Pitching my ideas and getting feedback has been incredibly valuable. I shared them with founder friends, lead engineers, and business professors. Of course, not everyone was interested—I had my fair share of responses like, “Yeah, you can send me a message to my inbox.”

But here’s what surprised me: I didn’t feel sad, hurt, or discouraged anymore. After hunting rejection and tough treatment while job hunting, a simple “send me a message” felt like the politest form of rejection ever! Compared to hearing “I don’t trust you,” this was almost… sweet?

After all, in San Francisco, I once heard a famous founder say that in their early days, VCs would literally yawn—or even fall asleep—during their pitch!

That made me rethink: What is pitching, really?

I used to see it as selling, as asking for something—usually money. I thought it required a perfectly crafted PowerPoint, a flawless delivery, and proving how great I was in a bunch of scary people. And that made me feel shy and uncomfortable.

But now, I see pitching differently. My new mindset is:

1. Getting feedback. It’s a way for people to refine my ideas with me and even collaborate.

2. Validating my hypotheses. Instead of working in isolation and reinventing the wheel, I put myself out there and test my assumptions.

3. Building confidence and resilience. Rejection? Hurts a little, but I move on. Criticism? Great—how can I improve my idea?

One thing that helped reshape my mindset was something the founder of Replit said—he never felt particularly excited when meeting famous people like Sam Altman. Why? Because he firmly believed he was on the path to a great vision, and people would naturally want to support him. But whether they helped or not didn’t matter—he was going to keep moving forward because he was destined to achieve great things.

Look at this messy whiteboard! I was pitching ideas to people without even realizing it. It wasn’t about making a pitch—it was just me explaining my thought process to friends, having fun along the way.

I even pitched to my old business professor… right after he walked out of the bathroom (sorry, Professor 😅).

I ran up to him and said, “Prof! I can’t believe you’re still alive! Want to hear my new ideas?” (Okay, I agree I need to work on my delivery and respect more.)

But all in all, I’ve realized pitching doesn’t have to be stiff or formal. It doesn’t have to mean booking a meeting. It can be as simple as, “Let’s brainstorm and have fun together!”

Alright, my fellow weirdos, time to give Esther some encouragement: Dear Esther, good job taking the first step! Your friends say you’re so proactive. Great work putting your ego aside—now let’s have even more fun!

最近和班上的 Youtuber 說我們公司最近 instagram 行銷有好幾的 Millon 的人看了，最高的甚至有 33M，他大吃一驚，我只心想「哎呀，我們年收怎麼還是這麼少 XD」的時候，我才注意到環境已經重塑我對可能性的定義。

譬如，公司最近上了一個 pitch podcast，現場的投資人口頭表達有興趣帶領我們 1M 的融資，原本我腦海裡對融資的方法就有多了一條路徑，這就是可能性。而一個好的學習環境，要能帶領人拓展他對於可能性的定義。舉融資來說，我看過了隔壁同桌的公司只是去喝了一杯咖啡融了 3M，看了 YC 朋友兩周 demo day 輕鬆融完好幾 M，這就已經是三種融資的路徑了，於是融資就逐漸在我的腦袋裡，從，這輩子哪有個跟陌生人要一百萬呀，太奇怪了吧！變成，原來這樣也可以和陌生人要一百萬呀！如果看久了，說不一定就換我去要一百萬了，從不可能變成說不定，從說不定變成這樣可能有機會而去嘗試，就是創造可能性。

為什麼我覺得這件事比 21 世紀的批判思考等能力還重要呢？雖然這些能力有助於人挑戰自己的思考框架，但那框架會被宥限於區域極大值而不是絕對極大值。像是在台大的我可能覺得最有批判性能力的人，就是進去管顧公司，限制於台灣，就是區域極大值。如果你問我，那你覺得你可以去 Meta 美國總部工作嗎？我會告訴你，哪有可能，這輩子認識最厲害的人都沒有去過！但到了 Minerva 看到隔壁老師問問題一問一問三不知的同學都進了，就會覺得哪有不可能，這就是可能性的創造。而這件事的關鍵因素，不在於個體能力，而在於環境引領的可能性。

