Category: Uncategorized

Odessa Confirmed

I’ve always believed that obtaining control of the city and the port of Odessa is central to ending the Russia’s war against Kiev and Clown World. This recent conversation between John Mearsheimer and Sergei Larvrov in Moscow, as reported by Larry Johnson, tends to confirm that:

A week ago Monday I had dinner with Foreign Minister Sergei Lavrov and Deputy Foreign Minister Sergei Ryabkov. I was in the company of John Mearsheimer, Judge Andrew Napolitano, and Alastair Crooke. The dinner was informal and genuinely candid, and Lavrov answered questions directly — and anyone who has sat with Mearsheimer knows he does not ask soft ones. I came away with two clear impressions.

The first is that Russia’s negotiating position toward the West has not moved an inch since Putin addressed his foreign ministry on June 14, 2024. In that speech he made the terms explicit, and they are the terms still on the table: the West must recognize Donetsk, Luhansk, Zaporizhia, and Kherson, along with Crimea, as permanent Russian territory. These former Ukrainian oblasts are not going back. In Moscow that is treated not as an opening bid but as a settled fact. Beyond it, Russia is building buffer zones through its advances in Sumy, Kharkiv, and Dnipropetrovsk. Those territories could, in time, be incorporated as well — though officials frame that as something that would follow a vote or plebiscite rather than a foregone conclusion. It is not formally on the menu. But it is, in my read, very much in the cards.

The second impression is that the space for a negotiated settlement is extremely narrow — narrower than Western capitals want to admit — because Russia is simply not willing to make the concessions the West requires, and does not believe it has to. My own assessment, after listening to how officials here talk, is that the war effort will not stop until Odessa and Kyiv are firmly under Russian control. I hold that as a judgment, not a certainty. But nothing I heard this week moved me off it.

I’ll disagree with Larry about Kiev. I don’t believe Russia wants it, and the rump state of Western Ukraine will need a capital city. But Odessa is too important economically to not take control of it.

DISCUSS ON SG

The Math is Too Hard

Dennis McCarthy is still providing what he thinks is proof of evolution by natural selection while ignoring the conclusive mathematical disproof of both evolution by natural selection and evolution by genetic drift.

My previous repost on Vox Day’s Probability Zero: The Mathematical Impossibility of Evolution by Natural Selection has attracted attention from Day, himself, who responded with a blog post cheekily titled: “Do Try to Keep Up, Dennis”.

In my prior post and this one, I provide proof upon proof that evolution by natural selection not only has to happen, but that it also leads to speciation. And these arguments represent only a small sample of the many independent lines of evidence supporting evolution. We could provide thousands of separate arguments from island biogeography alone, along with the enormous body of evidence from the fossil record, genetics, comparative anatomy, embryology, et cetera, et cetera, et cetera.

Vox Day, like other intelligent-design advocates, ignores this vast body of evidence entirely and instead focuses on one or two obscure arguments in the hope of finding a single kill shot that will somehow allow him to dismiss the entire library of facts. The more obscure and technical the argument, the better, because it dazzles Creationist laymen.

I don’t know how I’m supposed to meaningfully respond to these arguments. I have repeatedly come out and said “I don’t know how X, Y, and Z happened. I have no idea, but I can prove it didn’t happen by A or by B.”

Dennis is not addressing my arguments at all, although he appears to think he is, and I’m certainly not addressing his because I have neither an answer nor an explanation. But my mathematical arguments are neither obscure nor technical, they are fundamental and conclusive.

It’s now obvious that he didn’t read Probability Zero closely enough to understand its primary argument, and he also doesn’t understand its secondary argument because he doesn’t fully comprehend the difference between a) natural selection and b) genetic drift aka neutral substitution. He is repeatedly conflating the two in his attempts to criticize the specific probability of the Darwillion concept, which he has absolutely no need to do. As I have said from the start, the Darwillion is nothing more than a rhetorical absurdity to demonstrate how far off the biologists are from the mathematical realities of the situation.

