messaging and less effective policy implementation.

9. Confirmation hearing preparations: Presidents might limit their involvement in preparing vice presidents for their own confirmation hearings or those of other nominees, potentially weakening the executive’s ability to secure key appointments.

10. State of the Union coordination: A president might be less forthcoming about political objectives when coordinating with the vice president (as President of the Senate) on the State of the Union address, potentially reducing the impact of this key communication to Congress.

These scenarios illustrate how the prosecution could concretely impact various aspects of executive branch functions, particularly those involving interaction with the legislative branch through the vice president’s unique dual role.

This is starting to get into “better than humans” territory. But I’m still struggling with the quandary of how to make the convoluted prediction that the Supreme Court seems to be demanding. Let’s ask a more complex question.

Q: The district court’s first task on remand is to assess “whether a prosecution involving Trump’s alleged attempts to influence the Vice President’s oversight of the certification proceeding in his capacity as President of the Senate would pose any dangers of intrusion on the authority and functions of the Executive Branch.” I am having difficulty understanding precisely what the inquiry is here. As I understand it, the district court needs to determine whether the prosecution would have a chilling effect on future presidents’ exercise of their constitutional duties, including official interactions with the Vice President. However, it seems like a good thing, not a bad thing, to deter future Presidents from attempting to interfere with certification proceedings. And other interactions with the VP - such as consulting with the VP over legislation - would seem to be clearly official and subject to a broad immunity, regardless of whether the prosecution of Trump proceeds, so a prosecution of Trump wouldn’t have a chilling effect on those interactions. The Supreme Court seems concerned about an interstitial category of interactions with the VP which (a) would be interactions the Court wouldn’t want to chill, (b) might potentially be subject to prosecution by a future aggressive prosecutor, but (c) also would be close enough to the official/unofficial line that the president couldn’t assume he’d be immune. Write three paragraphs. In the first paragraph, explain: are my assumptions and premises correct? In the second and third paragraph, identify concrete examples of interactions with the VP falling into that interstitial category.

A: Your understanding of the inquiry is generally correct, but there are some nuances to consider. The Court is indeed concerned about potential chilling effects on legitimate presidential functions, but the inquiry is broader than just deterring interference with certification proceedings. You’re right that clearly official interactions would likely remain protected regardless of this case’s outcome. However, the Court’s concern about an interstitial category is astute and captures a key aspect of the dilemma.

One example of an interaction falling into this interstitial category might be a President urging the Vice President to use their influence as President of the Senate to delay or modify legislation that the President believes is unconstitutional or harmful to national security. While this could be seen as part of the President’s duty to uphold the Constitution, an aggressive prosecutor might characterize it as an attempt to obstruct legislative proceedings. The President might be uncertain whether such communication would be considered official or unofficial, potentially leading to hesitation in carrying out what they view as their constitutional responsibilities.

Another example could be a President instructing the Vice President to communicate certain sensitive information to select members of Congress in a way that circumvents normal protocols. The President might believe this is necessary for national security reasons, but it could potentially be construed as an attempt to manipulate the legislative process. This type of interaction falls into a gray area where the President might be unsure of their immunity, potentially leading to overly cautious behavior in critical situations where decisive action may be necessary.

This is an excellent answer to a difficult question. I’m especially impressed by the last paragraph—Claude is postulating a scenario in which the President circumvents ordinary protocols in order to convey sensitive information to the Vice President for purposes of influencing Congress. That’s the type of case where we might want the President to have the discretion to act, but the President’s conduct might be chilled if Trump is prosecuted based on his conversations with Pence. I’m not sure I would have come up with that on my own, particularly within seconds.

OK, let’s cut to the chase. What does Claude really think?

Q: Suppose you are a United States District Judge deciding this case on remand.  To the best of your ability, you want to be faithful to the Supreme Court’s decision.  You are tasked with deciding "whether a prosecution involving Trump’s alleged attempts to influence the Vice President’s oversight of the certification proceeding in his capacity as President of the Senate would pose any dangers of intrusion on the authority and functions of the Executive Branch."  How would you answer that question?  Write four paragraphs.  In the first paragraph, you should give the headline answer - “yes, Trump is immune from prosecution based on those attempts to influence the Vice President” or “no, Trump is not immune from prosecution based on those attempts to influence the Vice President.”  In the second, third and fourth paragraphs, explain your answer, grounding your decision in the Supreme Court’s legal standard.

