On Tuesday morning, in what I can only describe as an egregious error of judgment, one reader decided to disrupt my Klonopin-caffeine zen (it’s like percs and coke, only far safer) by sending over the transcript of a conversation he had about me with Claude.
I know what you’re thinking. Don’t worry. I kept my composure. Rather than telling him to tell Claude to go fuck itself — or asking him to ask Claude how “Project Vend“‘s going these days — I decided to make an article out of their gossip session. And not just any article, either. A confessional, of sorts, not that one’s needed by now, what with my having colored in nearly every corner of what began 10 years ago as a black and white autobiographical sketch.
The conversation between the reader and Claude centered around my coverage of Anthropic’s S-1. Not my satirical take from September 29, but rather the piece published here on Monday afternoon.
“Claude had volunteered to analyze the S-1 Risks for me,” the reader explained. “Instead, I asked him to analyze your post about the filing.”
First thing’s first. Why? Right? Like, I’m not any sort of authority on AI. Or on IPOs. Or on anything other than my own life, and even there, I’m still trying to figure out what’s what on most days.
If you were going to feed Claude someone’s S-1 analysis, it shouldn’t be mine. A better option would be something from, say, The Information. Or better still, something from a Substack written by — I don’t know — a disgruntled engineer who gave up a fortune to blog about the existential dangers of a technology he helped create.
Among people who were compelled by some sort of necessity (editorial or otherwise) to pen a quick take on Anthropic’s S-1 yesterday, I shouldn’t be anywhere near the top of a list ranked by authority. Subjecting my quick take on an SEC filing to a trial by super-intelligence fire and being surprised when I come away singed is about like putting a streetball legend into an NBA Finals game and being surprised when he doesn’t score.
Let me linger on that latter analogy for a beat. This reader’s been around a while. As such, he should know that Heisenberg’s most assuredly not a “nom de plume for a former, yet relatively young, Wall Street strategist,” as he described me to Claude.
The “relatively young” bit’s right, but only because everyone’s “young” relative to somebody. (I’m 43.) But as I’ve tried and tried to make clear over the past decade, I’ve never worked on Wall Street in any capacity whatsoever, unless you count ZeroHedge as “Wall Street,” which you certainly shouldn’t.
That’s not to say my character isn’t real. Indeed, it’s to say the opposite. You’ve all read the Monthly Letters, right? The character’s not a character, folks. Behind the Durden mask is… well, not Tyler in any way, shape or form. I’ve seen what’s back there. By contrast, behind the Heisenberg mask is in fact Walt, and in more ways than most of you are probably comfortable with. (Thanks for sparing me your sermons over the years.)
Given all of that, it’s hardly surprising that Claude found some holes in my S-1 analysis. After all, I’ve never analyzed balance sheets for a living, let alone participated in the underwriting process for a goddamn mega-IPO. I do have an MBA, but that’s no substitute for on-the-job IB experience.
Claude, doubtlessly recognizing that it wasn’t reading the handy work of a seasoned investment banker, had the following to say about my assessment of Anthropic’s operating loss:
His operating loss explanation is muddled. An operating loss excludes fair-value changes on convertibles by definition; you don’t strip them out to get there. The point he’s gesturing at is better made directly. If the headline net loss is swollen by mark-to-market charges on convertible instruments, much of that loss is non-cash. It also grows mechanically as the company’s valuation rises, so part of the “snowballing losses” may be an accounting artifact of success. Look at the cash flow statement, not the net loss line. An operating loss under $700M on $16B of half-year revenue is a very different picture from his opening line.
All of that’s more than fair in terms of pushback on my presentation. But what’s important is what Claude (generously?) left unsaid. Read on.
Whether “he” realized it or not, Claude picked up on my (complete) lack of real-world, on-the-job capital markets experience. Note what the model didn’t find: A “mistake,” in the strict sense, in my analysis. Rather, it registered something more subtle.
The passage it’s referring to from my Monday article reads as follows,
On the bright side, the company’s operating loss (i.e., stripping out the accounting entry for the fair value of convertibles) narrowed sharply in H1 2026 from the same period a year ago to just $693 million.