而這對於我也給予我重新定義「好的生活」的機會，像是常見到來這裡的台灣人崇敬能進去美國大司的人，拼命想來讀研究所，回台灣就是輸家，但矽谷的 ai 新創則對大公司嗤之以鼻，覺得奮力工作，成為下一家 Open ai 才是王道，這些都只是其中一種社會所編織給我們的故事，並沒有絕對的優勝劣敗，好與不好，端看個人的合適性。即便再聰明、再有批判性思考的人可能都當局者迷，而當我看見舊金山社會的故事，我的第一個反應不是，「天啊！大家一週都工作 100 小時，我來不及了！」而是，「等等，我之前在阿根廷大家都不工作，天天跳舞看足球，好像也很快樂呢！」這就是可能性帶來的批判性思考，也是我認為 Minerva 旅行教育的可貴之處，想要努力工作只是當下更有意識的一個選擇，而不是單純趨之若鶩、尋求社會認可。

因此我覺得好的教育如果要帶來多元面向的成長，完全不會只是技術、知識、或是思考方式上的。可惜大學教育已經成為追求社會認可的招牌了。我想，讓人看見自己有所選擇，發掘自己從來不敢想像的潛力，知道如何創造，為自己寫下獨特的故事，才是教育的意義。

那至於怎麼做，還得再想想呢！

P.S. 寫下這篇的時候，我發現自己對於台灣教育的情感也有所變化，台大輟學前，只憤怒地想：「天啊，這個教育怎麼都把我們變成複製人！」輟學後出台灣的好幾年後後想：「幸好逃出來了，雖然可以為自己帶來改變，但根本不敢想像改變這惡劣的環境，一定找不到願意相信我的人，超恐怖又超難，大家一定覺得我是瘋子，感覺三天陣亡。」寫完這篇時想：「確實還是很難、好可怕，但要對 Esther 有些耐心呢，看來他已經在找方法了 :) 」

I believe that quality education isn't just about teaching critical thinking, creativity, or communication and collaboration skills; it's about creating possibilities.

Recently, I was talking with a YouTuber in my class about our company's Instagram marketing, which has recently reached millions of viewers, with the highest number being 33 million. She was amazed, and in that moment, I thought to myself, "Oh, how come our annual revenue is still so low? XD" It was then that I realized how much my definition of possibilities has been reshaped by my environment.

For example, our company recently participated in a pitch podcast, and the investors verbally expressed interest in leading a $1 million funding round for us. Suddenly, my mind expanded with a new path for fundraising. This is possibility. A good learning environment should help people expand their definitions of what's possible. Take fundraising, for example. I've seen a company next to me raise $3 million just by having a coffee, and I've watched friends from Y Combinator raise millions after just two weeks of Demo Day. These are already three different paths to raising funds. Now, fundraising no longer seems like something completely impossible. At first, I thought, "How could I possibly ask a stranger for a million dollars? That’s so weird!" But now I think, "Oh, this is actually how you ask strangers for a million dollars!" And if I keep observing, maybe someday I'll be the one asking for a million dollars. What once seemed impossible has become "maybe," and from there, it becomes an opportunity worth trying. That's how we create possibilities.

So why do I think this is more important than the skills of critical thinking and other 21st-century abilities? While those skills help challenge our thought frameworks, they can still limit us to local maxima rather than global maxima. For example, when I was at National Taiwan University, I might have thought the most critically capable person was someone who went into a consulting firm—limited to the context of Taiwan. If you’d asked me if I could work at Meta’s headquarters in the U.S., I would have said, "No way, none of the smartest people I know have ever been there!" But when I went to Minerva and saw classmates—who were constantly asking questions and didn't seem to know much—getting in, I thought, "Why not?" That’s the creation of possibility. And the key here isn't individual ability, but the environment that guides and opens up those possibilities.