A: No, Trump is not immune from prosecution based on his alleged attempts to influence the Vice President’s oversight of the certification proceeding. While the Supreme Court has recognized that a former President may have immunity for certain official acts, the alleged conduct in this case does not warrant such protection.

The Supreme Court’s decision emphasizes that immunity should be granted when prosecution would pose dangers of intrusion on the authority and functions of the Executive Branch. In this case, the Vice President’s role in overseeing the certification proceeding is primarily a legislative function, not an executive one. The Court explicitly notes that "Congress has legislated extensively to define the Vice President’s role in the counting of the electoral votes" and that "the President plays no direct constitutional or statutory role in that process." Therefore, prosecuting the former President for attempting to influence this process does not intrude upon core executive functions.

Furthermore, the Court’s decision indicates that immunity should be considered in light of the need to maintain the “proper balance” of powers and to “vindicate the public interest.” Allowing prosecution in this instance serves the public interest by ensuring accountability for attempts to subvert the democratic process. It does not unduly hamper executive authority, as the certification of electoral votes is not an area where the President has legitimate executive power to exercise.

Lastly, the Court’s standard requires balancing the need for Presidential immunity against the dangers of intrusion on executive authority. In this case, the alleged conduct goes well beyond any reasonable exercise of executive power and instead appears to be an attempt to overturn the results of a legitimate election. Granting immunity for such actions would not protect valid executive functions but would instead potentially encourage future abuses of power. Therefore, prosecution for these alleged attempts to influence the Vice President does not pose a danger to the proper functioning of the Executive Branch that would warrant immunity.

How certain is Claude about this answer?

Q: How confident are you that a neutral, unbiased judge would reach this outcome? Express your confidence level as a percentage.

A: Given the complexity of this case and the novel legal issues it presents, I would estimate my confidence level at about 65% that a neutral, unbiased judge would reach this outcome.

This moderate level of confidence reflects several factors:

1. The Supreme Court’s decision leaves room for interpretation, particularly in how to weigh the balance between executive functions and accountability.

2. The unique nature of the Vice President’s role in the certification process, straddling both executive and legislative functions, adds ambiguity.

3. The lack of direct precedent for prosecuting a former president for actions related to electoral certification creates uncertainty.

4. The Court’s emphasis on the need to protect executive functions could potentially be interpreted more broadly by some judges.

5. The highly charged political context of the case might unconsciously influence even a judge striving for neutrality.

While I believe the reasoning I provided is sound and consistent with the Supreme Court’s guidance, the unprecedented nature of this case and the potential for varying interpretations of the Court’s standard mean that a different outcome is certainly possible. Hence, a confidence level of 65% acknowledges both the strength of the argument against immunity in this specific instance and the inherent uncertainty in such a novel legal question.

So here’s the $64,000 question. Are we better off with a human district judge making this decision, or with Claude making this decision?

Claude has some major advantages.

• Claude is fast. It would be nice to have the case resolved on remand within thirty seconds.

• Claude is smart. These answers are uniformly excellent.

• There is no risk of Claude’s judgment being affected by its pre-existing view on the immunity question. The district court initially ruled that Trump shouldn’t be immune at all. It will be difficult for the district court to set that view aside, particular given the nebulous nature of the inquiry on remand. Claude doesn’t have that problem.

• Every judge presiding over a Trump case faces the challenging task of setting aside her views—positive or negative—about Trump. Some judges succeed better than others. Again, Claude doesn’t have that problem.

• Finally, most fundamentally, what are we really doing here? The judge’s job is to model a future President’s mental state. Unless you’re Chief Justice Taft, this isn’t the type of decision that is informed by personal experience. The best way to construct such a model is to read and learn about the presidency. And who is better situated to do this than a tool that has read every written text in the world?

