Roman Concrete, Brain "Cognitive Legos," DeepSeek, and Econophysics
Date
Jan 12, 2026
Hosted by Lester Nare and Krishna Choudhary, this episode jumps from ancient engineering to modern AI and markets. We start with the newly uncovered Pompeii worksite that finally shows how Romans mixed their concrete — and why it "self-heals." Then we pivot into a Princeton neuroscience idea that the brain builds complex thought like LEGO bricks (compositional neural subspaces). From there, we break down DeepSeek's "manifold-constrained hyperconnections" as a stability mechanism for scaling deep nets. And we close with econophysics: a Physical Review Letters result arguing the square-root law of market impact is strictly universal across stocks and time.
Featured Research
mHC: Manifold-Constrained Hyper-Connections
Episode 21
mHC stabilizes Hyper-Connections by constraining residual mixing onto a manifold that restores identity-like signal flow, enabling more reliable large-scale training with modest overhead.
Large Language Models
Neural Network Architecture
Deep Learning
AI
arXiv.org
Jan 5, 2026
mHC: Manifold-Constrained Hyper-Connections
Episode 21
mHC stabilizes Hyper-Connections by constraining residual mixing onto a manifold that restores identity-like signal flow, enabling more reliable large-scale training with modest overhead.
Large Language Models
Neural Network Architecture
Deep Learning
AI
arXiv.org
Jan 5, 2026
mHC: Manifold-Constrained Hyper-Connections
Episode 21
mHC stabilizes Hyper-Connections by constraining residual mixing onto a manifold that restores identity-like signal flow, enabling more reliable large-scale training with modest overhead.
Large Language Models
Neural Network Architecture
Deep Learning
AI
arXiv.org
Jan 5, 2026
mHC: Manifold-Constrained Hyper-Connections
Episode 21
mHC stabilizes Hyper-Connections by constraining residual mixing onto a manifold that restores identity-like signal flow, enabling more reliable large-scale training with modest overhead.
Large Language Models
Neural Network Architecture
Deep Learning
AI
arXiv.org
Jan 5, 2026
Strict Universality of the Square-Root Law in Price Impact across Stocks: A Complete Survey of the Tokyo Stock Exchange
Episode 21
Using complete account-level trading data across all liquid TSE stocks over eight years, the authors find the market-impact exponent is statistically consistent with 1/2 at both stock and trader levels—strong evidence for strict universality of the square-root law.
Statistical Physics
Market Microstructure
Quantitative Finance
Econophysics
Physical Review Letters
Dec 15, 2025
Strict Universality of the Square-Root Law in Price Impact across Stocks: A Complete Survey of the Tokyo Stock Exchange
Episode 21
Using complete account-level trading data across all liquid TSE stocks over eight years, the authors find the market-impact exponent is statistically consistent with 1/2 at both stock and trader levels—strong evidence for strict universality of the square-root law.
Statistical Physics
Market Microstructure
Quantitative Finance
Econophysics
Physical Review Letters
Dec 15, 2025
Strict Universality of the Square-Root Law in Price Impact across Stocks: A Complete Survey of the Tokyo Stock Exchange
Episode 21
Using complete account-level trading data across all liquid TSE stocks over eight years, the authors find the market-impact exponent is statistically consistent with 1/2 at both stock and trader levels—strong evidence for strict universality of the square-root law.
Statistical Physics
Market Microstructure
Quantitative Finance
Econophysics
Physical Review Letters
Dec 15, 2025
Strict Universality of the Square-Root Law in Price Impact across Stocks: A Complete Survey of the Tokyo Stock Exchange
Episode 21
Using complete account-level trading data across all liquid TSE stocks over eight years, the authors find the market-impact exponent is statistically consistent with 1/2 at both stock and trader levels—strong evidence for strict universality of the square-root law.
Statistical Physics
Market Microstructure
Quantitative Finance
Econophysics
Physical Review Letters
Dec 15, 2025
An unfinished Pompeian construction site reveals ancient Roman building technology
Episode 21
Pompeii evidence shows Romans "hot-mixed" quicklime with dry pozzolan, creating lime clasts that later feed crack-healing chemistry—explaining the durability of Roman concrete.
Materials Durability
Archaeology
Cement Chemistry
Roman Concrete
Nature Communications
Dec 9, 2025
An unfinished Pompeian construction site reveals ancient Roman building technology
Episode 21
Pompeii evidence shows Romans "hot-mixed" quicklime with dry pozzolan, creating lime clasts that later feed crack-healing chemistry—explaining the durability of Roman concrete.
Materials Durability
Archaeology
Cement Chemistry
Roman Concrete
Nature Communications
Dec 9, 2025
An unfinished Pompeian construction site reveals ancient Roman building technology
Episode 21
Pompeii evidence shows Romans "hot-mixed" quicklime with dry pozzolan, creating lime clasts that later feed crack-healing chemistry—explaining the durability of Roman concrete.
Materials Durability
Archaeology
Cement Chemistry
Roman Concrete
Nature Communications
Dec 9, 2025
An unfinished Pompeian construction site reveals ancient Roman building technology
Episode 21
Pompeii evidence shows Romans "hot-mixed" quicklime with dry pozzolan, creating lime clasts that later feed crack-healing chemistry—explaining the durability of Roman concrete.
Materials Durability
Archaeology
Cement Chemistry
Roman Concrete
Nature Communications
Dec 9, 2025
Building compositional tasks with shared neural subspaces
Episode 21
The study shows that monkeys flexibly switch tasks by reusing shared, low-dimensional "neural subspaces" for sensory features and actions, composing them differently depending on the current task.
Decision-Making
Representation Learning
Cognition
Neuroscience
Nature
Nov 26, 2025
Building compositional tasks with shared neural subspaces
Episode 21
The study shows that monkeys flexibly switch tasks by reusing shared, low-dimensional "neural subspaces" for sensory features and actions, composing them differently depending on the current task.
Decision-Making
Representation Learning
Cognition
Neuroscience
Nature
Nov 26, 2025
Building compositional tasks with shared neural subspaces
Episode 21
The study shows that monkeys flexibly switch tasks by reusing shared, low-dimensional "neural subspaces" for sensory features and actions, composing them differently depending on the current task.
Decision-Making
Representation Learning
Cognition
Neuroscience
Nature
Nov 26, 2025
Building compositional tasks with shared neural subspaces
Episode 21
The study shows that monkeys flexibly switch tasks by reusing shared, low-dimensional "neural subspaces" for sensory features and actions, composing them differently depending on the current task.
Decision-Making
Representation Learning
Cognition
Neuroscience
Nature
Nov 26, 2025
More Episodes
