Monday, November 24, 2025

Google AI Studio (Playground) and App Builder

 Google has two tools to build artifacts. You have recently used the App Builder, which not only generates codes but also compile them. It takes more time to compile but less time to execute. Therefore, it is best for product delivery to  pilot production. If you are still playing with your idea and involve a lot of iterative revisions, you may want to use Playground, which does not compile the code so it will save time for waiting. 


Claude explains.

Google AI Studio (Playground)

Google AI Studio 是一個瀏覽器為基礎的環境,讓開發者可以快速測試和改進使用 Gemini 模型的提示(prompts) MediumGoogle AI。主要特點包括:

  • 快速原型開發: 這是開始使用 Gemini 建構應用最快的方式 Google AI Studio
  • 互動式測試: 可以即時測試不同類型的提示(freeform、structured、chat prompts)
  • 程式碼產生: 測試完成後可以選擇「Get code」並選擇您偏好的程式語言來使用 Gemini API
  • 適合對象: 開發者、研究人員、想要實驗 AI 功能的創作者

App Builder 工具

根據搜尋結果,Google 提供幾種不同的 App Builder:

1. AppSheet 的 Gemini for App Creation

允許您透過簡單描述業務流程或想法的自然語言來建構應用程式 Google Support。例如:

  • 描述一個管理設施檢查的應用程式
  • Gemini 會返回應用程式結構供您審查和編輯
  • 自動生成資料表、欄位和基本功能

2. Google AI Studio 的 Build Mode

Build Mode 讓您能比以往更快地添加 AI 功能,自動連接正確的模型和 API Google

主要差異總結

特點Playground (AI Studio)App Builder
用途測試和實驗 Gemini 模型建構完整的應用程式
輸出提示設計、API 程式碼可部署的應用程式
技術門檻需要基本程式設計知識可使用自然語言描述
開發方式手動編寫程式碼AI 輔助自動生成
適用場景API 整合、模型測試快速開發商業應用

簡單來說,Playground (AI Studio) 著重在「實驗和測試 AI 模型」,而 App Builder 則專注於「快速建構完整的應用程式」。您會先在 Playground 測試您的 AI 互動方式,然後可能使用 App Builder 或手動編碼來建構完整的應用。



Gemini explains.

Here is the difference between the two modes when creating an engineering animation:

1. Playground Mode (The "Prompt & Test" Phase)

This is the default interface where you chat with the AI (Gemini).

  • Goal: To test if the AI understands the physics or logic of your animation.

  • What you do: You describe the engineering concept (e.g., "Write a Python script to animate a double pendulum with chaotic motion").

  • Result: The AI gives you code snippets or text descriptions. You have to copy-paste this code into another tool (like VS Code or a Python IDE) to actually see the animation run.

  • Best for: debugging the math, refining the physics formulas, or generating the initial code logic.

2. Build Mode (The "Vibe Coding" / App Phase)

This is the new "Build" feature (often called "Vibe Coding").

  • Goal: To create a standalone, running web app that displays your animation immediately.

  • What you do: You type the same prompt (e.g., "Build a web app that simulates a hydraulic press crushing a can with adjustable pressure sliders").

  • Result: AI Studio writes the code (usually React/Javascript), compiles it, and runs it in a preview window on the right side of the screen. You see the animation working live.

  • Best for: Visualizing the final result instantly, creating interactive sliders/buttons for your engineering model, and deploying it as a shareable link.


Summary Comparison

FeaturePlayground ModeBuild Mode
Primary ActionChatting & Code GenerationApp Creation & Previewing
OutputText / Raw Code SnippetsA Functioning Web App (GUI)
VisualizationNone (must run code externally)Live Preview (runs in browser)
Engineering UseVerifying formulas/logicVisualizing the simulation

Wednesday, November 19, 2025

Build agents to work for you

 Gemini 3 Pro 

Optimize circuit design using A*

Two stage Amp


 






Make it fully optimized by A*















Using A* search to design a two stage amplifier using BJT.

Hint: A* guide  using A* search algorithm to design a two-stage BJT amplifier. This is a fascinating application of AI search to circuit design optimization!

