YUYANG ZHANG’s Hatched Ideas

Why Learning Gets Better When You Add a Little Noise and a Little Constraint

Why Learning Gets Better When You Add a Little Noise and a Little Constraint

Jun 5, 2026

Thinking Is Not a Style of Speaking, It Is a Way of Searching

Thinking Is Not a Style of Speaking, It Is a Way of Searching

Jun 4, 2026

88%

Why Hallucinations Start with Missing Coordinates

Why Hallucinations Start with Missing Coordinates

Jun 3, 2026

Why Better Models Often Need More Parallelism, Not More Care

Why Better Models Often Need More Parallelism, Not More Care

Jun 2, 2026

1 views

区块链真正的隐喻，不是货币，而是GPU

Jun 1, 2026

5 views

当世界有 29 万个字符时，谁拥有互联网的“地址簿”

当世界有 29 万个字符时，谁拥有互联网的“地址簿”

May 31, 2026

Why Training a Brain Looks Like Training a Policy

Why Training a Brain Looks Like Training a Policy

May 30, 2026

1 views

78%

Why the Future Belongs to Systems That Can Teach Themselves

Why the Future Belongs to Systems That Can Teach Themselves

May 29, 2026

The Hidden Art of Keeping Everything Busy

The Hidden Art of Keeping Everything Busy

May 28, 2026

1 views

68%

Why the Best Learners Hide the Answer From Themselves

Why the Best Learners Hide the Answer From Themselves

May 27, 2026

1 views

90%

When Learning Becomes an Arms Race: What Noise Teaches Us About Inference

When Learning Becomes an Arms Race: What Noise Teaches Us About Inference

May 26, 2026

2 views

87%

When Variability Becomes a Steering Wheel: Learning from Noise and yAntVelocity

When Variability Becomes a Steering Wheel: Learning from Noise and yAntVelocity

May 25, 2026

1 views

34%

The Hidden Contract Between Sequence Models and Generative Models

The Hidden Contract Between Sequence Models and Generative Models

May 24, 2026

1 views

92%

中心被自动化之后，真正值钱的不是权力，而是地图

中心被自动化之后，真正值钱的不是权力，而是地图

May 23, 2026

1 views

Why Every Digital System Needs a Theory of Empty Space

Why Every Digital System Needs a Theory of Empty Space

May 22, 2026

Why Smart Search Beats Raw Instinct, Until It Does Not

Why Smart Search Beats Raw Instinct, Until It Does Not

May 21, 2026

1 views

去中心化为什么需要一颗更好的大脑

May 20, 2026

1 views

68%

Why Better Gradient Flow Needs Fewer Accidental Complications

Why Better Gradient Flow Needs Fewer Accidental Complications

May 19, 2026

1 views

58%

Why Small Updates Beat Total Rewrites: The Hidden Philosophy of Fine-Tuning and Training Loops

Why Small Updates Beat Total Rewrites: The Hidden Philosophy of Fine-Tuning and Training Loops

May 18, 2026

1 views

83%

Why Good Models Win by Grouping Reality First

Why Good Models Win by Grouping Reality First

May 17, 2026

2 views

64%

Why Good State Design Is the Secret to Both Smart Planning and Smarter Vision

Why Good State Design Is the Secret to Both Smart Planning and Smarter Vision

May 16, 2026

1 views

When the Prior Stops Being Fixed, the Problem Starts Moving

When the Prior Stops Being Fixed, the Problem Starts Moving

May 15, 2026

2 views

27%

Why Hallucination Is Often a Training Problem, Not a Thinking Problem

Why Hallucination Is Often a Training Problem, Not a Thinking Problem

May 14, 2026

1 views

Why Good Generative Models Learn to Cheat Less

Why Good Generative Models Learn to Cheat Less

May 13, 2026

1 views

87%

Why Speed and Deception Belong in the Same Theory of Intelligence

Why Speed and Deception Belong in the Same Theory of Intelligence

May 12, 2026

2 views

82%

Why Intelligence Starts by Grouping the World Into Fewer Pieces

Why Intelligence Starts by Grouping the World Into Fewer Pieces

May 11, 2026

3 views

91%

What Ghosts and Food Can Teach Us About Better Gradient Flow

What Ghosts and Food Can Teach Us About Better Gradient Flow

May 10, 2026

2 views

46%

当区块链遇上随机推断：DAO真正改变的不是组织，而是我们如何管理不确定性

当区块链遇上随机推断：DAO真正改变的不是组织，而是我们如何管理不确定性

May 9, 2026

2 views

87%

Why Learning Models Must Learn from Uncertainty, Not Just Data

Why Learning Models Must Learn from Uncertainty, Not Just Data

May 8, 2026

2 views

Why a Good Interval Model and a Good Autoencoder Are Solving the Same Problem