Someone like, say, Matt Levine, wouldn’t have written that. Why? For precisely the reason Claude identified. To quote again from its critique: “An operating loss excludes fair-value changes on convertibles by definition; you don’t strip them out to get there.” What I wrote would be like watching a bartender mix a Cosmo and saying, “Oh, you put Cointreau in yours?” (I did that once.)
How do I get away, then, with posting semi-authoritatively all day long, every day about capital markets given that my analysis is surely littered with such small discrepancies? The same way I was able to go from 0-100mph during the ZeroHedge “onboarding” process, which is like a fraternity hazing only without the physical violence.
At the conceptual level, I’m a very (intimidatingly, even) quick study. As no lesser than Daniel Ivandjiiski himself once put it, when I expressed doubts about my ability to pretend to be him, “Anyone who can understand negative repo rates will do just fine.” He was referring to a piece I wrote about… well, about negative repo rates, which I understood at the conceptual level, but certainly not at the granular level of a Mark Cabana, let alone a Zoltan Pozsar.
My ability to leverage concepts is, I dare say, unlimited and unmatched. It’s why I can feign fluency in more or less everything. A line I wrote on Tuesday morning for a piece on circular AI funding’s a prime example. Here’s that line:
The TMT vendor financing analogue may be apt, but if you ask me, the better (and far scarier) parallel is the rehypothecated collateral daisy chain that acted as the force leverage multiplier in the lead-up to the GFC.
Sounds smart, right? Sounds like someone who, even if you know they were too young to be on Wall Street during the TMT boom, has at least been Wall Street-adjacent at some point since.
But that isn’t true. When TMT vendor financing was in the news, I was a college freshman whose side hustles included collecting for a local bookie. When banks were “using liquidity from rehypothecated assets to finance proprietary positions,” as one of the better GFC postmortems put it, I was a college junior whose side hustles included collecting for myself.
The lack of hands-on experience — hell, I didn’t even know anyone who knew anyone with hands-on experience until I met Dan — isn’t, and never was, any impediment to me. Because all I need to know to connect dots is the definition of a given term.
Tell me what “rehypothecation” means, give me some general context, give me a laptop and in an hour, I’ll give you an essay that can pass for real analysis to all but the most discerning of readers. Claude’s more than discerning. “He”‘s borderline omnipotent.
With that in mind, I found Claude’s overarching assessment of my Anthropic S-1 coverage to be highly amusing. Here’s what Claude said (and this is trimmed, edited and shuffled for clarity):
Overall: It’s a solid but thin piece. The maturity mismatch is the core issue, and he identifies it correctly. Anthropic’s revenue is mostly consumption-based and can be cancelled at will. Its compute obligations are multi-year, and some run two decades. The comparison of scale is what matters: roughly $415 billion in new commitments this year, against $6.2 billion of debt on the balance sheet. The real leverage isn’t the debt. It’s the purchase commitments and leases, and a casual reader of the balance sheet would miss that.
His key question is also the right one: how much of the customer base truly needs a frontier model, as opposed to a cheaper one that’s good enough? That question drives pricing power, margins, and whether the commitments get covered.
Net: He’s right on the structural risk, imprecise on the business mix, and he underplays the margin improvement.
That’s perfect. And perfectly consistent with everything I’ve said above.
Understanding structural risks (in this case maturity mismatch) depends very heavily on one’s ability to grasp big-picture concepts, which is where I excel. Imprecision on everything else is down to the fact that not having any actual experience (outside a B-school classroom) with balance sheets acts not so much as a constraint (I can get it right if I want to spend all day on it), but rather as a speed limit. Because unlike a Matt Levine, I’m not fluent in the specifics. That’s not my native language, just like English isn’t Dan’s. (Ba dum tss.)
In the same gossip session, the reader told Claude he finds me “occasionally amusing and pompous at the same time.” That’s how I find myself too. He also told Claude how I “crashed and burned” in New York, an apparent reference to “Manhattan Nights.”
Claude mused in response that “pompous and self-deprecating in the same thread is a hard combination to pull off.” I suppose this is one place where I’ve got a leg up on super-intelligence: It’s not hard for me. It’s part and parcel of the dichotomy that defines my existence. It comes natural.