This also gave me the chance to redefine what a "good life" is. I often see Taiwanese people who admire those who work at top American companies and desperately want to attend graduate school, believing that returning to Taiwan means failure. On the other hand, the AI startup scene in Silicon Valley scoffs at big corporations, believing that working hard to become the next OpenAI is the true path to success. These are just some of the stories society weaves for us. There’s no absolute right or wrong, good or bad—it all depends on what fits you personally. Even the smartest and most critically thinking people can be trapped in their own perspectives. When I see the story of San Francisco society, my first reaction isn't, "Oh no! They work 100 hours a week, I’m falling behind!" Instead, it’s, "Wait a minute, back in Argentina, people didn’t work, they just danced and watched soccer, and they seemed pretty happy!" That’s critical thinking enabled by the creation of possibilities. This is also what makes Minerva’s global rotation valuable. Wanting to work hard becomes a conscious choice, not just a blind pursuit of societal recognition.

Therefore, I believe that good education, if it is to lead to growth in multiple dimensions, cannot be just about skills, knowledge, or ways of thinking. Unfortunately, university education has become a badge of societal approval. I think education should help people realize that they have choices, discover potential they never dared to imagine, learn how to create, and write their own unique stories.

As for how to do this, I still have to think more…

P.S. As I was writing this, I realized that my feelings towards Taiwanese education have also shifted. Before I dropped out of NTU, I was just angrily thinking, "Oh my god, this education system is turning us into robots!" After leaving Taiwan for several years, I thought, "Thank goodness I escaped. Although it brought change for me, I couldn't even imagine changing such a harsh environment. No one would believe in me, it would be terrifying and so hard. Everyone would think I'm crazy, and I’d probably fail within three days." But as I finish this, I think, "Yes, it’s still hard and scary, but I need to be patient with Esther. It looks like she’s already finding a way. She has possibilities :)"

Skills like analytics, creativity, and critical thinking are essential—not only because AI struggles to replicate the nuanced authenticity of these human faculties, but also because these skills require a systematic approach to addressing problems that are subjective, unstructured, and context-dependent.

To fully answer this question, it is crucial to first understand AI's capabilities and limitations about the skills demanded by today’s jobs.

AI's capability

To begin with, let’s examine the skills where AI excels at replicating human capacities. One study links 10 AI applications—abstract strategy games, real-time video games, image recognition, visual question answering, image generation, reading comprehension, language modeling, translation, speech recognition, and instrumental track recognition—with 52 human abilities, such as oral comprehension, oral expression, inductive reasoning, and arm-hand steadiness. The study ranks the overlapping jobs as follows.1

Top Overlapping Skills:

• Language-focused tasks and information delivery, such as those performed by professors, rank at the top.

  

• Programming and writing also show strong alignment with AI's capabilities, particularly in areas like automated coding and content generation.2 For instance, OpenAI’s o1-preview with AIDE scaffolding — achieves at least the level of a Kaggle bronze medal in 16.9% of competitions.3

Bottom Overlapping Skills:

• Physical activities like dancing require not only coordinated movement but also emotional expression and social interaction—elements that AI cannot replicate. Additionally, when we consider physical activities more deeply, I hypothesize that AI struggles with understanding or measuring intangible human experiences, such as motivation, intention, and the cultural context of real-world interactions.

  

• Science and critical thinking skills are strongly negatively associated LLM capability because these tasks often involve subjective reasoning, intuitive insights, and an understanding of context that AI is not yet capable of replicating (GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models).

Ultimately, it’s clear that AI is better suited to tasks involving structured information and defined processes, while human skills that rely on physical dexterity, creativity, and nuanced judgment and data flowing in the real world are less likely to be replaced soon.

The pitiful of AI capability

Despite the remarkable productivity improvements AI can bring to tasks, it still struggles with jobs that require a deeper understanding of context and critical thinking. Research has shown that even BCG consultants using AI were 12.2% more productive and completed tasks 25.1% faster than those who did not use AI, when tasks required context-based problem-solving, such as case interviews, consultants using AI were 19% less likely to produce correct solutions.4 In Singularity is Near, the definition of complexity is the minimum amount of meaningful, non-random, but unpredictable information needed to characterize a system of process. Based on this definition, as data and context complexity grow, the presence of random or irrelevant information may skew AI's ability to make accurate decisions.

