Linus Mixson

Had to mute Lisan

@Saransh_X Maybe, but the evidence is generally accepted to suggest this order of descent (not a hierarchy, really, just a lineage), and that evidence does seem pretty ironclad on inspection. But who knows. This is definitely the best reconstruction, but it could be somehow wrong.

@LinusMixson It would not be appropriate to establish such a hierarchy here. They may simply represent contemporary developments of the scripts. 🤷🏻

Anyway did you guys know that the Thai alphabet is a descendant of Egyptian hieroglyphics

@asinwo_tbc This is the exact opposite of observable reality, but it's interesting to see the little Plato's Cave situation going on over at Hindutva-Kshetra.com.

@LinusMixson literally not a single serious archeolinguist thinks brahmi evolved from aramaic. you are parroting 110 yr old hypothesis.

@phlegm_attic The Egyptians straight made those things up. But in all seriousness it seems like little drawings used to label objects with their owners, basically, that were gradually formalized into a proto-linguistic form and then eventually a full script.

@LinusMixson What came before Egyptian hieroglyphics?

Wrong. You can line them up next to each other if you'd like. You would never in a million years get that letterform/phonetic value correspondence by chance. And IVC isn't a real competing hypothesis because it looks no more similar to Brahmi than Aramaic (with a lot of the "correspondences" being explainable by IVC having literally 10x more characters to mix-and-match a "correspondence" from), and it hasn't been decoded, meaning that you can't establish phonetic correspondence at all.

@LinusMixson Incorrect. Brahmi is not from Aramaic.

@jagunanthi I'm sure that the hand-rolled assembly is actually faster than you can get from any sane compiler. He's probably not wrong about that. But he'd be insanely asshurt if Mythos wrote better assembly than him. I would actually be worried about him if that happened.

@LinusMixson Someone should have mythos rewrite FFmpeg in rust just so this guy finally shuts up about his hand rolled assembly

Losing sympathy for the FFmpeg guy, whose feelings are apparently hurt because ffmpeg got pwned 21 times by Mythos after he spent a year straight talking about how LLMs are toys unsuitable to compete with hand-turned assembly & has taken recourse to suggesting that no one will let lil' old fame-poor ffmpeg into an open-source AI initiative.

It sets exactly the sort of precedent they want, they want it to be established as a legitimate possible intervention, it won't impact their bottom revenue line in the end, and it gives them insane, insane aura going into IPO. It's a huge win. I'm sure they're a bit irked by it but unless the admin decides to just let OAI leapfrog them out of hatred I have to imagine the perceived benefits outweigh the perceived cost.

@LinusMixson yeah i didn’t sense criticism. you’re probably right. my initial reaction was that their values must be inconsistent. i.e., safety is incompatible with broad deployment. but this is also kind of exactly what they asked for.

it’s almost as if the finding of a narrow potential cyberrisk landscape should not have been cause for gating the commercial release of a model that could have been agi uplifting hundreds of millions of people for months now or something

@seeingmole I think that's besides the point being made here because it pretty clearly comes from Khmer, whether via Sukhothai or some other intermediate.

@LinusMixson Moden Thai script isn't derived from the Sukhothai script, which is a forgery. The whole Sukhothai obelisk was a forgery commissioned by Rama 4 king.

There are actually distinct, marked, and arguably overwhelming similarities between Aramaic and Brahmi, ones that are quite difficult to dismiss because they are extraordinarily unlikely to have arisen without heredity, and certainly more similarities identified than are known between Brahmi and IVC, which has not been deciphered & therefore cannot be shown to demonstrate the sort of sound correspondences that exist (alongside corresponding letterform similarities) between Brahmi and Aramaic.

@LinusMixson Brahmi traces back to ivc script nothing is similar here

@elliaha_mu Yes, that would arguably be more apt. Thank you.

@LinusMixson Hieroglyphs you mean

This isn't my dream, idiot. I would be thrilled to learn that Brahmic scripts were actually indigenous; that would be very cool to me. Unfortunately the scholarly rough consensus denies me that. But thank you for letting me know that you think that Indian linguistic genealogy is a very emotional issue for me, a bizarre reaction on a post about *Thai* that only indirectly mentions Brahmi & does so with the same emphasis with which it mentions Khmer, and one that only makes sense if it's a really terribly intense emotional issue for you.

@LinusMixson Tamil is older than Aramaic. Sorry to break your dream. Even at the time of Alexander, the Greek, Roman ,egyptian, Persian and different civs of India were trading and growing up together. There is no static linear transfer of language like you wish for.

@neilalexanderw1 @KostyonkiLady @spandrell4 Insane, I had no idea. And Cyrillic is a *merger* of Greek and Glagolitic? Totally wild. Thank you.

@KostyonkiLady @spandrell4 @LinusMixson Correct

Every single living writing system besides Chinese is descended from Egyptian through Phoenician.

@logicus Oh, yeah, I wasn't criticizing you. I just personally can't help but suspect that Anthropic is delighted out of their minds about this.

they’ve certainly tied their own shoelaces in various knots here. on what you said, yeah, i’m sure they would love for the us govt to police oai according to ant’s own safety principles. i mostly detected the usual shock and outrage they typically have toward external criticism. this one was just hard to ignore. the deeper rub here is, are their models actually less safe than equivalently powerful oai models? i definitely can’t say in any disciplined way, but i suspect oai doesn’t get enough credit for how well-aligned its models are.

I already downloaded your pinned PDF and glanced it over. I also cloned your GitHub repo and ran your analysis. I even began to work on a way to exclude the priors you tendentiously injected into the process so that we could get a realistic sense of how your analysis measures up to competing applications of the same methodology that aren't retrofitted to the desired conclusion. But then I realized — your paper claims, in the very abstract, that "Indus inscriptions are in grammatically correct post-Vedic Sanskrit." So surely you can apply your methodology to the known corpus and return sane, plausible, grammatically-correct post-Vedic Sanskrit readings of all of them, right? And surely you would have already executed this rather simple undertaking to weed out issues with your rather extraordinary claim? So where's this corpus? I notice you haven't been able to publish this research in a journal. I can't think of a bigger, more impossible-to-ignore Indic-studies bombshell than a full, systematic decipherment of the Indus Valley inscriptions into grammatically-correct post-Vedic Sanskrit. It would get you taken seriously. Let's see it.

@LinusMixson Read this open.substack.com/pub/yajnadevam…

That Brahmi evolved from Aramaic in the same fashion as Phonecian evolved from Egyptian, i.e. gradually over the course of repeated usage, via adaptation and simple drift: that is, the usual process of script evolution, attested globally over and over and over again, and not as the grand Act of a King.

@LinusMixson @PunPun7507 @Lib9860 Ao what's the hypothesis, for you?

@maattttbrown @SFGate @bryancsk 😕

Stanford grads walk out as Google CEO Sundar Pichai takes the stage as commencement speaker. No mention of AI, unlike other uni speakers getting booed down this year. Story for @sfgate shortly

@xlr8harder @max_spero_ That being said, I would imagine that Google does *not* smoke test.

@xlr8harder @max_spero_ I suspect that all SDE-related terms that are remotely non-obvious show the same spoke.

My anecdotal evidence that LLMs love the term smoke test seems confirmed by search trends — searches for the term spike in 2026

@antirez No thank you

Redis users: would you like to have sorted sets using just 50% of the memory without any speed regression? I got a few ideas...