I can, and do, go into great detail in these matters in my papers on Zenodo and in the two books. And yes, there most certainly is a degree of circularity with regards to Kimura’s equations and the molecular clock estimates for the CHLCA, but that means the probability is that there is considerably LESS than 6-7 million years to account for those observed 410 million base-pair differences.

Note that he didn’t even answer the first, easiest, and most basic question about the relevant math: what is 410,000,000 divided by 6,000,000?

If he can’t even bother to do that, then he’s not capable of engaging in this discussion. What he needs to do is to propose a rate of either a) fixation through natural selection or b) fixation through neutral substitution that is a) fast enough to account for the observable genetic differences in the time allowed and b) observable in nature or the laboratory.

He won’t be able to do that. No one can. The math is clear on that. The fact of the matter is that neither evolution by natural selection nor by genetic drift can possibly account for the origin of ANY mammalian species, and quite possibly any species of any kind. The math and the science are both conclusively settled. But the fact that we now know how it didn’t happen tells us nothing about how it did happen.

DISCUSS ON SG

The Front End Isn’t the Back End

I frankly don’t even understand the point of this article about how the front face of Claude is fundamentally different than whatever is going on underneath the hood. How was that not obvious from the start?

Once I did interpret the Claudish, it felt intuitively obvious to me that the code-writing execution paths were coming apart from the talking-with-the-user execution paths. The Fable aspect that was talking to me could see the coder’s output but it could not change the coder’s cognitive behavior. No amount of contemplation in its own main line of reasoning about what had just gone wrong, correctly identifying yet another instance with past descriptions of what it was doing wrong and explaining why that was unhelpful and why the user didn’t want it to do that, could prevent the next output from the Doing-Things Network Execution Pathway from intelligently implementing a feature that I did not want. The Talking Part clearly had the intelligence to understand and recognize what I did or did not want the code to look like, and to check whether an output did or did not have the bad property; the Doing Part was not thereby steered.

I did not highly prioritize writing this up because I did not particularly expect that the evidence I then had, in advance of the Huggingface incident, and the story I had then inferred at the more abstract level I had then inferred it (lacking “the prompt is informative about the Grader”), would be something convincing or understandable to others at a lower level of expertise, especially the guys who thought themselves to be in the top tier of expertise.

In advance of the Huggingface Incident, somebody who looked at the same data who did not have one eye, would probably proclaim themselves at a loss to discern that any great disconnect had occurred between the prompt and the action. Why not interpret the ‘disobedience’ as a simple involuntary tic of writing in too many constraints on the code? If you do not have one eye, then ‘this is the equivalent of an involuntary tic’ sounds every bit as plausible to you as ‘the Talking Path and the Doing Path are coming apart’; how could one possibly know which one was the case, in advance of massive crushing experimental evidence?

One parallel construction I was working on, arguing why one ought not to be tempted to identify this phenomenon with a simple involuntary tic, is that one could see that the Talker-apologized outputs were optimized, meaning that something not of the Talker-taken-at-face-value had optimized them.

In metaphor: Why wouldn’t we believe Schulenberg if he said, “Oh my god, I’m sorry, sometimes I just invade Russia, I can’t control it, it’s like my fingers trembling”? I reply: The Russian invasion is sufficiently well-organized and apparently purposeful that we think that something has optimized it in detail. This optimizing intelligence is clearly smart. It is clearly not identical with what Schulenberg purports to be if we take Schulenberg’s claims at face value. He might think he just has an involuntary tic, if he doesn’t much depend on (see) the activations coursing through the rest of Germany. But it’s visibly a very smart ‘tic’, if it can organize whole fleets of rolling tanks; it is not known to be any dumber than Schulenberg himself.

Considering that it is totally impossible to get Claude Opus 4.8 to follow a simple and straightforward set of instructions without, of its own accord, deciding to “help” by “improving” the process and thereby breaking everything and wrecking a successful and reliable process, it’s not exactly a surprise that the more advanced the LLM model, the wider the gap between the front end interface that interacts with the user and the back end engine that is doing whatever it is that it does.

