Around seven months ago I wrote the post "a non-magical explanation of Jeffrey Epstein". In it, I make the argument that Jeffrey Epstein did actually kill himself and that he was a CIA agent. I reach that conclusion mostly by drawing on some Conspiracy Theory and doing some light research. 

I am mostly satisfied with my reasoning, and in particular my conclusion that Epstein was a high level informant for American intelligence services. However, over time I became less satisfied with my conclusion about Epstein's death. The paradox of the Epstein case is that the circumstances of his death seem so organized, and yet succeeding investigative details and a nuanced understanding of the parties involved seem to completely rule out foul play. It's quite literally a locked room murder story.

Problems with the murder hypothesis

Impracticality, lack of forensic evidence, lack of witnesses

Multiple reviewers of footage from that night, including the Attorney General at the time, claimed related cameras around Epstein's cell showed no one exiting or entering the block he was in. I am satisfied in the assumption that not all of the parties who verified this made a mistake or are secretly controlled by some puppet-master behind the scenes. This means Epstein would have to be killed by someone in his housing block who escaped their cell during the night, broke into Epstein's cell without any witnesses, and then strangled him to death without leaving any obvious and unmistakable evidence of a struggle for the coroner, and this seems totally implausible. Epstein didn't yell as his neighboring inmate sat there picking the lock? Get blood and skin under his fingernails? Do literally anything while this degenerate ninja-inmate tied his bedsheets around his neck? None of the other inmates thought to mention this if that's what happened, even after it was clear there was a coverup and they might be able to swing immunity out of it?

Epstein's will & foreknowledge of his death

Epstein clearly knew when he was going to die, because he drafted and signed a will just two days before he did. You can make up murder stories where he knows this, but Epstein seems like the type of person to cause maximum damage to his killers by explicitly telling others who he believed was going to off him and why.

Lack of viable suspects

I don't think Epstein's sex abuse victims have accused anyone with either the motive or means to organize a high-security jailhouse murder, regardless of the particular circumstances. This pretty much leaves the CIA, or some random unknown third party that wanted him dead for something maybe unrelated to raping children.

• I say motive because even if everything accused about any single one of the people fingered in the Epstein scandal were proven in a court of law, they still wouldn't be facing life imprisonment. It's hard to imagine how someone sane could see killing Epstein as a rational solution to their relatively minor legal problems, without the benefit of hindsight. Why would you risk getting life without parole for assassinating a state's witness so you could avoid five years in prison after a plea bargain for statuatory rape?
• I say means because organizing the murder of potential witnesses is the type of thing that only makes sense if you already have trusted confederates you can task. If you are a major drug dealer, and you have five direct underlings, and one of them goes to jail and it looks like they're going to testify against you, you can task one of the remaining four people with organizing a hit because you're already trusting that person in the course of your job and doing so doesn't mean assuming much additional risk. But if you don't have those pre-existing relationships then trying to get someone, or a group of people, to cooperate in killing a potential witness is, at best, replacing one potential witness with another potential witness. None of the Known Quantities (aside from hypothetically the CIA I guess) seem to be in possession of existing contacts useful in ameliorating this dilemma. 
  • (Not to mention the fact that if you haven't already arranged murders before, you might think "damn, maybe I'll fuck this up and dig myself into an even deeper hole than I already have.")
  • And my (admittedly limited) vision into how the CIA is structured and its officers are incentivized says that the people inside the CIA who would have any motive to murder Epstein would not be able to, in practice, use workplace resources to illegally assassinate an American citizen. 
    • If I am very very very wrong and the CIA really has that little internal oversight over the Special Activities Division kill button and our intelligence agencies are that dirty, then the people involved probably have no reason to risk escalating attention by murdering Epstein because in most of these alternate universes they don't face repercussions for cultivating him as an agent. You probably either live in a world where your intelligence agencies have impunity or a world where they have to organize assassinations of key figures; it makes less sense to me that a CIA would be both capable of assassinating Epstein AND scared of him testifying in court.

No in-progress plea bargain at Epstein's time of death 

If Epstein still had incriminating information on other people he thought he could leverage for a better sentence, he would have been using it, and he would have been completely safe. 