Use A* to design a class AB amplifier. The first stage is class A. the second is class B. freq response cut off at 10Hz and 25 KHz



schematic enhanced by Gemini 3.0 Pro




artifact enhanced by Claude (share)

Tuesday, November 18, 2025

HW#9 Electronics

  AI for solving amplifier electronics (課堂講義內有參考解答)


 建議工具

使用 Claude Sonnet 4

使用 ChatGPT 5

使用 Gemini 2.5 Pro 免費額度最高 1M tokens

使用 Grok 4



 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#9  [your id, your name]

How to share a Claude link

How to share a ChatGPT link

How to share a Grok link

How to share Gemini Link


 AI for solving amplifier electronics (課堂講義內有參考解答)

任選兩題


1. Plot frequency response for the following amplifier circuit using matplotlib

Compare two methods Claude Sonnet 4.5 Freq resp (all in one) vs. staged



AI Model Results:

Note: Lower cut-off comes from coupling capacitors (not shown here).


Generate professional grade Bode plot.

Using AI tools to generate professional plot using Matplotlib


Bode plot

ChatGPT integrates Python environment so it directs generates Bode plot.










Bode plot 

1. React (Claude) - AIStudio - Python w/ Matplotlib - Colab - Plot

2. Reasoning (AIStudio) - AIStudio - Python w/ Matplotlib - Colab - Plot



2. solve for the gain


 (source)



share (Claude running Python internally on VM)






3. Solve for the gain. Plot frequency response for the following  amplifier circuit.



share (Av=-2.31, Lower Cutoff 20 Hz, Upper Cutoff 10.46MHz)

Analysis with Cπ = 22.1 pF, Cμ = 3.0 pF

20Hz is assumed, not by calculation

Analysis with Cπ = 22.1 pF, Cμ = 3.0 pF: 








Design Alternative




 以上參考答案
(1) -4  (2) -165 (3) -2.31 答案誤差10%以內都是正常的


Friday, November 14, 2025

Simulating Random Process

Visualization of Manufacturing Process,  simulation of pipelines (inspired by Edward Chang)


Drone deployment using React


Submarine sonar detection 




simulate submarine detection with random sensor arrays
Case study teaching guide


simulation, artifact share



Wednesday, November 12, 2025

Claude Code 使用文字介面而非 IDE

 Claude Code 使用文字介面而非 IDE 有幾個策略性原因:

速度與效率 - CLI 工具可以立即啟動,在你已經在工作的終端機中運行,不需要啟動圖形化 IDE 的額外負擔。這讓快速委派任務或迭代程式碼變更更加快速。

通用整合性 - 它可以無縫整合到你現有的任何開發環境中。你可以繼續使用你喜歡的編輯器(VS Code、Vim、Emacs 等),同時讓 Claude Code 在背景處理特定的編碼任務。它不會強迫你切換環境。

符合命令列工作流程 - 許多開發者,特別是從事後端、DevOps 或複雜系統的人,大部分時間都在終端機中工作。Claude Code 可以自然地融入現有的工作流程,與 git、套件管理器、建置工具和部署腳本搭配使用。

專注的委派模式 - CLI 介面強化了「給 Claude 一個特定任務讓它執行」的範式,而非持續來回互動。你描述你想要什麼,Claude Code 自主執行,然後回傳結果。這與聊天介面的對話式、探索性質不同。

資源效率 - 文字介面非常輕量。不需要渲染複雜的 UI 元素,沒有 Electron 框架的負擔——只有你和 AI 編碼系統之間快速、直接的溝通。

可腳本化與自動化 - CLI 工具可以輕鬆整合到腳本、CI/CD 流程或自動化工作流程中,這是圖形化 IDE 無法做到的。

設計理念似乎是:讓開發者使用他們喜愛的工具,並將 Claude 作為強大的命令列助手加入,而不是試圖取代整個開發環境。

Saturday, November 08, 2025

Follow up to #8

 1. Gemini 2.5 Pro for 2-stage amp, Freq Response





staged calculation leads to better, more accurate results.