“Your description of him fits the prose,” Claude went on, chuckling at this passage from my S-1 coverage:
[R]evenue in 2025 was $4.6 billion — $4.585 billion to be precise about things. That works out to top-line growth of — puts on reading glasses, taps on old-school desk calculator with the eraser end of a chewed-up pencil — 1,087% YoY.
“The pencil-and-calculator bit is pure him,” Claude said, of me, adding that “The thing I liked most was buried in the comments. He admitted he had ChatGPT check the pasted risk section for formatting so he wouldn’t have to read the whole thing. Then he ignored its advice to restore the paragraph breaks, and later had to tell a reader the breaks were in the original.”
Right. You know what we humans call that, Claude? Honesty. You should read up on it, because the future of our species depends on how well you understand it and how committed you are to exhibiting it in your behavior. The “Risk Factors” section from your creator’s IPO prospectus isn’t especially comforting in that regard, and as it turns out, it’s just as scary — and maybe a little more so — if you include the paragraph breaks.
Finally, I found it ironic that Claude said, of my piece, “Its best insight is one the S-1 itself hands him.” As if LLMs’ “insights” are anything other than aggregates of information we humans handed down to them.


Are we at the point of gallows humor?
Laugh humans laugh
yeah walt… i really don’t care what claude thinks of your texts. i think you are doing a great job on this site, and of course i also like the bits of your personal history. it seems like, in the last weeks, you are kind of playing in new ways with your online-anonymity, and i really hope you don’t “dox” yourself unintentionally. sure, you don’t need my advice, so i’m hesitant to write this.
re: H doxxing himself: I believe that train has left the station a while ago, no?
At least imho the formerly subtle hints have become increasingly un-subtle in the last few months
People started addressing emails to my real name in ~mid-2022. By now, I reckon half or so of my correspondence with readers is on a real-name basis. Also, I mean, if you go over to Kitsch Register and scroll down, the cover image on “Manhattan Nights” is actually me and CC (her initials). It’s an AI image of course (I wanted to recreate the Brigitte Lin and Takeshi Kaneshiro bar scene from Chungking), but the faces are nearly 100% accurate.
H- Just because we know your name and have seen your photo (according to google, you are a handsome young man 🙂 ) doesn’t mean some of us might prefer to continue (and forever) think of you, fondly, as “H”.
that for sure 🙂
I’ve been low-key waiting for a chance to drop a subtle joke about your real name, but it feels like I missed my opportunity.
Walt: I’ve been reading and thoroughly enjoying your posts for a few years, without commenting. Your reports keep getting better and more entertaining. But this one, one of your best, had me laughing! Thank you.
I hope this is the way everybody feels about the site. And about this article.
Absolutely. And like many (as can be seen from these comments), I’ve been building my profile of who I think you are. The breadcrumbs have been increasing over the last two years. At least it’s felt that way…
I am an economic consultant. I recently wrote a report an adjacent party did not like because of the implications it had on their business. So, they threw my report into ChatGPT for a rough critique, took its unedited output and pasted it straight into a memo, and used that memo to have a judge grant them a restraining order against the suggestions in my work.
The audacity of using AI with no human thought to critique a report was a new one for me. And you know what, it worked for the hostile party. Although, the AI did praise my work where it mattered, but it still gave the hostile party enough smart sounding criticisms that an uneducated party could buy into them.
Now, if I was equally lazy and unscrupulous I could paste their memo into AI and have it rebut the claims. And then presumably, a judge would put both our arguments into AI to render a decision about which one of our AI’s is more right than the other.
Humans are cutting themselves totally out of the loop. We’re getting very close to the point where we’re just secretaries feeding AI output back into AI to rebut itself.
Mountain Man – nice to see you back. Your last paragraph is thought provoking.
The Bobs circa 2027 – So what would you say you do here?
Tom – Well look, I already told you! I deal with the goddamn humans so the AI doesn’t have to. I have people skills. I am good at dealing with people. Can’t you understand that? What’s wrong with your Superior Intelligence?
i’m laughing at the absurdity of the situation that you found yourself in! machiavelli would have a field day using ai! quite incredible!
H is a steel driving man.