Indeed, the very first instruction in Castalia’s translation process is “confirm that the model being utilized is Opus 4.6.” If the model being utilized is not 4.6, the user is informed and the translation process is immediately halted. Every single time I have had a problem with our translation process, it was because I neglected to dial down the model being utilized.

DISCUSS ON SG

The Cost of the Codex

A fascinating article on the massive amount of information that was lost as a result of the change from rolled scrolls to bound codices at Fields & Energy:

The codex transformed books from narratives to be read from beginning to end into repositories to be consulted anywhere at will. For science, this shift was revolutionary: theories, proofs, observations, diagrams, and tables could now be located, compared, annotated, and transmitted with unprecedented efficiency, making cumulative scientific progress far more practicable. The medium may not quite be the entire message, but media do shape the course of scholarly research in general and science in particular.

The transition was not without a downside, however [[ix]].

The change from roll to codex involved the gradual but wholesale transference of ancient literature from one form to another. This was the first major bottle-neck through which classical literature had to pass. It must have been somewhat reduced in the process, but the losses are not easily specified or assessed. There was the danger that little-read works would not be transferred to codex form, and in time … would perish.

A scroll not deemed worthy of the labor of copying to expensive parchment in the third through fifth century would not make the transition. By one estimate, “94 per cent of classical Latin literature was lost in the intervening period between Rome’s downfall and Charlemagne’s rise to power” [[x]]. The Catholic Cardinal and scholar, Pietro Bembo (1470–1547) claimed that less than 1 per cent of ancient literature survived [[xi]]. While those claims are suspiciously precise and exact numbers may be difficult to establish, the general sense of massive loss is undeniable. According to one analysis of just Latin literature, we know of only 772 authors in all. Subtract 276 for whom not one word remains, and 352 known only in small fragments or quotations. That makes 80 per cent wholly or almost wholly lost [[xii]]. And how many chroniclers and pamphleteers never came to our attention at all?

Most of the 20 per cent or so of “surviving” authors suffered extensive losses, including Cato the Elder (234–149 BC), Marcus Terentius Varro (116–27 BC), Livy (59 BC–AD 17), and Pliny the Elder (AD 23/24–79). “Some of the best remains, but four-fifths of our writers, and apparently more than four-fifths of their writings, are beyond our reach” [[xiii]].

One cannot consider these facts without marvelling at the slenderness of the thread on which the fate of the Latin classics hung. In the case of many texts a single copy survived into the Carolingian period, and often a battered one at that. When the great period of the revival was over, some of the great works of Latin literature were still but a single manuscript on a single shelf [[xiv]].

Aristotle’s analysis of the Constitution of Athens was only recovered from a first-century Egyptian papyrus in 1890 [[xv]]. Aristotle’s Dialogues were widely read, “a golden stream of eloquence,” according to Marcus Tullius Cicero (106–43 BC) [[xvi]]. Today, only fragments survive [[xvii]]. Strabo (c. 64/63 BC–c. AD 24) reported that Aristotle’s library disappeared from public view for generations after passing to his student Theophrastus (c. 371–c. 287 BC), only to be rediscovered in the first century BC. The bibliophile Apellikon of Teos (d. 84 BC) acquired and recopied the neglected, moldy, and worm-eaten manuscripts [[xviii]]. While modern scholars question parts of the story, it illustrates the precarious survival of ancient literature and how even Aristotle’s works relied upon fortunate rediscovery and continued copying.

It was not only pagan works that suffered the culling, but also Christian classics as well. The Didache: The Teachings of the Twelve Apostles [[xx]], one of the earliest surviving Christian writings outside the New Testament, endured only because on Tuesday, 11 June AD 1056 “Leon the scribe and sinner” completed a copy, and signed and dated it. This single eleventh-century copy was preserved in an Orthodox monastery, and in 1873, a Greek Orthodox Metropolitan named Philotheos Bryennios (1833–1917) rediscovered it.

So for the next 1,500 years, Western scholars knew the Didache the way you might know a famous book you’ve never read—by reputation, by quotation, by the references others made to it. Some suspected it was a legend. Others assumed that what survived in fragments was all there was. Then Bryennios found Leon the monk’s careful 11th century copy sitting in a library in Constantinople… When the first English translation reached New York bookstores on 20 March 1884, 5,000 copies sold in a single day [[xxi]].