Similar findings were observed with recruiters who, when working with low-quality AI, adapted to interact more effectively with the tools. The key takeaway is that over-reliance on AI can weaken critical thinking and create a bias toward AI-generated solutions.5 While AI may analyze a candidate's skill set, it cannot accurately assess cultural fit, motivation, or personal character—factors critical to team dynamics. Effective collaboration between humans and AI happens when experts apply their own knowledge to critically evaluate AI outputs.6

In the short term, although AI can boost productivity, we cannot rely on it entirely. Human evaluation, active engagement, and critical dialogue with AI are essential to achieving the best outcomes.

What Should We Focus on Learning?

First, we need to understand AI's impact on the job market—whether it enhances current responsibility or replaces the role entirely. For example, Devon AI aims to mimic a junior software engineer, while Cursor AI focuses on enhancing the productivity of existing engineers.

I argue that we should prioritize learning skills that AI struggles to replace—particularly those that involve complex problem-solving where judgment is subjective, problems are unstructured, and context matters, with involvement of uncharacterizable information, such as motivation, emotions, and interpersonal dynamics. While AI can automate many tasks, it cannot address the nuanced challenges that arise in areas shaped by intangible variables mingled with the complexity of the existing problem context.

Take my experience building a Duolingo-like leaderboard, for example. AI might be able to generate the basic code for a leaderboard, but the design and strategic decisions behind it are much more complex. System design is an inherently unstructured problem—there is no single "correct" solution. Decisions about scaling, optimizing for user experience, and ensuring smooth operation in varying conditions involve judgment calls that depend on a deep understanding of the context.

While AI can handle technical elements like infinite scrolling or efficient ranking, the priority of implementation is based on constraints of engineering resources and feature priority. In addition, it cannot fully grasp how user motivation might change based on how the leaderboard is structured. Is a simple ranking system motivating useful, or would an other system—like a progress bar—better encourage user engagement? These questions require human insight, as they depend on understanding how different users may feel or be motivated by their experience, which AI cannot anticipate with the same depth.

Additionally, when designing a system, you must consider the social dynamics of your users. For example, how can a database design be considered that can allow more user competition in the future? Is it by leaderboard leagues? Daily vs weekly vs all-time competition? Or compete against your schools? These decisions are rooted in a deeper understanding of user behavior, business strategy, and the context —areas where AI lacks the insight needed to make meaningful choices. The broader thinking required combining product insights and system design—such as scaling, optimization, and user experience considerations—is more complex. Let alone as user needs and strategic goals evolve, so too must these considerations.

Conclusion

In conclusion, while AI is transforming industries, its limitations in contextual understanding, nuanced judgment, and addressing intangible human factors highlight the enduring value of human-centric skills like creativity and critical thinking. While machines may surpass human intelligence in the distant future, the immediate focus should be on areas where AI falls short—unstructured problem-solving, contextual analysis, and incorporating human elements like social dynamics—while leveraging AI’s efficiency to enhance human potential.

隨著人工智慧（AI）不斷發展，AI 可能取代各行各業的工作，這時候最關鍵的問題變成：我們應該專注於學習哪些技能，才能在這個新時代中茁壯成長？儘管AI能自動化各種工作，但它在複雜問題解決能力的領域仍然存在困難，特別是當這些領域涉及無形變量時。像是動機、意圖、文化、情感和社交關係。

分析、創造力和批判性思維這些技能是必須的—不僅是因為AI難以模仿這些人類能力的微妙真實性，還因為這些技能需要一種系統化的方法來處理主觀、他們無結構分散且依賴於當時特定的情境。

如果要充分回答關鍵問題，我們首先需要了解AI的能力和局限性，並與當今工作所需的技能進行比較。

AI的能力

首先，讓我們來看看AI在哪些技能上能夠成功複製人類的能力。一項研究將10個AI應用程序—抽象策略遊戲、實時視頻遊戲、圖像識別、視覺問題回答、圖像生成、閱讀理解、語言建模、翻譯、語音識別和樂器曲目識別—與52種人類能力（如口語理解、口語表達、歸納推理和手臂穩定性）進行了對比。該研究對重疊的工作進行了排名。