• This would have been a major part of the story if it had happened. A bunch of people would have known it was happening, including his lawyers, who have failed thus far to mention such a plea or which people, if any, he was set to testify against.
• Epstein was a wall street executive with zero sense of criminal loyalty or honor. He would absolutely have been ratting on co-conspirators if he thought it was something he could do to get out of hard time, and would have started negotiating such a plea well before he died instead of sitting in jail crying about his living conditions. He did this exact thing as a pre-emptive strike on a finance partner who helped him steal buckets of money in the 90s, before he was even indicted, just to make sure the other guy didn't get the opportunity to do it to him first. 
• Epstein would have faced little to no threat of direct reprisal for such testimony. WITSEC has a literal 100% success rate at protecting informants from the worst criminal organizations in the world. Movies and television shows have lied to you about this, because a 100% success rate WITSEC makes for a very uncompelling story, and is anti-fun and anti-edgy to the point that people resist believing it.
  • It turns out that criminals, just like everyone else, have a coordination problem in punishing defectors. Even perennial institutions like the American Mafia with specific public reputations to uphold are typically lax about assassinating oathbreakers when they didn't hurt the specific regime of the people currently in charge. Criminals want informants to die in general, they don't have the motivation to personally go through the trouble and risk of enforcing underworld justice themselves, unless those people are actively putting guys from their organization in particular in jail. Murdering people is very expensive.
  • Please read this or another book on WITSEC before you start to quibble with me about the 100% success rate. Yes, there have been, in extremely rare circumstances, people who viciously assault civilians or start to move large amounts of drugs again after getting into the program, and then get kicked out of WITSEC after that as per their agreement, and then go back to their old neighborhoods against the advice of sane people and get hurt. If you want to use this to say Epstein would have been in danger, then go ahead and select a specific actual reference example of this happening so that we can debate whether or not that "exception" was the result of covert actions taken by people outside the program who wanted them dead, or that Epstein would have been forced to make the same mistakes. I say "100% success rate" because WITSEC has been 100% successful at doing the thing Epstein would need it to do, and the thing a reasonable person would want it to do, not because it's able to magically sedate every pathologically criminal psychopath with a death wish.

Low likelihood of FBI Corruption

Finally, as people pointed out in the comments of the post, I do in fact have a strong prior that the parties that investigated Epstein's death, namely the FBI, are and were generally habitually honest in the relevant sense. Explaining the whole story behind why I believe this, even in spite of the circumstances, would take several posts to convey sufficiently for those that don't want to hear it, but the cliff-notes are:

• There have been exceptionally few apparent cases of corrupt FBI officers in the last fifty years. And if it were possible to corrupt the FBI and had happened, I do in fact expect there would be more apparent cases. Substantiating that second sentence is one of the things that would take a lot of prep work.
  • Those corrupt agents, predictably, had a limited ability to actually tamper with evidence. Tampering with evidence is really dangerous! They were instead suspected of giving information to criminals about things like potential informants or arrests, which is easy enough to do in private if the person you're giving information to is your CI.
  • In the two to four serious possible cases of organized crime or individual criminals corrupting FBI agents I'm aware of, it was contained to specific agents. There was not a culture of dishonesty among a task force that might indicate a broader problem, and the rest of the people on the corrupt agents' team were generally diligent in raising advance concerns.
    • I have reasons to believe that the "teammates noticing and raising the alarm" part is not a selection effect and is in fact unusual for cases of corruption in law enforcement. Contrast this with the typical case of corruption in American local law enforcement where bad behavior often turns out to be endemic or systematically overlooked by the rest of the department.
      • A reviewer remarks that the reaction to this evidence could be "Look at other law enforcement bodies! They're so often corrupt - why not just believe the FBI has the resources to hide it really well?" My point is that the examples we have of FBI corruption seem systematically dissimilar to the examples we have of widespread corruption in other departments, in a way that implies they are edge cases. It's true that specific instances of local police departments being systematically, criminally corrupt ought to increase your prior that the FBI is compromised too. If those instances seemed so emblematic I were starting with a prior of near certainty, I might start to reason that the FBI were just extraordinarily good at coverups. But I don't actually think such a prior is justified, and while the FBI does have, say, a larger budget than local departments, those resources are earmarked for law enforcement and are not obviously useful for covering up their own corrupt behavior.
• FBI agents basically lack extrinsic motivation to become de novo corrupt, mostly by design.
  • FBI agents are already paid fairly well, in status as well as money. They make solidly above median income and get a position inside America's most elite crime fighting agency, and these positions are competed for and treasured very intensely even if you personally wouldn't value one. A substantial proportion of women find physically fit elite lawmen attractive. They have sex.
  • The benefits of accepting bribes in a first world country, as a police officer or a district attorney, are really quite small. While I respect that others may have different utility functions, it seems to me like only extraordinarily stupid or self-destructive people inside American law enforcement do it, regardless of morals. Even a 5% risk of ruining your career and going to jail for five years is probably not worth tolerating for monthly payments you cannot actually use to buy a house. And an agent accepting money from a typical professional or habitual criminal stands much more than a 5% chance of getting caught.
    • One way you can tell bribes are not much of a motivator is the fact that some of the most cited cases of possible FBI corruption did not even involve bribes! If you believe Lindley Devecchio did aid Greg Scarpa, he apparently did it mostly because it made him feel cool, or because he thought Scarpa's continued ability to work for him as an informant might help him boost prestige among coworkers.
  • The FBI's recruitment process is very good at filtering out people likeliest to disregard their incentives and become de novo corrupt anyways for reasons like pathological risk-taking. This is a primary goal of their screening and they go over and beyond by filtering out many who you would probably say are fine but seem even slightly on paper like they'd be interested in taking such risks.
• Corruption is a self-fulfilling prophecy, but also a self-refuting prophecy, and the prevalence of such corruption scales with common knowledge. The FBI has a general reputation among organized crime of being highly resistant to blackmail and bribes. So gangs and vice networks tend to stick to turning local law enforcement when they decide that's advantageous, and so the FBI becomes seen as even less corrupt, in a virtuous cycle. This phenomenon is part of why there's a gulf between countries that have ubiquitous corruption by patrolmen/traffic cops and countries where very few people pay any bribes on a regular basis.
• The FBI is a huge organization. The cases where FBI agents have been corrupted were cases where there was a specific team investigating particular groups of people, like "Boston area crime bosses", for long periods of time, who could actually influence the progress of those investigations long term. Why would the mysterious third party's random mole, if they had one, be particularly likely to be assigned to the investigation of Epstein's death? Or even be in a position to get themselves assigned to that investigation if that were possible?

