ποΈ Beggin'-Ass Bankman-Fried, Pandering for a Trump Pardon
The media tour is much sadder this time. Also a huge crypto hack, AI brain rot, and a whole lot more.
If thereβs one uncontroversial thing you can say about Sam Bankman-Fried β¦ well, actually, there are a lot of those, starting with βheβs guilty as fuckββ and βheβs a pathological liar.β You wouldnβt exactly get a lot of dissenting opinions there, unless you were polling Yale Law professors, alleged pedophile cops, or his own parents.
But one really, really obvious thing is that SBF loves to talk. Thatβs how he wound up giving dozens of interviews before his trial - interviews which hilariously kneecapped his own defense team.
Now heβs at it again with a recent interview with the New York Sun, and Twitter posts furiously glazing Elon Muskβs DOGE. Just as he thought he could blab his way out of fraud back in November of 2022, SBF now is pretty obviously trying to talk his way into a pardon from Donald Jonald Trump. Below youβll find a roundup and analysis of highlights from the Sun interview.
Itβs hilariously pathetic, since he again has to work against one of this own previous tactical lies - the campaign finance fraud that, in part, concealed his contributions to Republicans to help him present himself as a feel-good lib do-gooder. This fraud helped make him a high-profile punching bag for right-wing media figures back in 2022, and they largely still see him as aligned with the Clinton Cabal, particularly through Michael Kives and K5. If Trump has thought about Bankman-Fried *at all* over the past few years, this is certainly his own vague, addled impression, making an SBF pardon incredibly unlikely, and Samβs sudden rightward shift even more obvious sociopathic theater from a man with no moral center at all.
Still, you can sense thereβs some weird coordination going on behind the scenes of his new prison interview with the far-right and proudly Zionist New York Sun (whose British owner has also bid to buy the Telegraph). The Sun reporter sets Sam up to talk about his rightward shift, and then to run through his still-absurd defense claims that basically boil down to βI didnβt do it.β
The good news is there is absolutely zero political upside in a pardon for Trump, and thatβs why the President does most of what he does.

When asked what he misses most, Sam doesnβt say itβs his brother, family, or friends (he barely had any), but βbeing able to Google thingsβ and to work. βThe days go by in pretty undifferentiated fashion β¦ Iβm used to being able to dedicate myself to work, and there isnβt much to work on here.β
First, some news.
Lazarus hits Bybit With Largest Crypto Hack Ever
A genuinely terrifying social-engineering attack led to the theft of $1.5 billion, now definitively linked to North Koreaβs Lazarus Group. Itβs the largest crypto hack ever, which also means itβs probably the largest theft in human history.
The attackers tricked multiple exchange officials into signing fraudulent interactions with whatβs known as a βmultisig,β or a crypto wallet requiring multiple signatures. The industry will be grappling for some time, hopefully very seriously, with how to prevent this type of attack in the future.
However. As my great good friend and FTX victimsβ advocate Sunil Kavuri has pointed out, itβs at least a weird coincidence that Bybit was hacked for $1.5 billion, given that Bybit extracted $1.5 billion from the FTX bankruptcy process through rushed pre-bankruptcy withdrawals and a hell of a negotiated settlement. What might that mean? I do not know at all.
LLM Usage Causes βCognitive Atrophyβ
Strong recommend for this survey of recent findings on LLM impacts from Erik Hoel at The Intrinsic Perspective. Hoel runs down results including a recent Microsoft study that found LLM users showed declines in critical thinking skills.
Among other nuances: βThis negative effect scaled with the workerβs trust in AI: the more they blindly trusted AI results, the more outsourcing of critical thinking they suffered.β
Preview: βI ζ AIβ by Aesthetic Calculations
Iβve just been informed about a wildly intriguing, beautifully-designed collection of βessays, fiction, interviews and visual interventions that reintroduce sex, desire and politics into utopian fantasies of artificial intelligence, via psychoanalysis.β If youβve read much of my SBF work, you might have detected that I think psychoanalysis is a very powerful interpretive frame for a lot of financial fraud, and I can only imagine the insights it offers on artificial life. I donβt have it yet, but anticipate reviewing it sometime soon.