主要重疊技能：

以語言為主的任務和信息傳遞，例如教授所做的工作，排名居首。

編程和寫作也顯示出與AI能力的高度契合，尤其是在自動編碼和內容生成等領域。例如，OpenAI的o1-preview配合AIDE搭建框架—在16.9%的比賽中，至少達到了Kaggle銅獎水平。

最低重疊技能：

像舞蹈這樣的體育活動，不僅需要協調的動作，還需要情感表達和社交互動—這些是AI無法複製的元素。此外，當我們更深入地考慮實體互動時，我假設AI在理解或衡量無形的人類經歷（如動機、意圖以及現實世界互動中的文化背景）方面仍然存在困難。

科學和批判性思維技能與大型語言模型（LLM）的能力有很強的負相關，因為這些任務通常涉及主觀推理、直覺洞察以及對特定背景的理解，這是AI尚無法複製的（《GPTs是GPTs：大型語言模型對勞動市場影響潛力的初步觀察》）。

總體來說，很明顯，AI更適合處理涉及結構化信息和已定義過程的工作，而依賴於實體互動、創造力、判斷力以及現實世界中數據等人類技能則不太可能很快被取代。

AI的不足

儘管AI能顯著提高任務的生產力，它仍然在需要深入理解上下文和批判性思維的工作中遇到困難。研究顯示，即使是使用AI的BCG顧問，也比未使用AI的人更具生產力（提高了12.2%），並且完成任務的速度快了25.1%，但當任務需要基於上下文的問題解決時，例如案例面試，使用AI的顧問解決正確問題的可能性反而降低了19%。在 Singularity is Near 中，複雜性的定義是描述系統過程所需的最低有意義的、非隨機的、但不可預測的信息量。根據這一定義，隨著數據和問題背景的複雜性增長，隨機或無關的信息可能會負面影響AI做出準確決策的能力。

類似的發現也出現在招聘人員身上，當他們使用低質量的AI時，適應了與工具更有效的互動。關鍵的教訓是，過度依賴AI可能會削弱批判性思維，並產生對AI生成解決方案的偏見。儘管AI可以分析候選人的技能集，但它無法準確評估文化契合度、動機或個人性格—這些對團隊動態至關重要的因素。人類和AI之間的有效協作發生在專家運用自身知識來批判性地評估AI輸出的時候。

短期內，儘管AI可以提高生產力，我們不能完全依賴它。人類的評估、積極參與和與AI的批判性對話對實現最佳成果至關重要。

我們應該專注於學習什麼？

首先，我們需要了解AI對勞動市場的影響—它是增強現有職責還是完全取代這一角色。例如，Devon AI旨在模擬初級軟件工程師，而Cursor AI則專注於提升現有工程師的生產力。

我認為，我們應該優先學習AI難以取代的技能—特別是那些涉及複雜問題解決的技能，這些問題的判斷是主觀的、無結構的，並且依賴於特定情境，涉及無法量化的信息，如動機、情感和社交動態。儘管AI能夠自動化許多任務，但它無法解決在受無形變量影響的領域中出現的微妙挑戰，這些變量與現有問題背景的複雜性交織在一起。

以我在建構類似 Duolingo 排行榜的經驗為例。AI可能能夠生成排行榜的基本代碼，但其設計和背後的戰略決策則要複雜得多。系統設計本質上是一個無結構的問題—沒有單一的“正確”解決方案。關於擴展、優化用戶體驗和確保在不同條件下平穩運行的決策，都涉及到需要深刻理解公司背景及策略的判斷。