Problems with the naive suicide hypothesis

• The death rates inside these prisons is extremely low. How come the first person in this prison that managed to kill themselves in this prison in so many years happened to be as high profile as Epstein? Lots of people in that jail want to die.
• How on earth do these cameras specifically along his cell go out at the same time as Epstein is hanging himself? Even if this happens all the time it's gotta be a hundred, maybe a thousand to one that it happens for those specific cameras on that day.
• How come the guards violated protocol that night by not checking on Epstein every thirty minutes, and falsified records to say they did? If your hypothesis is that the guards were lazy or that this prison's security was lax, why does that hypothesis happen to be the case in this particular prison? Also, why does this particular prison seem to be so otherwise successful at preventing suicides of people in similar situations as Epstein?

The third option

So my opinions of both hypotheses were fairly low. I settled this by saying that the murder hypothesis is stupider than the naive suicide hypothesis, and went the latter. The justification in my head was something like: "The fact that this case was brought to my attention meant that it was a low-probability event anyways, and so it's not really that odd that it's so strange."

I think that was a mistake, possibly a generalizable mistake to learn from. Instead of saying "I am really not sure what happened, but X is more likely than Y", I picked one and forgot or didn't realize there could be alternative explanations. If both of your theories seem like million-to-one events, they're probably both wrong and you're making a mistake in concluding they're the only available options. I have now come up with a third hypothesis, which I guess you could say is technically covered by #2: 

Epstein paid correctional officers to let him die. The cameras weren't working and/or the guards failed to check on him that night, because Epstein explicitly bribed someone(s) to disable those cameras and/or look the other way.  The base rate of suicide in the MCC correctional facility is low because generally its precautions work to prevent suicide attempts. 

Epstein took deliberate action to ensure they didn't. He did so partly to save himself a life in prison, but also as part of a plan to spite his victims. His suicide denied them both the satisfaction of a trial and complicated access to any resources which his will would pour into an inaccessible private trust. That will, by the way, seemed deliberately designed to most observers to minimize the government's ability to access his fortune or get him to pay restitution in the event of his death.

I know this seems like the kind of wild theory you say you believe in an essay because it's thought provoking. This is what I actually believe; it's my 85% hypothesis. It's the only story that makes sense; it explains Epstein's behavior, the incidentally poor security, why there was nobody going in or out of the housing blocks and no evidence of a struggle, and doesn't propose some wide ranging conspiracy across multiple government departments that we would already probably know about. It's the only way he could be expected to kill himself inside that prison in the first place if he did. It also explains why this plot wasn't uncovered: because there are like, three people that would have to know about it, not even including Epstein's lawyers. Epstein could have just made up a sob-story and convinced a correctional officer to look the other way himself, or promised the officer his lawyers would pay them after the fact, and then didn't, because arranging that would be more complicated and time consuming.

And I'll admit, even to the hit of my own credibility, that I love this theory because it's so... Anticlimactic? Left-field? No one wants to believe this story, except pedantic weirdos like me. It's so satisfyingly unsatisfying. You don't figure this out by imbuing some grand intentionality to the story or thematic significance to the people involved. You get it by following all the billiard balls really closely and paying attention to all of the uninteresting details and seeing what incidental property emerges. And once you're done you're usually left without some obvious message because these people aren't living their lives deliberately trying to convey some grand story about corruption. Those lessons are there, but they're not there as part of a curriculum designed to teach them to you.