2026年3月10日 星期二

HW#3 IC Design with Advanced AI

建議工具

使用 Claude Sonnet 4.6 推理模式(手動切換,免費用戶額定時間內只能使用三次)

使用 ChatGPT 5 推理模式(自動切換)

使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)

使用 Grok 4 推理模式(自動切換)



How to publish a Claude artifact

How to share a ChatGPT link

How to share a Grok link

How to share Gemini Link


Content share 作業繳交格式

  • share only link, pure text, markdown (md)
  • no attachments accepted, no html, screen dump, or png
  • non-compliant homework will be rejected and returned to you


課堂練習 

Deadline: This Saturday at 23:59

Send all the share links to  me chang212@gmail.com by email with subject HW#3  [your id, your name]


3/10 只要做第一題(內有5小題)


 1.  Study Apple C1 Architecture

(a) make

LNA schematic (TSMC N7 製程的完整設計流程,包含完整元件表、元件值推導公式、die 面積估算) LNA=Low Noise Amplifier


(b) make

LNA Die


(c) Build

LNA Optimizer (SMITH Chart, Frequency Response)





(d) Build

LNA Optimizer with Die Synced  (UI with synced sliders)


PromptsCont. Prompts (same as last)

(e) Build Sub 6 GHz Power Amplifier Optimizer with Die Synced

2. (a) On Claude, Run NGSpice for the LNA using given parameters. If the results do not match, explain why.

(b) Run an MNA simulation (Modified Nodal Analysis, as used in Cadence Spectre engine)

 

Optimized parameters (share)


MNA simulation


(c)  Perform DRC (Design Rule Check) for the  3.5 GHz LNA. If it is not entirely design rule compliant, redesign to make the layout more DRC driven.







(DRC, share) (DRC-driven, artifact)




3. (a) Run the Simplified LNA optimizer on Die, 3.5 GHz LNA TSMC N7. Use MNA (Modified Nodal Analysis, as used in Cadence Spectre engine)  to optimize again. 

Hint: (GD/(A*)+MNA Optimizer on Die artifactshare)


DE+MNA Optimizer on Die artifact


(b) To achieve the 15-20 dB gain target and improve the LNA performance to match the schematic specs, use Differential Evolution Optimization to optimize the LNA performance.

Hint: (DE+MNA Optimizer on Die artifact

Comparing A* & DFS/BFS

 A*/DFS

A*/DFS/BFS

2026年3月3日 星期二

JEPA Prediction for robot collision avoidance walking pedestrians

 Voronoi start-up A* + Repelling (example illustrated)





Core Navigation Repalced with JEPA + A*


HW#2 Reasoning and Search

 

建議工具

使用 Claude Sonnet 4.6 推理模式(手動切換,免費用戶額定時間內只能使用三次)

使用 ChatGPT 5 推理模式(自動切換)

使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)

使用 Grok 4 推理模式(自動切換)



How to publish a Claude artifact

How to share a ChatGPT link

How to share a Grok link

How to share Gemini Link


Content share 作業繳交格式

  • share only link, pure text, markdown (md)
  • no attachments accepted, no html, screen dump, or png
  • non-compliant homework will be rejected and returned to you


課堂練習 

Deadline: This Saturday at 23:59

Send all the share links to  me chang212@gmail.com by email with subject HW#2  [your id, your name]


1.   

Solve Pz 2 using chain of thought (CoT) reasoning with each step whether correct or incorrect, showing backtracking. 

Animate the CoT process in the above synced with river crossing scenarios.  

Animate search tree with synced progression (A*/BFS side-by-side)

 2. 

Solve Pz 20 using chain of thought (CoT) reasoning with each step whether correct or incorrect, showing backtracking. 

Animate the CoT process in the above synced with river crossing scenarios.  

Animate search tree with synced progression (A*/BFS side-by-side)

2026年2月24日 星期二

Chain of Thought (CoT)

HW#1 Use AI Professionally

 

建議工具

使用 Claude Sonnet 4.6 推理模式(手動切換,免費用戶額定時間內只能使用三次)

使用 ChatGPT 5 推理模式(自動切換)

使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)

使用 Grok 4 推理模式(自動切換)


課堂練習 

Deadline: This Saturday at 23:59

Send all the share links to  me chang212@gmail.com by email with subject HW#1  [your id, your name]

How to publish a Claude artifact

How to share a ChatGPT link

How to share a Grok link

How to share Gemini Link


Content share 作業繳交格式

  • share only link, pure text, markdown (md)
  • no attachments accepted, no html, screen dump, or png
  • non-compliant homework will be rejected and returned to you


任選一題



1. 

(a) Solve River Puzzles  (Pz 3Pz 20) by Chain of Thought and Visualize/Animate the process

(b) Verify each step not violating constraints and correctness.


2. Comparing LLMs: Claude 4.6, Gemini 3.0 Pro, ChatGPT 5.0 (三選二) for the following task.

Animate in 3d (in three.js) Yagi-Uda Antenna, AI pipe inspector, or things you choose. (擇一)

The animation must be scientifically accurate




Hints

  • Reasoning/Think/Extended Thinking Mode of the AI platform you use may be required for all the problems here
  • Pz 3 solutions on various platforms
  • Experiments on Pz 20 

River Crossing 20

 problem

animation23d

Animation of thoughts, not just solution, 2

(包含問題分析、每個嘗試的死路、突破點、逐階段的約束驗證,以及最後的最優性證明和經驗總結。)

Analysis of strategies



Part II 

2026年2月23日 星期一

AI Platforms & Variations

 


推薦 Opus 4.6 搭配思考模式(extended thinking)

我的使用體驗:

一次到位 vs. 反覆修正:這是 Opus 和 Sonnet 之間最核心的差異。Sonnet 在簡單任務上表現不錯,但面對複雜問題——像我們常做的 Three.js 地質動畫、LNA 教學投影片、A* 搜尋演算法視覺化這類高度整合的工作——Sonnet 容易在細節上出錯,需要來回修正好幾輪。Opus 因為推理深度更深,往往第一次就能把架構、邏輯、細節都處理到位。

Token 消耗的取捨:誠實地面對 Opus 單次消耗更多 token,這是因為思考模式下模型會在回應前進行更長的內部推理鏈。但從「總成本」角度來看,如果 Sonnet 要錯三次才修正到對,累積的 token 消耗和時間成本反而可能更高。這其實是一個「單次成本高但總效率高」vs.「單次便宜但需要多輪迭代」的經典取捨。

「比較一陣子才固定使用」:經過實際比較後的結論,更有參考價值。


AI Basics

 最好的程式語言 (Python? JavaScript?)

Prompt in English vs. Traditional English (中文版)

為什麼我的AI跟你不一樣: factorsbasic skills美學



supplemental


archive

2026年2月20日 星期五

2026年1月27日 星期二

學生回饋意見

 (稍微改寫去可辨視化)


2026年1月22日 星期四

健康照護科技專題 - AI 應用總覽

 

1. World Model(世界模型)- 最新 AI 研究方向

課程涵蓋 JEPA-WM(Joint Embedding Predictive Architecture World Model)與 A* Voronoi 的比較,包括 2D/3D 導航、MPC(Model Predictive Control)、以及 Block Stacking 任務規劃(TAMP vs JEPA-WM) blogspot

這是目前 AI 領域非常前沿的研究方向,JEPA 是 Yann LeCun 提出的架構!

2. 強化學習應用

PCB Layout design using RL - 使用強化學習進行電路板佈局設計,包含 2D trace、3D 視覺化、以及 hole drilling 優化 blogspot

Portfolio Mag 7 - 使用 Rule-Based Agent 和 RL Agent 進行 Magnificent 7(科技七巨頭)股票投資組合管理(2024-2025) blogspot

3. AI 財務管理

Financial analysis 系列涵蓋:基本技能、網路搜尋、Agent 代理、以及結合 RL 的投資組合管理 blogspot

4. 預防性維護

Virtual Factory 虛擬工廠模擬與 Preventive Maintenance Optimization 預防性維護優化 blogspot

5. 其他主題

  • Navigation - 導航系統
  • Carbon capture - 碳捕捉
  • River crossing puzzles - 推理問題
  • AI Engine Leap 71 - AI 引擎設計
  • 南島語族擴張視覺化 - 歷史/人文主題

部落格歷史(2013-2026)

年份文章數主要內容
2013-2017~93篇早期健康照護科技
2023-2024~35篇開始整合 AI
202547篇大量 AI 應用
2026 (1月)16篇World Model、RL、財務 AI

結論:張老師的 AI 研究深度

領域內容
World ModelJEPA-WM vs A* Voronoi
強化學習PCB 設計、股票投資、交通控制
Agent 系統Rule-Based Agent、RL Agent
工業 AI預防性維護、虛擬工廠
財務 AIMag 7 投資組合優化

張老師不只是「用 AI 工具」,他實際上在教授和研究當前最先進的 AI 技術,包括 World Model(這是 Meta AI / Yann LeCun 主推的方向)!這個部落格的內容技術深度相當高。

2026年1月3日 星期六

Navigation: Voronoi navigation with mobile obstacles

 Voronoi navigation with mobile obstacles 

path finding A*

path finding Dijkstra


Carbon capture

 Mg-MOF-74

CALF-20

River crossing Puzzle 1

     Five people - Alice, Bob, Carlos, Diana, and Elena - must cross a river in a small boat. The boat can only hold three people at a time. The following conditions apply:

* Alice refuses to be in the boat with Carlos

* Bob and Diana must travel together if either one is in the boat

* Elena can only operate the boat when Carlos is not present

* At least one person in each crossing must know how to operate the boat

* Only Bob, Carlos, and Elena know how to operate the boat

Infer an efficient plan.





update


Opus 4.6 (on the go), share


_____________________________

2026年1月2日 星期五

River crossing Puzzle 2

  渡河問題2 (statement),Problem via 許元銘











Search
 A* (share) what each node on A* tree means




Gemini 3 Pro did it by reasoning

ChatGPT non Think failed ChatGPT Think led to Python gen BFS and exe


______________________________________________________________________________
ChatGPT 5 works 4/5.  (Three claimed to use Python code execution. Therefore, is it true code was generated and then executed?) 
*1 Reasoning

*4 BFSvalidateviz,
*5 BFS



Code Gen

Use ChatGPT 5 to generate BFS search


 search that works


Use Claude to generate A*

 it takes 15 steps. (thanks to the python code generated by Claude)




another 15-step solution



yet another 15-step

one more 15-step

last 15-step


River crossing puzzle 3

     problem statement















update 

Opus 4.5 did it by reasoning, Viz  update 
Opus 4.6 A* (share)



ChatGPT non Think failed, Opus 4.5 fix it. ChatGPT Think led to Python gen and exe

Gemini 3 Pro did it by pure reasoning




_________________________________________________________________


Code gen (Python)

code gen python A* search for ::River Crossing Puzzle ...



change to js



run js



use React to visualize the steps





ChatGPT 5 create BFS to solve it. 17 step solutionvisualizevisualizer

ChatGPT BFS

ChatGPT BFS

If BFS option is trurned off, ChatGPT has a hard time doing it manually.