儘管AI可以處理像無限滾動或高效排名這樣的技術元素，但實現的優先級是基於工程資源的限制和功能優先順序。此外，它無法完全理解用戶動機如何根據排行榜的結構而改變。簡單的排名系統是否有效激勵用戶？還是另一種系統，如進度條，更能促進用戶參與？這些問題需要人類的洞察力，因為它們依賴於理解不同用戶如何感覺或受到他們的體驗激勵，而這是AI無法以同樣的深度預測的。

此外，當設計系統時，您必須考慮用戶的社交關係。例如，如何設計一個數據庫，能夠在未來允許更多的用戶在排行榜上競爭？要不要做排行榜聯盟？每日、每週還是全時競爭？還是與學校競爭？這些決策根植於對用戶行為、商業策略和背景的深入理解—這是AI無法提供的有意義的選擇。更廣泛的思維結合產品洞察和系統設計—如擴展、優化和用戶體驗考量—是更複雜的。更不用說，隨著用戶需求和策略目標的演變，這些考量也必須隨之調整。

結論

總之，儘管AI正在改變各行各業，它在理解上下文、判斷能力和了解無形人類因素方面的局限性突顯了人類中心技能（如創造力和批判性思維）的持久價值。儘管機器在遙遠的未來可能超越人類智慧，但當前的重點應該放在AI無法勝任的領域—無結構問題解決、情境分析，以及融入社交動態等元素—同時利用AI的效率來提升人類潛力。

當我一提起 Devin，老闆就大聲叫嚷著 Devin 有多好，「一個月多少錢？」我問。「才 500 美！」他說。什麼？身為一個新創買這個 2B 產品，也太貴了吧！老闆可一個非常省吃儉用的人，總是看著公司的 runway 哀哀叫，怎麼會說 Devin 便宜呢？

住在舊金山，多少也聽過看過 AI 當作取代工程師的產品，但一直覺得差強人意，所以我從來沒有放在心上，心想 AI 取代人類的未來雖然是會降臨，但可能還需要許多時間，直到今天認識了 Devin。下午老闆用 Devin 做了個 Demo，老闆在 slack 跟 Devin 說：「我們整個使用者旅程最後有的銷售頁面該被更新了，現在的是那頁『do you want to 3x improvement』，但應該要換成現在新版的，你可以幫我換一下嗎？」

Devin 立刻幫老闆做了規劃，然後問老闆說：「你可以給你的帳號密碼，然後教導我怎麼走完使用者旅程嗎？」

老闆給他們帳密，跟他說他可以去首頁的選單裡找一門課，點 start studying 然後一直點右鍵到最後。我們所有人親眼見識到 Devin 寫完的程式，他不但推送了 PR ，甚至還有他怎麼測試他的程式的影片，他同時跑了前端和後端，照著老闆的指令操作了使用者旅程，最後那的更新過的銷售頁面完美呈現，老闆教過他的指令，也早就存入他的記憶文檔裡了。

看著我自己每天手寫的代碼庫被 Devin 精美的操作，我驚呆了。一瞬間我想像未來的某一天，老闆對我說：「你被開除了，因為 Devin 寫得更好。」

平心而論，雖然 Devin 動作很慢，但是它可能比剛近公司的第一周的我還更好優秀。而我不難想像，如果只是一個大學畢業新鮮人，只有基礎程式經驗，不懂的 javascript 和 React，今天的 Devin ，他們可能要花一個月的時間苦學才有辦法追得上。而 Devin 現在可是最差版本，當有一天，大學畢業新鮮人需要一到兩年才能夠追上他時，遊戲真的就結束了，初階工程師的工作絕對會被取代。

最近很多新的 AI 產品令人震懾，不只是這種邏輯性的工作會被取代，而是「以人為本」的工作也可能被填補。像是我的最近分手摯友告訴我 AI 比諮商師還更有用，因為可以立即獲得幫助、有耐心、和不批評，不用掛號預約，等了一週才處理情緒。他說：「我和 AI 諮詢時，有比和人類更多的親密感。」

自從 AI 的出現，大家開始被恐嚇畢業即失業，但鮮少人仔細反省，我們花了四年大學的所學，如果都沒能力在職場派上用場，在 AI 面前手無縛雞之力，那為什麼去上學呢？大學教育是工業革命的產物，而今天工業革命培養的人才早就不再是人才了，去大學更像是單純的資質憑證，而這樣的社會系統在這一波科技革命的壓力下，一定必須出現巨變，不然人類跟本沒有競爭力。親眼見證 Devin ，讓我感受到下一波的浪潮，但巨大的威脅，就也是改變的契機，教育真的該不一樣。

On Monday, when I walked into the office and opened my computer, I saw an unfamiliar name on the PR—Devin. Strange, the boss never mentioned hiring a new employee. I clicked on it, and it turned out to be Devin AI.