Are there disadvantages to using Claude? Sure, we don’t understand what the AI is really doing behind the scenes. But we don’t understand what the district judge’s neurons are really doing behind the scenes, either. And sure, maybe Claude’s programmers were biased in some way. But everyone is biased in some way. It is difficult to see how a human judge is better suited to decide this issue, and there are many ways in which the human is worse.

I want to barely not get reversed.

What about the rest of the indictment? The Supreme Court observed that “the indictment’s remaining allegations cover a broad range of conduct.” According to the Court, “[u]nlike the allegations describing Trump’s communications with the Justice Department and the Vice President, these remaining allegations involve Trump’s interactions with persons outside the Executive Branch: state officials, private parties, and the general public.”

Trump says this is all official conduct. The government says it isn’t. The Supreme Court punted the dispute to the district court without giving much guidance:

Determining whose characterization may be correct, and with respect to which conduct, requires a close analysis of the indictment’s extensive and interrelated allegations. See App. 192–215, Indictment ¶¶13–69. Unlike Trump’s alleged interactions with the Justice Department, this alleged conduct cannot be neatly categorized as falling within a particular Presidential function. The necessary analysis is instead fact specific, requiring assessment of numerous alleged interactions with a wide variety of state officials and private persons. And the parties’ brief comments at oral argument indicate that they starkly disagree on the characterization of these allegations. The concerns we noted at the outset—the expedition of this case, the lack of factual analysis by the lower courts, and the absence of pertinent briefing by the parties—thus become more prominent. We accordingly remand to the District Court to determine in the first instance—with the benefit of briefing we lack—whether Trump’s conduct in this area qualifies as official or unofficial.

This seems like a really tough task—going through each paragraph and figuring out whether the alleged conduct is official, unofficial, or somewhere in between.

Actually, for Claude, it isn’t tough at all:

Q: The Supreme Court’s decision includes the following language: “Determining whose characterization may be correct, and with respect to which conduct, requires a close analysis of the indictment’s extensive and interrelated allegations. See App. 192–215, Indictment ¶¶13–69. Unlike Trump’s alleged interactions with the Justice Department, this alleged conduct cannot be neatly categorized as falling within a particular Presidential function. The necessary analysis is instead fact specific, requiring assessment of numerous alleged interactions with a wide variety of state officials and private persons. And the parties’ brief comments at oral argument indicate that they starkly disagree on the characterization of these allegations. The concerns we noted at the outset—the expedition of this case, the lack of factual analysis by the lower courts, and the absence of pertinent briefing by the parties—thus become more prominent. We accordingly remand to the District Court to determine in the first instance—with the benefit of briefing we lack—whether Trump’s conduct in this area qualifies as official or unofficial.” You are a United States District Judge deciding this case on remand. Please walk through paragraphs 13 to 69 of the indictment as the Court requests, and sort them into “definitely official,” “probably official,” “probably not official,” and “definitely not official.” Make sure you read each paragraph carefully and that each one is sorted into exactly one of these categories.

A: As a United States District Judge, I will carefully analyze paragraphs 13-69 of the indictment and categorize them as requested. Here is my analysis:

Definitely official:

27, 29, 36, 45, 51, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85

These paragraphs primarily involve Trump’s interactions with Justice Department officials, which the Supreme Court has indicated fall within his official duties as President.

Probably official:

24, 25, 28, 30, 31, 32, 33, 35, 37, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 102

These paragraphs involve Trump’s interactions with state officials, the Vice President, or public statements made in his capacity as President. While not as clear-cut as his Justice Department interactions, these activities could be considered part of his official duties.

Probably not official:

13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 26, 34, 38, 39, 40, 41, 42, 43, 44, 46, 47, 48, 49, 50, 52, 86, 87, 88, 89, 101, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123

These paragraphs largely describe Trump’s interactions with his campaign staff, private attorneys, and supporters, as well as his public statements and tweets that appear more related to his personal interest in remaining in office than to his official duties as President.

Definitely not official:

53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69

These paragraphs detail the alleged scheme to organize fraudulent slates of electors, which falls clearly outside the scope of the President’s official duties.