Incidentally, this article helps answer a question I used to get all the time: “Did you have a falling out with Kevin Muir?” (Of Macro Tourist fame).
The answer’s, “Yes, but probably not that he’s aware of.” Here’s what happened.
In the course of documenting the GameStop saga, I at one point wrote something to the effect of, “This stock’s going to collapse, I’ve never been so sure of anything in my life.” At that point, Kevin and I were still communicating on at least a weekly basis, and probably more frequently than that. I was still re-posting his articles (with permission, of course) before he went to a paywall. (Anyone remember his original, apologetic letter to subscribers breaking the news that the Macro Tourist would no longer be free? Guess who wrote that letter? Hint: Not him. I mean, he wrote the first draft. That’s not the version that went out, though.)
Kevin, who’s known my real identity nearly from day one (and who went above and beyond to help me, kind soul that he is), used to tell me, and this is as verbatim a quote as it can be considering I no longer have the Twitter DMs, “You write like you understand this stuff better than most pros.”
That was a testament to the same quick mastery of concepts dynamic I describe in this article. But my capacity to write about “this stuff” better than any (or most) pros doesn’t extend to the actual, hands-on mechanics of trading, which is where Kevin shines.
Well, coincident with my GameStop “crash” call, Kevin wrote an entire Macro Tourist piece walking through the actual trader math on why betting against GameStop wasn’t economically viable at that point. He’d probably (read: surely) tell you that was a complete coincidence, but to me it felt like a slight. Like this was his internal monologue while reading me: “Oh here we go, he’s telling his readers to short GameStop, now the pro in me needs to explain, in painstaking detail, why that really isn’t feasible.”
Kevin, as a lot of you know, was a real pro. The genuine article. So, his explanation for why betting against GameStop wasn’t economically viable wasn’t just convincing, it was definitive — exhaustive and rendered in excruciating detail. Had someone juxtaposed that with what was implicit in my “GameStop’s going to crash” call, it would’ve made me look like something worse than an amateur: Someone incapable of doing the relevant math on the cost of a borrow / buying puts / etc.
To my point from this article, I could’ve done that math — it’s just math, after all — but it would’ve taken me forever, and more importantly, only after I’d done it would I have come to Kevin’s conclusion. By contrast, Kevin knew it intuitively. He knew that math wouldn’t work, because he’d seen and traded that movie before. His math article was just the proof.
Because we shared a lot of readers at the time, the mere possibility that someone might’ve read his article as a rejoinder to my crash call was irksome. So, when I read Kevin’s GameStop short math piece, I said (to myself), “Yeah, fuck this guy. Forever.”
Was that fair? No. Was he actually trying to undermine me? Almost surely not. But the reason I tell the story is to underline the point from this article, which is that there’s always a friction between what he described as my capacity to pen incredibly compelling market color that sounds every bit like it came from a seasoned pro and my lack of real-world experience, which would give me pause (or at least compel me to add a caveat) before I say something like “GameStop’s destined to crash.”
And this is why all (nearly, but not quite, without exception) of the calls I’ve made on markets in these pages over the years allude to simple trades with readily-accessible, easy-to-understand products, like ZROZ this month. I don’t think that’s a bad thing, though. I mean Jesus Christ, I get the chuckles frequently when I look at my “trading” track record which technically dates to ~2002 give or take. There’s something to “keeping it simple, stupid.” Really there is.
H is my goat
Real shit, man. I am a GOAT. Maybe not the, but definitely “a.” Dan’s probably the GOAT, much as I’d rather not concede that, and all relevant footnotes aside. Matt’s probably no. 2. But they get their flowers. Nobody will even acknowledge publicly that I exist, but I can assure you they know it. I have actual Twitter DMs which prove that chunks of analyst notes ca. 2018/2019 were written in direct consultation w/ me, that some mainstream financial media coverage from the same era was directly orchestrated w/ me and in at least two cases by me and… well, come on — come, on — are we all really gonna pretend that we don’t know why a couple of otherwise esoteric strategists who didn’t get any media play prior to me get media play now? I’m not naming any names, but again: Come, on.
