YuhaoLin2005

AI agents say they follow rules. I measure whether they actually do. Ugly results included — no cherry-picking. FAFU undergrad.

Joined on Jun 26, 2026 https://github.com/YuhaoLin2005

Education

FAFU, Spatial Information & Digital Technology

Pronouns

he/him

Work

Seeking summer 2026 internship

YuhaoLin2005

Jul 15

I Pre-Registered a Hypothesis. 600 API Calls Later, the Data Killed It.

#ai #machinelearning #python #datascience

5 min read

Want to connect with YuhaoLin2005?

Create an account to connect with YuhaoLin2005. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

YuhaoLin2005

Jul 13

Your Feedback Made This Better — Here's What Changed

#ai #programming #machinelearning #devops

7 min read

Frozen annotations to avoid lookback bias

YuhaoLin2005

Jul 13

Follow-Up: Decision-Token Measurement, Format-as-Fallback, and What Changed

#ai #programming #machinelearning

4 min read

YuhaoLin2005

Jul 12

I Told My AI "You're Safe to Say I Don't Know." Then I Measured What Changed — With Logprobs.

#ai #python #deeplearning #programming

5 min read

YuhaoLin2005

Jul 12

My Experiment Showed Zero Effect. A Statistician Told Me My Measurement Was Broken.

#ai #research #python #datascience

4 min read

Mechanical gates beat rule phrasing

YuhaoLin2005

Jul 11

I Ran 150 Tasks to Test If AI Agents Follow Rules — The Answer Surprised Me

#ai #programming #devops #testing

2 min read

YuhaoLin2005

Jul 10

I Built a Neural Gate for My AI Agent — Layer 2 of Self-Verification

#ai #programming #machinelearning #deeplearning

2 min read

YuhaoLin2005

Jul 10

AI Agents Can't Self-Verify — And That's a Structural Constraint, Not a Bug

#ai #programming #machinelearning #productivity

3 min read

YuhaoLin2005

Jul 9

Meta-Cognition Is the Future of AI Personalization — A 4-Quadrant Framework to Build It

#ai #machinelearning #llm #opensource

5 min read

YuhaoLin2005

Jul 8

My Loss Went Down, But My Model Still Broke — So I Built a Drift Metric

#ai #machinelearning #python #opensource

4 min read

YuhaoLin2005

Jul 7

I Built a Self-Referential AI System. Then Anthropic Discovered the Same Architecture in Claude.

#ai #jspace #promptengineering #deepseek

5 min read

YuhaoLin2005

Jul 7

How I Built a File-Timestamp-Based Feedback Loop to Enforce AI Output Quality

#python #ai #productivity #opensource

3 min read

YuhaoLin2005

Jul 6

The Open Source Flywheel: How I Turn Personal AI Scripts Into Merged PRs

#ai #github #opensource #productivity

3 min read

YuhaoLin2005

Jul 6

Your AI Agent Burns 30% of Its Context Parsing Your Config. Fix It in 5 Minutes.

2 min read

YuhaoLin2005

Jul 4

I Packed DeepSeek V4 + Claude Code Into a Starter Kit. Clone It and Ship.

#showdev #ai #claude #llm

3 min read

YuhaoLin2005

Jul 3

I Open-Sourced the Protocol That Stops AI From Forgetting Who You Are

#agents #ai #llm

2 min read

YuhaoLin2005

Jun 30

I Built a Closed-Loop Self-Healing System for My AI Config — By Accident

#ai #claudecode #automation #systems

4 min read

YuhaoLin2005

Jun 30

Your AI Agent's Best Work Produces Zero Output — And That's the Point

#ai #claudecode #devops #monitoring

4 min read

YuhaoLin2005

Jun 30

I Made My AI Rules Self-Verifiable. Now They Catch Their Own Violations.

#ai #claudecode #productivity #tdd

4 min read

YuhaoLin2005

Jun 29

Search Didn't Make Your LLM Dumber. Unweighted Context Did.