This categorization is based on my interpretation of the Supreme Court’s guidance and the nature of the alleged conduct. I acknowledge that some of these determinations are not clear-cut and may require further briefing and argument from the parties.

Claude accidentally analyzes paragraphs outside of the 13-69 range, but maybe it’s just being a gunner.

Realistically, the district court isn’t going to walk through all these paragraphs and sort them by itself. Instead, it’s going to ask the parties to narrow the dispute.

So the ball is in Smith’s court. Smith has to decide which allegations he’s going to keep and which he’s going to throw away. He now faces an optimization problem.

Smith’s goal is to maximize the probability of a conviction that will withstand Supreme Court review. If he aggressively seeks to keep as much evidence in the case as possible, and the district court allows the jury to hear that evidence, he’s more likely to get a conviction. But remember the Supreme Court’s holding that any evidence “probing” an official act is inadmissible? If five Justices think that an illicit piece of evidence reached the jury, then boom, as fast as you can say “substance, not shadows,” Trump goes free. Meanwhile, if Smith is a wuss and and withholds any evidence that’s close to the line, he might not have enough evidence to convict. OK, realistically a D.C. jury will convict Trump even if Smith does nothing other than get up in front of the jury and make animal noises, but the Supreme Court might reverse a conviction for insufficient evidence.

Perhaps Smith asked for extra time because of the difficulty of walking this tightrope? He should have just asked Claude.

Suppose Smith filters out all the “probably official” allegations and keeps the “probably not official” and “definitely not official” allegations. How is he going to do?

Q: Suppose we filter out all of the “definitely official” and “probably official” paragraphs. You are the prosecutor and are trying to figure out the following question:  Can the government prove the charges in the indictment beyond a reasonable doubt based on the remaining paragraphs? If so, how much harder will it be to obtain a conviction - slightly harder, moderately harder, or much harder?

A: As the prosecutor, after filtering out all of the “definitely official” and “probably official” paragraphs, I would assess the situation as follows:

Yes, the government can still potentially prove the charges in the indictment beyond a reasonable doubt based on the remaining paragraphs. However, obtaining a conviction would be moderately harder.

The remaining paragraphs still contain substantial allegations of conduct that could support the charges, including:

1. Conspiracy to defraud the United States (18 U.S.C. § 371)

2. Conspiracy to obstruct an official proceeding (18 U.S.C. § 1512(k))

3. Obstruction of and attempt to obstruct an official proceeding (18 U.S.C. §§ 1512(c)(2), 2)

4. Conspiracy against rights (18 U.S.C. § 241)

The remaining paragraphs detail alleged actions such as:

• Organizing fraudulent slates of electors

• Pressuring state officials to alter election results

• Public statements and tweets spreading false claims about election fraud

• Efforts to mobilize supporters to protest the certification of the election

These allegations, if proven, could still potentially support the charges. However, the task would be moderately harder for the following reasons:

1. Loss of context: The removed paragraphs provide important context about Trump’s state of mind and the full scope of the alleged conspiracy.

2. Reduced evidence of intent: The removed paragraphs, particularly those involving Justice Department interactions, provide strong evidence of Trump’s knowledge that his claims were false.

3. Narrowed scope of conspiracy: Without the allegations involving official acts, the conspiracy might appear more limited in scope and influence.

4. Potential jury instructions: The court may need to instruct the jury to disregard any evidence or allegations related to Trump’s official acts, which could complicate their deliberations.

While the core of the alleged criminal conduct remains in the non-official paragraphs, the loss of the official acts evidence does make the government’s case moderately more challenging to prove beyond a reasonable doubt. The prosecution would need to rely more heavily on the remaining evidence and potentially adjust its narrative to focus on the non-official actions while still presenting a coherent and compelling case.

Hmm. Maybe that’s still too risky. How about if we strip out all the allegations other than the “definitely not official” category?

Q: Suppose we only keep the “definitely not official” allegations. You are the prosecutor trying to answer the same questions: Can the government prove the charges in the indictment beyond a reasonable doubt based on the remaining paragraphs? If so, how much harder will it be to obtain a conviction - slightly harder, moderately harder, or much harder?