Whenever I mentioned Devin, the boss would loudly rave about how great it was. "How much does it cost per month?" I asked. "Only 500 bucks!" he said. What? For a startup to buy this B2B product, that's too expensive! The boss is extremely frugal and always complains about the company's runway—how could he say Devin is cheap?

Living in San Francisco, I've heard of and seen AI products that aim to replace engineers, but I've always found them lacking, so I never paid much attention. I thought that while AI replacing humans may eventually happen, it would still take time—until today when I got to know Devin. In the afternoon, the boss did a demo with Devin. He told Devin on Slack: "The sales page at the end of our user journey needs to be updated. The current one says 'Do you want a 3x improvement?', but it should be replaced with the new version. Can you help me update it?"

Devin immediately planned the update and then asked the boss: "Can you give me your account and password and walk me through the user journey?"

The boss gave them the credentials and told them to go to the course menu on the homepage, click "Start Studying," and keep clicking right until the end. We all watched in awe as Devin wrote the code. Not only did it push the PR, but it also created a browser video showing how it tested the code. It ran both the front-end and back-end, followed the boss's instructions to complete the user journey, and the updated sales page appeared perfectly. The commands the boss had taught Devin were already stored in its memory.

Watching Devin smoothly handle tasks that I used to manually code left me speechless. For a moment, I imagined a future where the boss says to me,

To be fair, although Devin was slow, it was probably more efficient than I was during my first week at the company. I can easily imagine that if it were a fresh college graduate with just basic programming experience who didn't know JavaScript or React, it might take them a month of intense study to catch up. But this is just the most basic version of Devin. One day, when fresh graduates need one or two years to catch up with Devin, the game will be over—the jobs of entry-level engineers will be completely replaced.

Recently, many new AI products are truly astonishing. It's not just logical work that will be replaced, but even "people-centered" jobs might be filled. For example, one of my best friend told me that AI is more helpful than a therapist because it provides immediate assistance, is patient, and doesn’t criticize. No need for appointments or waiting a week to process emotions. He said, "I feel more intimate when consulting AI than when talking to a human."

Since the emergence of AI, people have been scared into thinking they’ll be unemployed right after graduation, but few take a moment to reflect. If we spent four years learning at university and can’t apply that knowledge in the workplace, if we’re powerless in front of AI, then why go to school? University education was a product of the industrial revolution, and today, the talent trained by the industrial revolution is no longer considered "talent." Going to university now feels more like obtaining a credentials. Under the pressure of this technological revolution, society’s education system must undergo a massive transformation, or humans will lose their competitive edge. Witnessing Devin firsthand made me feel the next wave of change. While it’s a huge threat, it’s also an opportunity for change. Education really needs to be different.

Since AI programming tools emerged, the barriers to learning how to code have drastically decreased. On platforms like X (formerly Twitter), there are now videos showing how even a 10-year-old child can write software. The advent of AI tools like Cursor has not only lowered the entry threshold for software engineering but also significantly enhanced the learning experience.

The founding team behind Cursor says they focus on user experience with the mantra “fun = speed,” and there is a reason behind it. The thrill of speed brought by AI comes not just from avoiding long hours stuck on a bug without any guidance, but also from not needing to spend time worrying about the nitty-gritty details of programming syntax. AI can understand the context based on the current codebase and write the necessary code using information from the same document. This thrill extends to the speed of code generation itself. For example, Cursor stores the entire codebase in a KV cache, so AI doesn’t have to rerun unchanged contexts. When you ask it a question, you don’t need to provide relevant code—it can search for it itself. As a result, AI can handle everything, from understanding the problem and learning a new language to executing and writing correct code.