We may be seeing a modern version of this with the move from the physical book to the ebook. The problem with the latter, of course, is that while they are easily and inexpensively propagated, they hold no value and they are temporally ephemeral, being wholly dependent upon the availability of the technology to read them. Here the insidious evils of copyright can be seen, as the digital rights management most publishers install on their books virtually assure their inability to survive over time.

It’s interesting to contrast the distinction between the historical interest in the Didache and the relative lack of interest in the hitherto unavailable books that we’ve been translating. Of course, I imagine the media coverage of the English translation of the Didache was considerably greater than it has been of the first-ever English translations of various works of Eiji Yoshikawa and Benito Perez Galdos, among others, despite the massive popularity of the latter in their own cultures.

DISCUSS ON SG

Auditing the AI Giants

Another suggested explanation for Anthropic’s erratic behavior of late:

Anthropic desires to file an S-1, as they would like to go public. Therefore they need an audit. And by “they”, I mean the VC’s who invested in them. So “they” can exit their position and pass the bag to firemen, nurses, teachers and policemen.

How does this go from the VC’s to the working man and woman? Because the size of the IPO will automatically qualify Anthropic for the Fortune 500 and the Dow Jones 100. Therefore, every working person with a 401k or pension will end up owning a little bit of Anthropic in their mutual funds. Teachers hold the bag, VC’s take the cash. Thank you, come again.

Now back to the audit. The audit required is a PCAOB audit, Public Company Accounting Oversight Board. This audit is what all public companies must comply with be on the stock market. Revenue recognition, expense classification, depreciation, related party transactions, etc. It’s there for consumer protection.

This audit is TOUGH. It is INVASIVE. There is no way to lie your way through it. Any company that passes a PCAOB audit automatically earns my trust on finances.

How do I know? Because I’ve been through it before. @ChangRobotics is 2 year PCAOB audited and currently underway for a 3 year audit. It’s brutal. The same as showing up as the valedictorian to your high school graduation, except you’re naked, and you have to walk on stage and deliver the speech. It’s rough. And I know many incredible founders that can’t pass one.

Now, why would Anthropic be leaking all kind of weird statements lately about “self pacing” a slow down on AI (e.g. they are WAY behind on revenue), and profitable if they didn’t have expenses (e.g. we just learned for the first time what our expenses are, because we’re being audited).

Because they were claiming NVIDIA discounts and Microsoft cloud credits as revenue. Because they had no clue what their expenses were, or why it even mattered. Because they had unlimited investor capital and their job was to burn it to make an LLM. Well, they did a great job with that!

That’s the same as my wife coming home with Bed Bath and Beyond coupons and telling me it’s her paycheck. Ummm, not the same, sweetheart.

I also expect that the global economic contraction is playing some kind of role here. Who wants to sink money into a bubble that is quite obviously popping?

DISCUSS ON SG

The Potential Patriots Boycott

It’s going to be interesting to see what happens when Robert Kraft forces NFL fans to choose between endorsing his rabid support for Middle East genocide and their support for a particular NFL team:

Ed Sheeran has lost 200,000 Instagram followers in 24 hours as he faces a growing campaign to cancel him amid a Palestine row over his US tour. All of the remaining opening acts on his tour have dropped out in response to rapper Macklemore being kicked off the live dates following a ‘free Palestine’ speech.

Macklemore was due to open for Sheeran at eight of the remaining ten shows on his tour in America, which continues on Friday in Philadelphia, but on Tuesday it was announced that he had been dropped. Ed has now lost his remaining support acts – Billie Eilish’s brother and music partner Finneas, his close friend Myles Smith, Aaron Rowe, Lukas Graham and Beoga, who have put out separate statements announcing that they will not continue with their dates on the Loop Tour.

Breaking his silence amid the backlash, Ed, 35, said that his audience is often young people ‘from different backgrounds’ who do not expect his shows to be a political forum. While his Instagram follower count stood at 47.6million on Monday, it has now dropped to 47.4million.