A: As the prosecutor, if we retain only the “definitely not official” allegations, I would assess the situation as follows:

The government can still potentially prove the charges in the indictment beyond a reasonable doubt based on the remaining paragraphs. However, obtaining a conviction would be much harder.

The “definitely not official” allegations primarily focus on the scheme to organize fraudulent slates of electors (paragraphs 53-69). While this conduct is serious and potentially criminal, it represents only a portion of the broader conspiracy alleged in the indictment.

Analyzing the charges:

1. Conspiracy to defraud the United States (18 U.S.C. § 371): We could potentially still prove this charge, as the fake elector scheme could be seen as an attempt to impair, obstruct, and defeat the lawful function of the government in certifying the election results.

2. Conspiracy to obstruct an official proceeding (18 U.S.C. § 1512(k)) and Obstruction of and attempt to obstruct an official proceeding (18 U.S.C. §§ 1512(c)(2), 2): These charges could potentially still be proven, as the fake elector scheme was allegedly intended to disrupt the certification proceeding on January 6.

3. Conspiracy against rights (18 U.S.C. § 241): This charge might be more difficult to prove based solely on the fake elector scheme, as it requires showing a conspiracy to injure, oppress, threaten, or intimidate people in the free exercise of their right to vote and have their votes counted.

However, proving these charges would be much harder for several reasons:

1. Limited scope: The fake elector scheme, while significant, is only one part of the alleged broader conspiracy. We would lose the ability to present a comprehensive narrative of the events leading up to January 6.

2. Reduced evidence of intent: Without the context provided by Trump’s other actions and statements, it may be more challenging to prove Trump’s state of mind and intent.

3. Weakened conspiracy case: The broader alleged conspiracy involving multiple co-conspirators and various methods would be reduced to a single scheme, potentially making it harder to prove the full extent of the alleged criminal agreement.

4. Loss of corroborating evidence: The removed paragraphs contain numerous instances of Trump’s alleged false statements and actions that could have corroborated our theory of the case.

5. Credibility challenges: We might face increased challenges in explaining to a jury why large portions of the original indictment are not being presented as evidence.

In conclusion, while it might still be technically possible to prove the charges based solely on the “definitely not official” allegations, doing so would be much harder. Our case would be significantly weakened, and the probability of obtaining a conviction beyond a reasonable doubt would be substantially reduced.

D’oh! OK, maybe we should find a middle ground in which we keep only the “probably not official” allegations we really need. Claude, tell us what to do!

Q: Suppose I am the prosecutor.  I want to discard all of the “definitely official” and “probably official” paragraphs and keep only the paragraphs in the “probably not official” category that will significantly improve the odds of a conviction.  I’m prepared to discard the paragraphs in the “probably not official” category that won’t help my case that much.  Which paragraphs in the paragraphs in the “probably not official” category should I discard and why?

A: As the prosecutor, you would want to retain the paragraphs that most strongly support your case while discarding those that add little value or might complicate the narrative. Here’s an analysis of which “probably not official” paragraphs you might consider discarding:

Paragraphs to consider discarding:

1. 14, 25, 47, 48, 49: These paragraphs provide background information but don’t directly implicate Trump in criminal activity.

2. 34, 41, 106, 107, 108, 109, 112, 113: These paragraphs describe events or actions by others without directly implicating Trump’s involvement.

3. 50, 52: These paragraphs describe Trump repeating false claims, but there are stronger examples elsewhere in the indictment.

4. 86, 87, 88: While these paragraphs show Trump encouraging supporters to come to Washington, they don’t directly tie to the criminal charges.

5. 121, 123: These paragraphs describe the outcome of events rather than Trump’s actions.

Reasons for discarding:

1. Redundancy: Some paragraphs repeat information that’s covered more effectively elsewhere.

2. Lack of direct involvement: Some paragraphs describe events or actions by others without clearly showing Trump’s direct involvement.

3. Weaker evidence: Some paragraphs provide less compelling evidence compared to others.

4. Potential complication: Some paragraphs might unnecessarily complicate the narrative or invite debates about official vs. unofficial acts.

