George3d6
@george3d6.bsky.social
Opinions are legally binding to all my employers, retweets are deep endorsements.
I write at: https://cerebralab.com
I write at: https://cerebralab.com
Read the original at: cerebralab.com/read/8
January 28, 2026 at 3:04 PM
Read the original at: cerebralab.com/read/8
mathematics—and won't be simplified for similar reasons (incentive problems). But it remains to be seen if there is any practical application for vibe maths, and my impression is that there will be; but it will reside in helping people apply "trivial" mathematics to "trivial"
January 28, 2026 at 3:04 PM
mathematics—and won't be simplified for similar reasons (incentive problems). But it remains to be seen if there is any practical application for vibe maths, and my impression is that there will be; but it will reside in helping people apply "trivial" mathematics to "trivial"
unlike mathematics, code produces useful artifacts that people can iterate and refine. I am a fan of vibe coding, since it leads to a world where people stop needing to learn the (overly complex) rules for coding, which in itself is pointlessly complex for similar reasons to
January 28, 2026 at 3:04 PM
unlike mathematics, code produces useful artifacts that people can iterate and refine. I am a fan of vibe coding, since it leads to a world where people stop needing to learn the (overly complex) rules for coding, which in itself is pointlessly complex for similar reasons to
compile. Assuming a mathematics that did compile, we'd run into the same issue that we run into with LLMs writing code: great ability to run through walls, inability to understand the global context that makes the code useful or the product good. Vibe coding is "a thing" because
January 28, 2026 at 3:04 PM
compile. Assuming a mathematics that did compile, we'd run into the same issue that we run into with LLMs writing code: great ability to run through walls, inability to understand the global context that makes the code useful or the product good. Vibe coding is "a thing" because
lossy—it's not a "definitive" abstraction. It falls down to human judgement whether a certain thing is "true" or "false". This human judgement involves relative agreement among 20 to 500 specialists depending on the field, but it's human judgement nonetheless—mathematics doesn't
January 28, 2026 at 3:04 PM
lossy—it's not a "definitive" abstraction. It falls down to human judgement whether a certain thing is "true" or "false". This human judgement involves relative agreement among 20 to 500 specialists depending on the field, but it's human judgement nonetheless—mathematics doesn't
Symbolic Confusion LLMs generalize to maths for two distinct reasons: They are hyper-optimized for generalizing on mathematics. Mathematics, much like language, is a form of lossy symbolic reasoning. I wrote more about the second reason, the fault of mathematics is that it is
January 28, 2026 at 3:04 PM
Symbolic Confusion LLMs generalize to maths for two distinct reasons: They are hyper-optimized for generalizing on mathematics. Mathematics, much like language, is a form of lossy symbolic reasoning. I wrote more about the second reason, the fault of mathematics is that it is
of general intelligence; it's an impressive accomplishment and serves to make us understand what certain parts of intelligence are, but you could make the algorithm infinitely better at DOTA or Starcraft (or Go), and it would be nowhere closer to general intelligence. Coding And
January 28, 2026 at 3:04 PM
of general intelligence; it's an impressive accomplishment and serves to make us understand what certain parts of intelligence are, but you could make the algorithm infinitely better at DOTA or Starcraft (or Go), and it would be nowhere closer to general intelligence. Coding And
other fields, e.g. music, where the best instrumentalists are often autistic savants who have crumpled their minds into a hyper-specific shape only applicable to their obsession. If I show you an RL algorithm that is great at DOTA or Starcraft, you will not assume that is a sign
January 28, 2026 at 3:04 PM
other fields, e.g. music, where the best instrumentalists are often autistic savants who have crumpled their minds into a hyper-specific shape only applicable to their obsession. If I show you an RL algorithm that is great at DOTA or Starcraft, you will not assume that is a sign
arbitrary puzzles. Mathematics is, in other words, a very bad benchmark. Not for humans, mind you, but for LLMs. Often, we find (with both pro-gamers and mathematicians) that it's not the best participants who generalize to other fields—it's the tiers below that. This applies to
January 28, 2026 at 3:04 PM
arbitrary puzzles. Mathematics is, in other words, a very bad benchmark. Not for humans, mind you, but for LLMs. Often, we find (with both pro-gamers and mathematicians) that it's not the best participants who generalize to other fields—it's the tiers below that. This applies to
be so much a function of intelligence as a function of caring. It's about as impressive as being the best DOTA or Starcraft player—curious, to say the least, but not indicative of an ability to reason, just indicative of grit and a strange proclivity for a hyper-specific set of
January 28, 2026 at 3:04 PM
be so much a function of intelligence as a function of caring. It's about as impressive as being the best DOTA or Starcraft player—curious, to say the least, but not indicative of an ability to reason, just indicative of grit and a strange proclivity for a hyper-specific set of
where the stakes are too low. Or rather, about 20-odd games or so, united by loose threads but fundamentally different, since even specialist mathematicians seldom communicate across their specialty gaps. A Bad Benchmark Thus, "solving" a lot of hard mathematical problems may not
January 28, 2026 at 3:04 PM
where the stakes are too low. Or rather, about 20-odd games or so, united by loose threads but fundamentally different, since even specialist mathematicians seldom communicate across their specialty gaps. A Bad Benchmark Thus, "solving" a lot of hard mathematical problems may not