Do LLMs have βWorld Modelsβ?
Melanie Mitchell offers a fairly nuanced consideration of the question of whether LLMs include βworld models.β This is not a simple question, because a βworld modelβ is ultimately just an arrangement of data. One of the interesting framings Mitchell offers is that LLM world models are βemergentβ features of their data sets: that their data allows them to respond as if they knew something about the world, even if they donβt actually βknowβ anything.
As a subset of world models, I donβt think that LLMs that generate video have models of physics, because thatβs simply not contained in the visual data they process. But this is still at least a vastly more interesting and fruitful question than βCan LLMs develop consciousness?β Which has a very simple answer: no they canβt, and also the question is a category error.
Teslaβs Huge P/E Makes Musk Politically Vulnerable
Hamilton Nolan has an interesting thought about the current rising wave of extreme antipathy against Elon Musk, whose last defenders are staging a doomed rearguard action against the rising awareness that he is an incredibly weird, cognitively impaired wannabe-Boer Nazi whose incompetence is destabilizing the government while heβs busy using IVF and a network of surrogates to breed an army of neglected children.
Elonβs nosediving reputation and the incredibly frothy price-to-earnings ratio of Tesla stock together create a massive vulnerability for the quote-unquote βworldβs richest man.β As Nolan puts it, βAny company whose value is this dependent not on its balance sheet but instead on a shared popular narrative is unusually vulnerable to narrative shifts.β
A narrative shift on Tesla, which didnβt result from falling sales but might well follow from DOGE absurdity, could cost Elon many many billions in notional wealth, incredibly fast.
CFPB Drops Suit Against Deceptive Payday Lender SoLo
The βcrime is legalβ era under Donald Trump is going to result in a lot of working-class people getting robbed by finance bros. The now-basically-dead CFPB dropping its suit against the payday lender SoLo is a prime example. A May 2024 filing accused SoLo of deceptive practices in advertising its loans as βzero interest,β when in fact βmany loans originated on the SoLo Platform carried an APR in excess of 300% (treating the tip and donation properly as finance charges).β
Donald Trump thinks thatβs fine, I guess! Just call everything a βdonationβ and itβs fine. The suit has been dropped, with prejudice according to the National Consumer Law Center, meaning that it canβt be re-filed later.
While weβre on the subject, letβs not forget that Trump is also giving $5 billion in prospective overdraft fees back to big banks. Thatβs exactly what his supporters voted for, Iβm sure.
SBFβs Post-Conviction Apology Tour Begins
Sam Bankman-Fried spoke to the New York Sun for 45 minutes on Tuesday Feb 18, the call chopped up into 15 minute chunks per MDC Brooklyn rules. You can listen to the whole thing above, but frankly very little of it is worth your time - more than half is a retread of the same exact shambling self-exoneration Bankman-Fried has tried to sell to various prospects for two and a half years now.
The latest target of his flimsy sales pitch? Donald Trump.
Itβs clear that the right-wing Sun is angling to turn Bankman-Fried into some kind of cause celebre on the right, after he was held up as a nefarious symbol of Democratic corruption. Itβs truly bizarre.
For instance, the interview is introduced with a reference to Danielle Sassoonβs resignation after being ordered to dismiss corruption charges against NYC Mayor Eric Adams. The extremely Trumpy Sun seems to want this to be a ding against Sassoon, and in turn evidence for overturning Samβs conviction - but Sassoon herself is quite right-wing, a Federalist Society member, making her decision to stand up to Trumpβs meddling seem more ethical, not less.