2. 繪圖,動畫:  建議 使用 Claude Sonnet 4.5

推理,複雜問題:  建議使用以下 

使用 Claude Sonnet 4.5 extended thinking 付費版

使用 Gemini 2.5 Pro 免費額度最高 1M tokens

使用 ChatGPT 5

使用 Grok 4




Monday, November 03, 2025

ASML

ASML High-NA EUV Lithography: The Mechanical Marvel Atomic-precision motion control + 40 years of mechanical engineering mastery 🏭 CLASS 1 CLEANROOM - Everything inside costs $$$$ EUV Source (LPP) CO₂ Tin droplets → Plasma @ 500,000K → 13.5nm EUV Collector Mirror M1 M2 Complex Mirror Optics 6-14 mirrors, each with 40+ layers (0.1nm precision) Zeiss exclusive, $100M+ per set Reticle Stage (4× pattern) Reticle Mask Pattern Synchronized Stage (4:1 ratio) Projection Optics (0.55 NA) 4× demagnification ⚙️ WAFER STAGE: The Real Moat (40 Years of R&D) 300mm Wafer 6-DOF Motion Control x, y, z, Rx, Ry, Rz - all synchronized 4G acceleration @ atomic precision < 0.1nm positioning < 2nm overlay Laser Interferometers X-axis position Y-axis position Z-axis position Real-Time Control 1000+ corrections/sec Temp: 0.001°C Vib. isolation SYNC ASML's 40-Year Moat: Key Technologies 1. Motion Control ✓ Air bearing stages ✓ Magnetic levitation ✓ Voice coil actuators ✓ Piezo fine positioning ✓ 6-DOF simultaneous Dev time: 15+ years Japanese precision + Dutch integration Nikon/Canon couldn't replicate 2. Metrology ✓ Laser interferometers ✓ Encoder systems ✓ Alignment sensors ✓ Real-time correction ✓ Sub-nm accuracy 1000+ measurements/sec Proprietary algorithms Hardware + software IP 3. Synchronization ✓ Reticle + Wafer sync ✓ 4:1 speed ratio ✓ Overlay < 2nm ✓ Dynamic focus ✓ Vibration damping 4G acceleration sync'd Dual-stage architecture 200 wafers/hour throughput 4. Supply Chain ✓ Zeiss mirrors (exclusive) ✓ Japanese bearings ✓ US control systems ✓ Dutch integration ✓ 1000+ suppliers Ecosystem depth 40 years of relationships Cannot be replicated quickly This is why Intel with $100B+ budget still can't catch up. This is why Substrate's "solve the light source" strategy misses 99% of the challenge.

Substrate

Substrate XRL Mechanism (Animated) Watch: Electrons → X-rays → Pattern Transfer (with motion control challenge) 💰 CHEAP LAND (Outside Cleanroom) Particle Accelerator (Synchrotron / Linear) X Target ⚡ Electron Acceleration • Energy: ~GeV scale • Beam current: mA range • Pulse rate: MHz • X-ray flux: High ✓ Large size OK - no cleanroom needed here Cost savings: Build on cheap industrial land ~$40M per tool (estimated) X-rays (~1nm) Vacuum beam transport 🏭 EXPENSIVE CLEANROOM XRL Exposure System X-ray Mask ↕ Proximity gap (~10-50μm) Silicon Wafer + Photoresist ⚙️ Precision Wafer Stage 6-DOF motion control required ±0.1nm needed! ⚠️ THE CHALLENGE Atomic-level precision motion control: • <0.1nm positioning • Real-time correction • Overlay accuracy Process Flow: How XRL Works 1️⃣ Accelerate Electrons accelerated to GeV energies in particle accelerator ✓ Solved (National lab tech) 2️⃣ Generate e⁻ hits target → X-rays produced via Bremsstrahlung ✓ Solved (Physics understood) 3️⃣ Transport X-rays travel through vacuum tube to lithography chamber ✓ Solved (Standard engineering) 4️⃣ Pattern X-rays pass through mask, expose wafer (proximity printing) ⚠️ Partial (Alignment needed) 5️⃣ Control Nanometer motion control & alignment across full wafer ❌ UNSOLVED (The moat!) Animation shows: electrons → X-rays → patterned exposure → moving wafer stage (emphasizing control challenge)