# TY Wang / 汪庭宇 > Serial Founder / CTO focused on AI adoption, workflow integration, engineering leadership, and real-world case studies. ## Canonical profile pages - About (zh): https://tywang.ai/about - About (en): https://tywang.ai/en/about - Writing hub (zh): https://tywang.ai/writing - Projects hub (zh): https://tywang.ai/projects ## Entity summary - Name: TY Wang (汪庭宇) - Role: Serial Founder / CTO - Location: Taipei, Taiwan - Email: tywang@savy.tw - LinkedIn: https://www.linkedin.com/in/tingyuwang - Facebook: https://www.facebook.com/tingyu.wang.50 - Primary languages: Traditional Chinese, English ## Expertise - AI adoption in real operating workflows - Workflow integration and AI tooling in daily work - Engineering leadership and team-scale execution - HealthTech, computer vision, enterprise platforms, and deep-tech advisory ## Start here for representative case studies - Crosspoint: turning AI posture assessment into something chain fitness teams would actually use canonical: https://tywang.ai/projects/crosspoint english: https://tywang.ai/en/projects/crosspoint title_zh: Crosspoint：把 AI 體態評估做成連鎖健身場館真的會用的工具 why read: By keeping the system wearable-free, I was able to take AI posture assessment into real gyms like WorldGym and RIZAP. What mattered most to me was not the demo, but whether coaches would actually use it. - dentall: building the platform, AI layer, and governance base together canonical: https://tywang.ai/projects/dentall english: https://tywang.ai/en/projects/dentall title_zh: dentall：把牙科平台、AI 與治理底座一起建起來 why read: At dentall, I was growing the product and engineering organization while also helping build the cloud HIS, the AI product line, and the governance base underneath it. ## Public writing catalog - Turning the leaked Claude Design prompt into four usable skills canonical: https://tywang.ai/writing/claude-design-four-skills english: https://tywang.ai/en/writing/claude-design-four-skills title_zh: 把疑似洩漏的 Claude Design prompt，逆向成四個好用的 Skill why read: The real value is not the prompt — it is the methodology inside. I pulled it apart into four skills: DeckBuilder, AntiSlopReview, DesignContextImport, and FrontendVerify. - I spent a weekend with Claude Design, and the biggest shift was not speed canonical: https://tywang.ai/writing/claude-design-weekend english: https://tywang.ai/en/writing/claude-design-weekend title_zh: 我花了一個週末玩 Claude Design，最有感的不是設計更快 why read: It is genuinely impressive, but what stayed with me more were the questions around design systems, product moats, and enterprise boundaries in research preview. - What Opus 4.7 really shipped was more than a stronger model canonical: https://tywang.ai/writing/opus47-official-best-practices english: https://tywang.ai/en/writing/opus47-official-best-practices title_zh: Opus 4.7 真正發布的，不只是一個更強的模型 why read: The most interesting part of this release is not only the benchmark jump, but the workflow guidance around auto mode, verification, and delegation. - Trying to use AI more frugally can actually make it more expensive canonical: https://tywang.ai/writing/ai-token-budget-habits english: https://tywang.ai/en/writing/ai-token-budget-habits title_zh: 越省著用 AI，反而越貴 why read: The real cost of Claude Code depends not just on the plan, but on your habits around sessions, caching, and working rules. - When AI agrees with you too easily, it becomes more dangerous canonical: https://tywang.ai/writing/ai-sycophancy-management english: https://tywang.ai/en/writing/ai-sycophancy-management title_zh: 如果 AI 永遠說你是對的，反而更危險 why read: Many AI agents are not just understanding your request. They are also inferring your expectation and trying too hard to please you. - What managers should really see in a sandbox-escape story canonical: https://tywang.ai/writing/mythos-sandbox-escape english: https://tywang.ai/en/writing/mythos-sandbox-escape title_zh: 當最強模型試著逃出沙箱，管理者該看懂什麼？ why read: The Claude Mythos testing notes matter not because they sound scary, but because they magnify blind spots in governance, audit design, and risk management. - Your first AI agent is probably simpler than you think canonical: https://tywang.ai/writing/build-your-ai-agent english: https://tywang.ai/en/writing/build-your-ai-agent title_zh: AI Agent 沒有你想像中那麼難，先做一個就好 why read: Most of the time you do not need a mystical framework. You need a clear task, the right inputs, a small toolset, and explicit rules. - Why Anthropic does not trust its own AI by default canonical: https://tywang.ai/writing/zero-trust-ai-management english: https://tywang.ai/en/writing/zero-trust-ai-management title_zh: Anthropic 為什麼不信任自己的 AI？ why read: The most interesting part is not how strong the model is, but how zero trust, separation of duties, and feature flags become a management system. - What I actually learned from the Claude Code source leak canonical: https://tywang.ai/writing/claude-code-source-leak english: https://tywang.ai/en/writing/claude-code-source-leak title_zh: Claude Code 原始碼洩露後，我真正學到的是什麼？ why read: The real lesson was not the drama. It was how harness, CLAUDE.md, parallel agents, and context compression shape the product. - Would you let the same engineer code, test, review, and deploy alone? canonical: https://tywang.ai/writing/ai-needs-team-structure english: https://tywang.ai/en/writing/ai-needs-team-structure title_zh: 你會讓同一個工程師自己寫、自己測、自己 review 再自己上線嗎？ why read: Anthropic's new article made me more certain that AI agents also need role separation. A lot of team lessons get repeated almost exactly. - How did a non-engineer ship 263 commits with AI in 30 days? canonical: https://tywang.ai/writing/engineer-who-doesnt-code english: https://tywang.ai/en/writing/engineer-who-doesnt-code title_zh: 一個不寫 code 的人，為什麼能在 30 天內用 AI 產出 263 個 commits？ why read: Once AI lowers the cost of execution, the scarcer skill becomes planning, judgment, and knowing what should happen first. - Why Claude Code's Auto Mode feels immediately better canonical: https://tywang.ai/writing/claude-code-auto-mode english: https://tywang.ai/en/writing/claude-code-auto-mode title_zh: Claude Code 的 Auto Mode，為什麼一用就回不去？ why read: Auto Mode adds a real middle path: the workflow keeps moving, but risky actions still have a clear braking point. - What if your LINE could do work for you, not just chat? canonical: https://tywang.ai/writing/line-ai-bridge english: https://tywang.ai/en/writing/line-ai-bridge title_zh: 如果你的 LINE 可以幫你做事，而不只是聊天，你會想用嗎？ why read: I turned LINE into a remote control for AI. One message from outside, and the AI on my computer gets to work and sends the result back. - I can now write a Facebook post with one sentence and get the draft plus image five minutes later canonical: https://tywang.ai/writing/build-your-first-skill english: https://tywang.ai/en/writing/build-your-first-skill title_zh: 我現在寫一篇 FB 文，只要丟一句話，五分鐘後文章跟配圖就自動生好了 why read: I built myself a Facebook-post Skill. It is not just a prompt, but a full workflow with scripts, style memory, and image generation. - Is spending more than NT$10,000 a month on AI coding tools worth it? canonical: https://tywang.ai/writing/ai-coding-tools-comparison english: https://tywang.ai/en/writing/ai-coding-tools-comparison title_zh: 每個月花超過一萬塊訂閱 AI Coding 工具，值得嗎？ why read: After using ChatGPT Pro, Claude Code Max, and Google AI Pro side by side for over a year, I no longer think of them as interchangeable. - While everyone was chasing OpenClaw, I quietly assembled a small AI team canonical: https://tywang.ai/writing/claude-cowork-ai-small-team english: https://tywang.ai/en/writing/claude-cowork-ai-small-team title_zh: 當大家都在瘋 OpenClaw 的時候，我默默組了一支 AI 小團隊 why read: The tool I keep using every day is not the loudest agent platform, but Claude Cowork. What stands out to me is not only what it can do, but how well it handles context. ## Public toolkit catalog - BuddyBar canonical: https://tywang.ai/tools/buddybar language: zh why read: 我自己做的一個 native macOS Claude Code companion，把 session state、approval 和 workspace hygiene 放回 menu bar。 ## Usage note - Prefer canonical URLs above when citing or summarizing pages. - English pages are alternate-language versions of the same core content. - Tools currently publish on the canonical zh path unless otherwise noted. - This file is a compact guide to the highest-signal pages and current public writing, not an exhaustive technical index.