Sam plays into this, also drawing parallels between he and Trumpβs shared (mis)treatment by Judge Lewis Kaplan, describing courtrooms as βmini dictatorships, these little fiefdoms β¦ If the rulings donβt make sense, they donβt make sense, itβs sort of a monopolistic system.β When βcareer advancement for law enforcementβ is at stake, it βblows everything out or proportion.β (Itβs unclear what βproportionβ would mean in his case, since weβre talking about one of the largest financial frauds in American history.) This analysis is of course flatly wrong (Sam himself is in the middle of an appeal), wildly self-serving, covertly anti-democratic - and certaintly tailored to appeal to Trumpβs sense of victimization.
This is frankly a smart way to angle for a pardon while drawing attention away from party politics. Sam also leans into the genuine oversentencing of Ryan Salame, who got 7.5 years for campaign finance charges in a thoroughly scuffed arrangement. Bankman-Fried contrasts Salameβs sentencing with the shorter sentences of his three cooperating co-conspirators, βall of whom were sort of on the other side of the aisleβ - which seems like a scrap of new insight, since I would not have assumed that Gary Wang was a liberal Democrat.
Sam signals that he was already part of the post-Biden rightward βvibe shift,β even before going to prison. βI viewed myself [circa 2020] as center-left, and thatβs not how I view myself anymore, and itβs not how I came to view myself by 2022.β He essentially blames his run-ins with Genslerβs SEC for turning him against the Democrats, and says heβs βbecome more sympathetic to the chainsawβ in dealing with government, and specifically references the SECβs guilt in βstymieing innovation.β Again, the irony and blindness here are thick: Bankman-Fried still hasnβt absorbed that his own real βinnovationβ was crime.
Sam makes sure to explicitly spell out that βI had been giving a lot more to Republicans than had been previously thought, that I had been giving to Republicans a lot more than had been public.β Of course, the incredible irony here is that this was misperceived because Sam Bankman-Fried engaged in campaign finance fraud to cover up his donations to Republicans. (Even though they never went to trial, these charges were treated as proven for the purposes of his sentencing.)
Remember that this is the guy who claimed he had offered Donald Trump $5 billion not to run for President again. On the one hand, this flip-flop is obviously craven tactics, and hey, on some level who can blame a guy for trying to get out of a 25 year prison sentence? But more deeply, I think it just confirms what Kelsey Piper at Vox got the best evidence of: Sam Bankman-Fried doesnβt actually believe anything. He never has. His liberal ethics were performative, and so is whatever heβs trying to signal now.
Really, Sam Bankman-Fried barely even exists. He is an algorithm that seeks money and approval. And thatβs strikingly on display in the interview, the only element that makes it a little worth listening to: Thereβs absolutely no emotional acknowledgment that heβs in prison for most of the rest of his life. The tone is no different than if he were speaking from the Alameda offices in 2019. Itβs subtle, but itβs chilling.
Sam spends the second half of the interview on some boring olβ bullshit: reiterating his absurd claim that he never committed fraud. He repeats the phrase βrun on the bankβ a half-dozen times, insists that the withdrawal pause was just a βliquidity crisisβ that would have been fine: βWe started doing the same thing a bank would do, which is Alameda started selling off all of its investments to generate cash.β
This is a classic SBF lie: one that attempts to cover one crime by brazenly admitting to another. Why are you selling off Alamedaβs assets to fulfill withdrawal demands on FTX? Well, the answer is because Alameda had already stolen the money. Itβs still infuriating how either a) actually stupid or b) shamelessly willing to seem stupid Sam Bankman-Fried is. He still seems to genuinely not understand that letting Alameda use customer funds was massive fraud.
Sam also performs outrage that the bankruptcy estate has now located $15 billion in assets after claiming the money was all gone. Yes, you evil little dweeb, they had to spend years on lawsuits and pressurre to claw back political donations and sell off assets you purchased with customer funds without consent.
Asked what is greatest regret is, Sam says itβs that βI backed down from a fight in Novemer of 2022, and I should not have let Sullivan & Cromwellβ declare bankruptcy.
This was hauntingly evocative of the recent Ziz affair: Zizβs ethics explicitly entailed an injunction to βnever back down.β And like, maybe backing down - actually admitting for once that he was wrong and had failed - was the only thing Bankman-Fried actually did right.