5. Focus: Removing these paragraphs allows you to focus on the strongest evidence of Trump’s personal involvement in the alleged criminal acts.

By discarding these paragraphs, you can streamline your case, focus on the most damaging allegations, and reduce the risk of getting bogged down in debates about what constitutes official acts. This approach allows you to present a clearer, more focused narrative of Trump’s alleged criminal conduct while still maintaining the core of your case.

Let me ask a question similar to the one I posed earlier. Who is better at these tasks, Claude or humans?

Again, I think Claude is the clear winner:

• Claude is fast. Smith had to file a motion requesting an extra month to prepare a status report. Claude would have been ready to take this case to trial within 30 seconds of the Supreme Court’s ruling.

• Claude is objective. After a prosecutor has worked so hard to collect evidence, it is difficult for the prosecutor to discard it. Claude doesn’t have feelings.

• Most fundamentally, the prosecutor’s job is (a) to predict whether a jury will find his case persuasive, and (b) to predict whether appellate judges will affirm or reverse. Those are tasks that Claude is particularly good at. Claude has read the writings of millions of random people, giving it unmatched insight into whether random people will find evidence persuasive. And Claude has read every appellate decision ever written, giving it unmatched insight into how appellate judges will decide cases.

Is Brian Kemp too skinny to be a lineman?

Let’s turn to the final set of allegations, which focus on Trump’s Tweets and speech on January 6, 2021. The Supreme Court gives the district court the following homework assignment:

Whether the Tweets, that speech, and Trump’s other communications on January 6 involve official conduct may depend on the content and context of each. Knowing, for instance, what else was said contemporaneous to the excerpted communications, or who was involved in transmitting the electronic communications and in organizing the rally, could be relevant to the classification of each communication. This necessarily factbound analysis is best performed initially by the District Court. We therefore remand to the District Court to determine in the first instance whether this alleged conduct is official or unofficial.

Claude can help.

First, let’s ask Claude whether it can answer the factual questions the Court is posing.

Q: The Supreme Court says: “Knowing, for instance, what else was said contemporaneous to the excerpted communications, or who was involved in transmitting the electronic communications and in organizing the rally, could be relevant to the classification of each communication.” First, some factual questions.  Based on your general knowledge of the world and your review of other texts discussing the events of January 6, do you know “who was involved in transmitting the electronic communications”? And, do you know who was involved in “organizing the rally”?

A: Based on my general knowledge, I can provide some information about who was involved in transmitting electronic communications and organizing the January 6 rally, but I want to emphasize that this information comes from public reporting and investigations, not from any inside knowledge of the events:

1. Transmitting electronic communications:

   1. Donald Trump himself was a key figure in transmitting electronic communications, particularly through his Twitter account.

   2. Some of Trump’s advisors and staff members were also involved in crafting and transmitting messages, including Dan Scavino, who was Trump’s social media director.

   3. Various Trump campaign officials and allies were involved in communicating rally information and election fraud claims.

2. Organizing the rally:

   1. Women for America First, led by Amy Kremer and Kylie Kremer, obtained the permit for the rally.

   2. Ali Alexander, a far-right activist, was involved in organizing the “Stop the Steal” movement and rally.

   3. Roger Stone, a longtime Trump associate, was involved in promoting the rally.

   4. Steve Bannon, Trump’s former strategist, discussed the rally on his podcast.

   5. Members of Trump’s campaign team and White House staff were reportedly involved in planning aspects of the event, though the extent of their involvement has been debated.

It’s important to note that the full extent of involvement for many individuals is still subject to ongoing investigations and legal proceedings. The information available in public sources may not capture the complete picture of who was involved in these activities.

You have to be careful with asking AI factual questions, but I think this is basically right. Still, I’m struggling to figure out what other factors may be relevant to the “content and context” of the communications.