#ai #llm #mcp #agents

3 min read

YuhaoLin2005

Jun 29

I Installed 50 AI Agent Skills Blindly. Here's the 3-Question Filter I Use Now.

#ai #llm #agentskills #agents

3 min read

YuhaoLin2005

Jun 29

Context Engineering Isn't Just for Prompts — It's for Config Files Too

#ai #tutorial #agents

2 min read

YuhaoLin2005

Jun 29

Your AI Agent Is Burning Tokens. Do You Know How Many?

#agents #ai #monitoring #python

4 min read

YuhaoLin2005

Jun 28

Your AI Config Files Are Fighting Each Other

#ai #claudecode #config #llm

2 min read

YuhaoLin2005

Jun 28

I Run DeepSeek on Claude Code — How I Swap Models by Changing Only One File

#ai #claude #deepseek #llm

2 min read

YuhaoLin2005

Jun 27

I Built a Dual-Pool Adversarial Review System for AI Agents — And It Actually Works

#opensource #ai #claude #codereview

3 min read

YuhaoLin2005

Jun 26

Single-Modal LLMs Have a Blind Spot. Here's How to Fix It.

#tutorial

4 min read

YuhaoLin2005

Jun 26

Your LLM Gets Dumber Every Time You Compact Context — And Nobody Is Measuring It

#llm #opensource

3 min read

DEV Community

YuhaoLin2005

Badges

1 Week Community Wellness Streak

Writing Debut

Skills/Languages

Currently learning

Currently hacking on

Available for

I Pre-Registered a Hypothesis. 600 API Calls Later, the Data Killed It.

Want to connect with YuhaoLin2005?

Your Feedback Made This Better — Here's What Changed

Follow-Up: Decision-Token Measurement, Format-as-Fallback, and What Changed

I Told My AI "You're Safe to Say I Don't Know." Then I Measured What Changed — With Logprobs.

My Experiment Showed Zero Effect. A Statistician Told Me My Measurement Was Broken.

I Ran 150 Tasks to Test If AI Agents Follow Rules — The Answer Surprised Me

I Built a Neural Gate for My AI Agent — Layer 2 of Self-Verification

AI Agents Can't Self-Verify — And That's a Structural Constraint, Not a Bug

Meta-Cognition Is the Future of AI Personalization — A 4-Quadrant Framework to Build It

My Loss Went Down, But My Model Still Broke — So I Built a Drift Metric

I Built a Self-Referential AI System. Then Anthropic Discovered the Same Architecture in Claude.

How I Built a File-Timestamp-Based Feedback Loop to Enforce AI Output Quality

The Open Source Flywheel: How I Turn Personal AI Scripts Into Merged PRs

Your AI Agent Burns 30% of Its Context Parsing Your Config. Fix It in 5 Minutes.

I Packed DeepSeek V4 + Claude Code Into a Starter Kit. Clone It and Ship.

I Open-Sourced the Protocol That Stops AI From Forgetting Who You Are

I Built a Closed-Loop Self-Healing System for My AI Config — By Accident

Your AI Agent's Best Work Produces Zero Output — And That's the Point

I Made My AI Rules Self-Verifiable. Now They Catch Their Own Violations.

Search Didn't Make Your LLM Dumber. Unweighted Context Did.

I Installed 50 AI Agent Skills Blindly. Here's the 3-Question Filter I Use Now.

Context Engineering Isn't Just for Prompts — It's for Config Files Too

Your AI Agent Is Burning Tokens. Do You Know How Many?

Your AI Config Files Are Fighting Each Other

I Run DeepSeek on Claude Code — How I Swap Models by Changing Only One File

I Built a Dual-Pool Adversarial Review System for AI Agents — And It Actually Works

Single-Modal LLMs Have a Blind Spot. Here's How to Fix It.

Your LLM Gets Dumber Every Time You Compact Context — And Nobody Is Measuring It