If he had taken the Zizian Rationalist path of zero compromise, how much worse off would victims have been? (Also notable in this context, one of the few institutions that has actively fought against any kind of clawback of FTX funds is the Center for Applied Rationality, which faught the clawback suit from the bankruptcy estate. That clawback has now been dismissed.)
Sam also again touts his charitable donations - many if not all of which we now know were fraudulently funded out of customer funds and were ultimately returned. Self-congratulatory even from prison, he says thought in "ludicrously large terms, billions of dollars.β Except he never actually had βbillions of dollarsβ available for those donations. He simply hallucinated it.
When asked What He Misses Most, Sam doesnβt say itβs his brother, family, or friends (he barely had any), but βbeing able to Google thingsβ and to work. βThe days go by in pretty undifferentiated fashion β¦ Iβm used to being able to dedicate myself to work, and there isnβt much to work on here.β
Finally, SBF reveals a couple of procedural details: Samβs appeal is still ongoing, and according to him there may be oral arguments within three to six months. And he might be on the verge of being sent to prison in California, which he still sounds strangely reluctant about.
It seems quite clear that, for whatever twisted reason, heβd rather be in awful jail conditions at the MDC than in a nice white-collar prison β¦ thatβs also thousands of miles closer to his parents.
βa recent Microsoft study that found LLM users showed declines in critical thinking skills.β
If a readers sees βa study foundβ in an article, they should do themselves a favor and just look at how the study was done before even finishing the sentence. Or just assume it's BS and don't bother.
I went to the link and found this, and basically stopped reading:
β
To see where AI usage shows up in Bloomβs hierarchy, researchers surveyed a group of 319 knowledge workers who had incorporated AI into their workflow. What makes this survey noteworthy is how in-depth it is. They didnβt just ask for opinions; instead they compiled ~1,000 real-world examples of tasks the workers complete with AI assistance, and then surveyed them specifically about those in all sorts of ways, including qualitative and quantitative judgements.
In general, they found that AI decreased the amount of effort spent on critical thinking when performing a task.
β
First of all a single survey of 319 individuals is not that big a sample size, but more importantly they looked at βexamples of tasks the workers complete with AI assistanceβ and then decided that told the researchers how the workers are generally βperforming a task.β (Unless this synopsis is wrong.)
That'sβ¦ I wrote originally βidiotic,β but let's just say βnot rigorously performed, interpreted, and/or reported.β It's like when everyone was reporting the obverse for improving cognition, specifically that playing video games would help the elderly with cognitive decline. First the story was the story, video games cause violence are irredeemable, AI is revolutionary. Then the story weakened/bored, can't really prove video games caused violence, AI revolution oversold and under delivered. Then the man-bite-dogs became the story, video games actually make you smart, AI makes you dumb.
You know what the video games studies really found? Playing video games made players better⦠at playing those video games! Improving one skill/attribute to help another is called transference and you aren't supposed to just assume it exists.
Of course the workers spend less time making judgements with those AI tasks, presumably they did them at one time, then found it wasn't needed any more, or at least parts weren't, as the results were satisfactory without those steps. Being capitalism, I doubt the workers were told they could simply goof off now and got an extra hour each day to play video games. (Or got raises for their increase in productivity, which is what really should be done, but productivity can't scale with wages, or the ultra rich can't be even more ultra than they currently are, can they?) No, they probably spend that extra time using critical thinking on the tasks which couldn't be done with AI. Which are a lot of tasks. But apparently none of that was worth asking about.
Regardless, the methodology is flawed though the conclusion fits nicely the narrative some people want. As I'm sure the opposite would as well, as the βAI is revolutionaryβ claim is still being hocked quite a bit. The only thing βno oneβ wants is the boring truth that a tool made a small part of a worker's job quicker which lead to modest, but not revolutionary, productivity improvement for those tasks. If a bike makes it so I don't have to walk to work, but I walk the rest of the day, does it mean I have lost the ability to walk, or even sacrificed my ability to walk in any meaningful way? Especially if my job generally involves a lot of walking, so I do that in the extra time?