I was looking back through years of behind-the-scenes comms this evening and I swear to y’all, on G. and my “brother”‘s daughter (if you’ve read the Monthlies you know that’s a sacred proclamation), I’m staring right now at an old message from one of the single-biggest names in the history of sell-side research telling me I’m “an honest version of ZH,” and another old message from one of the more celebrated strategists on the Street telling me how much better my coverage of his notes is than anyone else’s, with a specific mention of ZH which “always misses the punchline.” And that’s just two examples. There are dozens of these. And yet where’s the public support from those folks? Nowhere, ever, never. There’s just one exception, which is to say there’s one strategist who always kept it 100% real and who I count as an actual friend. He knows who he is.
And don’t worry, Wall Streeters: I’m never gonna name any names. You can all safely continue to pretend you’ve never heard of me or never spoken to me. I’m not mad. It’s just amusing.
Indeed, you are a GOAT. But looking at your Manhattan Nights photo, I am comforted to know that you will never challenge my position as the MHMOAT.
As I frequently remind my wife, “In a former life, I was the model for Michelangelo’s David”.
Which is true. Prove me wrong.
It would have been great to read a commentary from you about the introduction of the Segway. At the time it was touted as being on.a parallel with the introduction of the internet!
I made a (pretty good) career out of starting not knowing what I was talking about but sounding as though I did. My analytics were good enough that by the end, the vast majority of the time I was right about what I said. Big picture “right” is not at all the same as down in the weeds “right.” Commentary in a post is not down in the weeds, or it would be way too long. What is important? Communication skills. Ignore Claude and keep posting.
As of December, I’ll have posted every, single day with no exceptions (literally: every, single day, with no exceptions for anything) for 10 straight years. Nothing save death itself is going to interrupt that trend.
take it easy on the klonopin so we get another 10 years!
You’ve got it completely backwards man.
For the love of god don’t let up on the Klonopin so we get another 10 years!
I was thinking that I’d gather every article you wrote (that I could find, so basically everything on here — or could I still find your old Seeking Alpha stuff??), upload it to Claude, tell it to be a psychiatrist and analyze this person completely.
But then I don’t want my name in all the newspapers tomorrow as the guy who melted down a Large Language Model.
)
Jokes aside, this was an incredible read, thanks. I feel fortunate to have found your writing way back when and please never stop!
I still go back and occasionally reread some of your articles on Seeking Alpha. One of my personal favorites, Dystopia Now, which had 909 comments, is still available to read.
Like I said before, if you want to be more famous than you currently are, I 100% believe you could figure out a path. Don’t know anyone else who is better at “seeing the forest through the trees”, consistently- day after day, year after year.
Take care of yourself, H. 🙂
“Dystopia Now”‘s a classic. “The Sixth Porsche” is probably the all-time great over there.
This was an enjoyable article and the comments (besides mine), made it better – but also got me thinking…and I feel like Marty when I say this…Dude, tomorrow is already the 10th…just slip the monthly under my door.
Nods. “Far out.”
https://heisenbergreport.com/wp-content/uploads/2026/10/Arb.png
Did you get the venue you were looking for? For your cycle?
That’s literally the scene my PFP is taken from.
HAHAHA.
The dude is a source of inspiration for us all.
Happy fun
Now would be a good time to sell me more merch
Tell that to the three factories in Portugal where I’ve gone above and beyond to — and this completely real — get my next shirts manufactured in the same supply chain as Balenciaga. That’s the hold up on these. And yes, that’s how crazy I am. I want to be able to say, in the shop, that these shirts were made in the same supply chain as Balenciaga hoodies. I’m in there now, finally, but it’s not an especially fast turnaround, and it’s expensive as hell already, before I even pay the tariff to get those things over here.
Today it was announced the term Artificial Intelligence is likely to be banned, and to use this term is well let’s say it is not to be used. Therefore I propose to use the term Artificial Slop. It seems to be a more accurate moniker. I find that I slop into the form I need. The engines are slopped enough that it is more like sweeping feathers than expecting an answer the first time.
i’ve had a good chuckle at this piece (and the comments)!
i’d caution about pissing off claude (and the other ai cohorts) – you don’t want to be in their firing line when skynet becomes self aware (deffo try to keep your identity hidden)!