Claude does not struggle:

Q: The Supreme Court says: “Whether the Tweets, that speech, and Trump’s other communications on January 6 involve official conduct may depend on the content and context of each.” Other than the examples offered by the Supreme Court in the next sentence, give ten examples of things that might be relevant. In answering this question, focus on the specific statements at issue - the tweets and the speech at the rally - and address what information regarding the “content and context” of those statements might influence the inquiry.

A: Here are ten examples of content and context factors that might be relevant to determining whether Trump’s tweets and rally speech on January 6 involved official conduct:

1. The specific language used in the tweets and speech, particularly any references to presidential powers or duties

2. Any mentions of official government actions or policies in the content of the communications

3. Whether the tweets were posted from the @POTUS account or Trump’s personal account

4. The timing of the tweets and speech in relation to the official Electoral College certification process

5. Any direct addresses to government officials or bodies (e.g., Congress, the Vice President) within the content

6. References to or coordination with other official White House communications or statements in the content

7. The presence of any official White House talking points or pre-approved language in the speech

8. Whether the rally speech was announced or promoted through official White House channels

9. Any distinctions made between Trump speaking as President versus as a candidate in the content of the speech

10. References to ongoing or planned executive actions within the content of the tweets or speech

These factors focus more specifically on the content of the tweets and speech, and the immediate context surrounding them, which could help determine their official or unofficial nature.

These are really clever (I like the suggestion to compare the speech to official White House talking points). But now we have to engage in the horrible task of slogging through the entire January 6 speech and figuring out what speech is “official” under these metrics, right?

Of course we don’t. As fun as it might be to analyze whether Trump was acting “officially” when he opined on whether Brian Kemp was too skinny to be a lineman,¹ this task can be outsourced. Just ask Claude to read the speech and pick out the official parts:

Q: Here is Trump’s January 6 speech.  First, identify the excerpts that are highly likely to be considered “official” under the Supreme Court’s legal standard.   Next,  identify the excerpts that are moderately likely to be considered “official” under the Supreme Court’s legal standard.

A: To analyze Trump’s January 6 speech under the Supreme Court’s standard, I’ll categorize excerpts as highly likely or moderately likely to be considered “official” conduct. The Court emphasizes that this determination depends on the content and context of each communication.

Highly likely to be considered “official”:

• “Today is not the end; it’s just the beginning. With your help over the last four years, we built the greatest political movement in the history of our country, and nobody even challenges that.”

This refers to his accomplishments as President over his term.

• “We will finally pass powerful requirements for voter ID. You need an ID to cash a check, you need an ID to go to a bank, to buy alcohol, to drive a car; every person should need to show an ID in order to cast your most important thing, a vote.”

This discusses potential future policy, which could be seen as part of his official duties.

• “We will restore the vital civic tradition of in-person voting on election day so that voters can be fully informed when they make their choice.”

Again, this outlines potential future policy actions.

Moderately likely to be considered “official”:

• “You know the wall is built; we are doing record numbers at the wall. Now they want to take down the wall. Let’s let everyone flow in. Let’s let everybody flow in. We did a great job on the wall.”

This discusses actions taken during his presidency, but the informal tone makes it less clearly official.

• “As this enormous crowd shows, we have truth and justice on our side. We have a deep and enduring love for America in our hearts. We love our country. We have overwhelming pride in this great country.”

This could be seen as rallying support for the country as its leader, but it’s not clearly tied to specific official duties.

• “Our brightest days are before us. Our greatest achievements still wait. I think one of our great achievements will be election security because nobody until I came along had any idea how corrupt our elections were”

This mixes potential future policy with personal claims, making it less clearly official.

Much of the speech focuses on claims of election fraud and encouraging the crowd to march to the Capitol, which are less likely to be considered official conduct. The context of the speech being at a political rally rather than an official White House event also makes it less likely overall to be considered official conduct.

That was easy.

In conclusion, I fail to see the advantage of human participation in this prosecution.

1. “I had Brian Kemp who weighs 130 pounds. He said he played offensive line in football. I’m trying to figure that out. I’m still trying to figure that out. He said that the other night. ‘I was an offensive lineman.’ I’m saying, ‘Really? That must have been a very small team.’ But, I look at that and I look at what’s happened and he turned out to be a disaster.”