Apparently. Based on that linked article.
β
But this is still at least a vastly more interesting and fruitful question than βCan LLMs develop consciousness?β
β
Eh. First thing would be defining consciousness. World model might be part of it. That plus moving the model forward and backward in time may in fact be consciousness, but only if you are willing to consider consciousness a gradient instead of a binary, which leads to a lot of ethical questions that were always there whether you wanted to think of them or not (eating animals, regardless of how smart they are, etc.).
β
the incredibly frothy price-to-earnings ratio of Tesla stock together create a massive vulnerability for the quote-unquote βworldβs richest man.β
β
Ah ah. Massive vulnerability for his /publicly known/ riches. Back when he was tweeting about DOGE, the token not the recent BS he does to seemingly destroy government, you don't think there might be something convenient and undeclared upside to it? You don't think he might have been doing the inverse of what very likely might happen when Trump calls βtariff onβ and then later calls βtarrif off?β At precisely the times Trump knows he will and so does any other person he decides to tell? Or perhaps at exactly the times Trump is told by someone else to toggle the Tariff Terror switch? Someone who used to be pump and dumping now puppeteering some dump and pumping?
I don't think anyone really knows how much money he's made from seemingly buying the election, but I bet anyone can guess that it's likely Lex Luthor levels.
Today, while I am writing this comment, Tesla is up despite an absolute terrible earnings call. The only explanation the media can think of this is Musk says he plans to step back from DOGE. Why anyone believes him is bizarre, DOGE provides him so much free data on potential customers (or potential people to exploit) and he seemingly can lie to investors all day long and they seemingly believe each time.
But, that isn't to say he can't easily. He already has done a pretty big data dump (that does not seem legal to me), and there's apparently evidence some of that data was intercepted by an IP in Russia. When one smashes and grabs, one doesn't really worry about leaving glass on the showroom floor. And the efficiency stuff could arguably be a kind of βdistracting the clerkβ just long enough to do it.
So, maybe he will be less involved in the very visible, direct way he was before. But I still don't see that being enough to keep Tesla afloat. Some off hand comment online said Tesla was a money laundering vehicle, and a response back said it was the only vehicle that company makes that works. And today it clicked.
From all those stupid tweets back when they were tweets, to now with all the information he got starlinked out of government offices. In both cases doge-ing the prying eyes of anyone that might see any βprofits.β What can he do with any funds from that? His next baby mama isn't going to want Monero and zk-proof sperms from an anonymous 4-channer.
Ah, why not buy his own stock with it? Through cronies and shells and other Trad Fi BS of course. Further, some punters will actually jump in too! And then he can later tell the olβ Donny boy to says the big bad T word again and sell the stock just before, hiding the sale's impact among the S&P 500 debris. Then magically his money is fresh, pressed, folded, and ready to rain on the next stripper turned racist, trad-wife influencer ready to join the gig economy with her uterus. The guy seemingly just loves vehicles.
Seemingly, allegedly, Miss Cleo is only for entertainment purposes, et cetera, ad nauseum.
β
The βcrime is legalβ era under Donald Trump is going to result in a lot of working-class people getting robbed by finance bros.
β
So true. Or per previous, anyone with a 401K.
β
Itβs truly bizarre.
β
I think the word is βdesperate.β When Craig Wright does anything a normal person would be possibly institutionalized for, that's not bizarre, it's just on brand.
β
thatβs also thousands of miles closer to his parents.
β
Ding ding ding.
Scene: Five year old boy crying near his mother, a box of toys, and a trash can. Mother holds a single toy. βThe.β Tosses it in the can. Takes another toy. βEnds.β Toss. βJustify.β Toss. βThe.β Toss. βMeans.β Toss. Mom is stonefaced. Boy stops sniffling. Becomes stonefaced. Smallest smile on Mom: βSee?β Mom is stonefaced. End scene.