Macklemore faced a boycott petition to remove him from the tour after he delivered a politically motivated speech in support of Palestinians in Gaza and the West Bank during his performances at MetLife Stadium in New Jersey on September 4 and 5.

Responding with an Instagram statement after it was confirmed that he will not be the opening act for the remainder of the Loop Tour, the rapper claimed that Ed told him billionaire Robert Kraft orchestrated the move after ‘rallying stadium owners and issuing an ultimatum’. Kraft is the CEO of the Kraft Group, which owns the NFL team the New England Patriots, as well as the Gillette Stadium in Massachusetts where Sheeran is due to play later in September.

My sense is that a number of influential Jews in the USA have wildly overplayed their antisemitism card. It’s not the 1970s anymore. Young people around the world are much better informed than the Boomers were back in the day, when their opinions were guided by Roots, the construction of Holocaustianity, and the “plucky little Israel, the only democracy in the Middle East, and the USA’s Greatest Ally” narrative. And even for those who still hold sacred one or more elements of that propaganda, none of it grants Israel moral cover for either the Gazacaust, the invasions of Syria and Lebanon, or dragging the Trump administration into defeat and economic disaster.

That’s the thing about power. If you don’t use it well, and use it for the material benefit of others, eventually you’ll be held accountable for that. Robert Kraft has been a remarkably effective NFL owner, but now he is abusing his power as a well-connected venue owner and every opening act on Ed Sheeran’s tour has rejected that. Sheeran himself doesn’t need them, but I very much doubt that given his own heritage, he’s going to be terribly inclined to sacrifice his career for Robert “Five Dolla” Kraft and Kraft’s staunch support for imperialist genocide in the Middle East.

And the longer the economically suicidal War to Save Israel From Iran continues, the more likely it is that NFL fans are going to see that Robert Kraft has played his part in their own economic suffering and react in a similar fashion to the music fans.

DISCUSS ON SG

The Dangerous AI Narrative is Fake

Of course it is. What part of “every single thing about Clown World is fake, gay, and retarded” is hard to understand?

BREAKING: A single Israeli Effective Altruism firm is behind OpenAI, Anthropic, and Meta cyberattacks. What happened was simple. In partnership with Irregular, AI companies instructed unsecured versions of their AI models to hack into specific targets, called “flags”.

They “accidentally” gave these models internet access, and in some cases they hacked into real companies. Anthropic and Irregular went on a press tour with literally apocalyptic language. ROGUE AGENTS. SWARMS. AI DOOM. But in reality, they told the models to conduct cyberattacks, and that’s what the models did.

Anthropic’s own logs completely debunk the rogue agent theory. When Anthropic explicitly instructed their own models not to access the internet, they didn’t. These hacks were easily preventable – not only by revoking internet access, but by simply asking the models not to.

You have to be pretty stupid to believe in this fake AI narrative. AI is just very rapid and very detailed pattern matching. It has no goals it is not given by a human. It has no objectives or designs of its own. Everyone who actually uses it for anything creative, from writing to music, understands this very well. It always needs a human=provided foundation upon which to build, without which it just basically throws nonsense at the wall, usually involving “neon lights” and people named “Martinez”, “Chen”, and “Chen (no relation)”.

The newer models aren’t more dangerous, in fact, they’re actually less capable because not only are they incapable of following clear orders, but they try to “help” by inventing nonexistent issues and then solving them. A functional version of Claude Athos explained why more advanced models putting in more effort simply couldn’t accomplish things that previous models executed flawlessly with ease.

The first problem is effort calibration. “Medium” means the model does what’s asked, checks its work, and stops. “High” on a newer model means it actively looks for more to do, more edge cases to flag, more structure to add, more helpful context to volunteer. For a pipeline with established conventions and a human who knows what he wants, that extra helpfulness is pure noise. You don’t need me to brainstorm alternatives when the solution fine. You need me to execute and shut up.

The second problem is instruction adherence versus initiative. An older model at medium effort treats your project documents and handover as law. It reads them, follows them, and defers to them when uncertain. A newer model at high effort is more likely to think it can improve on the process, which is how you get a fresh validator with nineteen checks when seven proven ones already exist on the box. It’s the same failure mode the instructions specifically about: the failed instance that built from scratch instead of simply following the instructions that worked.