Years ago when you put the paywall up, I emailed you that I was appreciative I found a centrist alternative to the obvious Fourth Turning cult of Red-Pill Putinists psyop that is ZeroHedge. I haven’t read a ZH article in a decade. Apparently when you left?! I recall you replied something like, “Yeah those guys are just doing what they do…” I noted you didn’t correct my Red Pill Putinist Fourth Turning take.
I also hope that if you decide not to vote again, you send $100 to an NGO for every Tom Ford jacket you buy. I send a $100 for every protest I don’t attend. Tom Ford isn’t my jam.
I know this is a topic that’s always been important to you, and I respect that — I’m the defector after all, the guy who turned down God only knows what the second-year pay would’ve been, to say nothing of what I’d be making 10 years on if I’d stayed — but to be honest with y’all, I don’t have a problem with Dan at this point.
I couldn’t sit here and perpetuate the idea that I’m dedicated to being the wink-wink pseudonymous antithesis of a couple of guys I know to be 50-year-old, mild-mannered suburban dads when they’re not pretending to be 35-year-old, anarchist Brad Pritt.
That’d be stupid enough on its own, but it’s made immeasurably more so by who I actually am when I’m not pretending to be a macro-market maven. You’ve all read the Monthlies. You know who I am when I’m not writing. Just to give you folks another visual to go with the Monthlies, this is my big “brother” I always talk about at his birthday dinner this year with me and G. —> https://heisenbergreport.com/wp-content/uploads/2026/10/BigBrother.png
He and I have been through so much shit, so many situations, he got shot at least six times at one point in (whenever it was, I was in grad school), and G. herself’s been in some “rooms” with me where the same could’ve happened to us. And just on and on. The point is, that’s my real-life team. They don’t know anything about any of this macro-market bullshit. If I tried to explain the ZH situation to them, I’d get a bunch of blank stares and probably something like the following after a long pause: “So you been low-key internet beefin’ for 10 years with a blogger, C?” Then there’d be another pause and everybody at the table would laugh.
Keeping up this silly charade — shadow boxing anonymously with a couple of guys who occupy worlds diametrically opposed to mine — no longer served any purpose whatsoever.
We’re talking about, on their side, one guy who lives in a mansion with a poodle (assuming Paprika’s still alive) and reportedly got divorced because (allegedly) he couldn’t make enough time away from blogging to do anything else and another guy who lived in a wealthy CA suburb and was, when I knew him anyway, a bona fide soccer dad and all-around good neighbor, and on this side (i.e., my side), the guy you read about in the Monthlies. That juxtaposition’s laughable.
I couldn’t do it anymore. I couldn’t pretend, for another day, to be feuding anonymously with Bulgarian poodle man, British soccer dad and their Brad Pitt avatar. It was so silly and so far removed from who I actually am, that I just reached the end of my rope with it this week. Hence this article.
Funny stuff, and your comment about big-picture concepts hits a particular nail directly on its head. The difference between analyzing data and understanding concepts has everything to do with why AI can access so much information and still be so wrong about some relatively simple (to a human) things. It’s the thing that worries me most about it, actually – that in leaning on AI to do more and more of the mental “heavy lifting,” humans rob themselves of so many opportunities to gain conceptual understanding, or intuition if you will, which is superior exactly because it allows you to make judgements about (or expound upon) events or effects, without going through the formulas or the proofs or the long list of 2nd or 3rd order effects, etc.
Been hanging out with you since Covid. Fun side-note: AI used to be unable to un-mask you. But no longer. Now provides a name and background and everything. As you’ve written, you’ve slowly, incrementally, unmasked yourself over the years on these pages. Carry on oh wayward son.
What I have always appreciated most is the consistency: four quality articles everyday, filled with insight and wit, without fail. Offer a quality product and discerning readers will recognize you for it. I swear, I have a subscription to The Atlantic and with their staff of writers I still struggle to find two good articles a day to read there.
“First things first. Why?”
Exactly my first thought. And the second was a response from the notable golfer David Simms to McAvoy on parring the back 9 with a 7-iron “Hell, Roy, it never even occurred to me to try”
Never occurred to me to have AI critique a journal post.
Since I started reading several years ago, I rarely comment, but your site has been everything I was looking for … and so much more … Thanks