lifetime to learn and are used in no practical field, mathematics is a beautiful representation of reality. Mathematics has become a nerdy circle-jerk, and the reason nobody cares to break it anymore is that simply "learning the rules" would take too long. It's a boring game
January 28, 2026 at 3:04 PM
lifetime to learn and are used in no practical field, mathematics is a beautiful representation of reality. Mathematics has become a nerdy circle-jerk, and the reason nobody cares to break it anymore is that simply "learning the rules" would take too long. It's a boring game
avoiding errors. One can hold two social theories as to why mathematical systems no longer "break" as often as they did in the early 20th century: Mathematical systems are nearly perfect, and in spite of the fact that the last hundred years of mathematical advances require half a
January 28, 2026 at 3:04 PM
avoiding errors. One can hold two social theories as to why mathematical systems no longer "break" as often as they did in the early 20th century: Mathematical systems are nearly perfect, and in spite of the fact that the last hundred years of mathematical advances require half a
theory. The "reaction" we have to these formal systems breaking is often to try and create a less error-prone formal system—but these new systems are necessarily more complex, and the drive to create them is not aesthetic or practical, but rather a nerdy myopia focused on
January 28, 2026 at 3:04 PM
theory. The "reaction" we have to these formal systems breaking is often to try and create a less error-prone formal system—but these new systems are necessarily more complex, and the drive to create them is not aesthetic or practical, but rather a nerdy myopia focused on
What the world understands today as mathematics is a narrow corner of what mathematical thinking was, historically. There are certain cases of individuals pointing out flaws in symbolic systems that "break" it; popular examples being Ramanujan with calculus and Gödel with set
January 28, 2026 at 3:04 PM
What the world understands today as mathematics is a narrow corner of what mathematical thinking was, historically. There are certain cases of individuals pointing out flaws in symbolic systems that "break" it; popular examples being Ramanujan with calculus and Gödel with set
Tariffs imposed on China are there primarily because they are the main exporter of highly concentrated flouride salts that can protect teeth from cavities and plaque formation.
It never made economic sense, for you, it's a conspiracy to protect big Organic and the detnists.
It never made economic sense, for you, it's a conspiracy to protect big Organic and the detnists.
April 10, 2025 at 6:47 PM
Tariffs imposed on China are there primarily because they are the main exporter of highly concentrated flouride salts that can protect teeth from cavities and plaque formation.
It never made economic sense, for you, it's a conspiracy to protect big Organic and the detnists.
It never made economic sense, for you, it's a conspiracy to protect big Organic and the detnists.
This is roughly why I expect tariffs to be economically bad for the US (or any advanced economy) -- and good, if used correctly, for a poor country -- but only "good enough" to get it to be middle of the pack.
April 2, 2025 at 4:04 PM
This is roughly why I expect tariffs to be economically bad for the US (or any advanced economy) -- and good, if used correctly, for a poor country -- but only "good enough" to get it to be middle of the pack.
Building/Scaling the product is tertiary and fully fungible.
This applies even to strict & authoritarian cultures, the only difference is that you need to place the constraints in a more top-down way.
This applies even to strict & authoritarian cultures, the only difference is that you need to place the constraints in a more top-down way.
April 2, 2025 at 4:04 PM
Building/Scaling the product is tertiary and fully fungible.
This applies even to strict & authoritarian cultures, the only difference is that you need to place the constraints in a more top-down way.
This applies even to strict & authoritarian cultures, the only difference is that you need to place the constraints in a more top-down way.
People without sufficient constraints have "lack", filled by whomever can produced constraints (needs) for them.
This is the highest-value economic activity in any market:
- Create constraints for other people
Creating a product that solves the resulting needs is secondary.
This is the highest-value economic activity in any market:
- Create constraints for other people
Creating a product that solves the resulting needs is secondary.
April 2, 2025 at 4:04 PM
People without sufficient constraints have "lack", filled by whomever can produced constraints (needs) for them.
This is the highest-value economic activity in any market:
- Create constraints for other people
Creating a product that solves the resulting needs is secondary.
This is the highest-value economic activity in any market:
- Create constraints for other people
Creating a product that solves the resulting needs is secondary.
I've spent like ~10 years training myself to do this and I am still bad at it, having needs is a skill (a very valuable one, people pay for it).
I know very few people that can formulate needs for themselves, and they usually use said skill to formulate needs for other people.
I know very few people that can formulate needs for themselves, and they usually use said skill to formulate needs for other people.
April 2, 2025 at 4:04 PM
I've spent like ~10 years training myself to do this and I am still bad at it, having needs is a skill (a very valuable one, people pay for it).
I know very few people that can formulate needs for themselves, and they usually use said skill to formulate needs for other people.
I know very few people that can formulate needs for themselves, and they usually use said skill to formulate needs for other people.