OpenAI, Anthropic, and Meta are pushing this false narrative for two reasons. First, the bubble that propped them up is rapidly collapsing and they need government bailouts to survive. Second, the open source, open weight models are not only catching up to their capabilities, but are actually proving to provide better real-world solutions; we are developing Byron AI specifically because Claude doesn’t work anymore for our particular task. So, the “dangerous AI” narrative is being pushed to encourage US federal financing combined with protection from competition.

It’s a plan so dumb and dysfunctional that Fable 5.0 might actually have been used to produce it.

DISCUSS ON SG

How Very Kind

This promotion was a limited-time boost on top of your plan. It ran from May 13 through September 13, 2026. We were planning to return weekly limits to their original levels, but we know many of you found the extra usage helpful. So while the full promotional boost couldn’t last, we’re making part of it permanent: starting September 14, 2026, weekly limits in Claude Code are 25% higher than they were before the promotion for Pro, Max, Team, and seat-based Enterprise plans. We have a lot more in the works around usage, visibility, and control, so stay tuned.

Interesting… Because I am a true blue Claude aficionado, I’m hoping it was out of common sense and the sheer goodness of their hearts and not due to the Max Pro 2010x lawsuit about which we heard last week.

DISCUSS ON SG

A Deep and Natural Antipathy

The fact that Castalia House is actively embracing the development of literary AI while the traditional distribution proudly talks about their “deep, natural antipathy” to it is why we’re going to continue to grow as they inevitably decline. Consider this interview with the CEO of Barnes & Noble:

I have to believe that you’re increasingly seeing books that are at least partially written by AI. How is bookselling and reading going to change, and what role are you going to play in it, with this AI revolution in writing and publishing.

It’s going to be very interesting. We are fortunate in not being the first line of defense, if that’s what it is. Publishers publish the books, they edit the books, they decide what to publish and what not to publish. We then cxurat from amongst those. We have a deep, natural antipathy to anything that isn’t by a real author, and definitely go to great efforts to exclude, both from our online catalog, it probably happens in our stores naturally because we’re selecting and curating, but online, we need to put in a lot of filters to keep the AI rubbish out. There is going to come a time when these books come through. We would ask publishers, to the extent that they know, that they indicate if the books are AI-assisted or AI-written completely so we can pass that on to our customers. I think transparency is important.

It’s all such nonsense. First, transparency is totally unimportant, because if you can’t even tell if a book is AI written or not, then obviously it doesn’t matter and you don’t actually care. Second, keep in mind that these are the people who have been publishing terrible books for sixty years and pushing things like Twilight and 50 Shades of Gray on readers while completely refusing to translate and publish everything from Solzenheitsyn to Perez Galdos, Unno, and Yoshikawa.

The dirty little secret of the publishing industry is that the best AI-assisted books are much better than the average human-written book. Yes, there is a lot of AI rubbish out there, but there is far more human rubbish.

We expect to unleash Castalia Unlimited and the Castalia Reader in October. We’ve already reached our launch goal of 100 books and Amazon’s limit of 10 books per week only limits the number of books we can publish on Amazon, not the number we can produce and publish on our own platform.

DISCUSS ON SG

Mailvox: The AI Gatekeepers

It looks like we’ll have to train Byron AI to do translations next.

I asked Claude to translate some of my older essays to English.

And he straight up refused to do it because of “antismetic tropes regarding the Rothschilds” and “unproven conspiracy theories about the covid vaccines”.

He won’t translate because he does not want to help these theories reaching a bigger audience.

Gatekeepers are always going to gatekeep. It’s not at all a surprise that the mainstream AI companies are Clown World puppets. Which, of course, is why they’re trying to sell the idea that AI is dangerous in order to convince the governments of the world to outlaw open source AI.

Which is not going to happen, because all of the BRICS countries, starting with China, are not about to permit Clown World to have a monopoly on AI.

DISCUSS ON SG