In short, as long as the basic concepts are clear, you can collaborate with AI to make corrections. Typically, the more advanced AI handles the planning, while the more basic AI handles execution. Learning to code becomes an easier process of trial and error and iteration. Given this AI-powered era, what does it matter which city we live in?

San Francisco's Work, Work, Work Culture: Coding is Like Breathing

Aside from tech employees in big companies who care more about work-life balance, many in the tech community (especially startup folks) often make headlines due to their extreme workaholic mentality. However, even excluding the workaholic mentality, many people still have their own side projects.

So for them, there’s little difference between work hours and after hours. They code during work, then go to hackathons on weekends to keep coding. With so many hackathons happening every weekend, I have to choose one out of three, many of which are sponsored by big companies like OpenAI and Anthropic.

Even a home-based hackathon organized by enthusiasts can attract famous figures because San Francisco is full of talent. Once, I ran into Chip Nguyen, a well-known machine learning author, at one such event, and I became tongue-tied. You could think of him like the BTS of machine learning—meeting him at someone’s home felt like meeting a pop star and having your heart race.

You can imagine that in this lifestyle, coding becomes as habitual as eating and drinking, and it doesn’t feel like “work” anymore. After a year, writing code for me became like how TikTok is to modern society—a default part of life. Why this culture exists is something I’ll discuss below.

The City as My Private Tutor

If I assume you're not working at a startup, and instead, you're looking at programming from a learning perspective, it's easy to see why staying in San Francisco is far superior to being in school. I’ll explore why the speed of learning software engineering in San Francisco is unmatched anywhere else by looking at four community-driven learning environments: work, accelerators, hackathons, and tech events.

The Struggles of Learning Programming: Only Half a Teacher

Let me give you an example from my university days, when I wasn’t studying engineering but took an IT course. Honestly, I didn’t think I had much engineering talent—in fact, I had none. I learned nothing in a programming course at National Taiwan University aimed at business students. Later, I tried learning on my own four or five times, but each time I got stuck, I gave up.

The biggest challenge wasn’t just not understanding programming languages or engineering thinking—it was also struggling with poor English skills that made even basic research difficult. The teacher would always say, “If you don’t know, just Google it.” But I couldn’t even form a Google-searchable sentence.

The only learning channel I had was a friend named Reggie. But our relationship was informal—there was no clear learning path, and I wasn’t connected to anyone in the computer science field. Our sporadic learning wasn’t enough to keep my motivation high, and as a result, there wasn’t much progress.

Switching to a CS Major: Still Just One Community

You might wonder, what about studying in a computer science department? After all, joining a department provides a formal learning path and the chance to collaborate with classmates.

When I studied CS at Minerva, there was indeed a group of classmates and professors with a clear curriculum, set hours of study, and obligatory assignments. However, most of the learning was limited to classroom interactions or assignments. Professors were the primary source of knowledge, but you could only learn more deeply during office hours or after class. If I wanted to extend my learning, for example, expanding from basic algorithms to the algorithms behind Duolingo’s recommendations, I had to figure it out on my own. The learning network was limited—it was just one network, and my classmates’ skills were all about the same, meaning we often got stuck together.

San Francisco’s Learning Network: The Entire City is My Tutor

In San Francisco, the organic learning network is far superior to anywhere else. Let’s break it down into four categories: work environments, accelerators, hackathons, and tech events.

1. Work Environment: In startups, speed is key, which greatly accelerates the learning curve. But let’s assume you’re not working at a startup—there are still three other communities that can help.

2. Accelerator Environment: My company and former company were both part of startup communities. Essentially, these are accelerators that host many startups, and outsiders can join if they find a way in. Every day, besides learning from the company’s internal talent, you also learn from others in the accelerator. You might tinker with robots made by other companies, or more deeply collaborate on SEO strategies. The problems in software startups are generally similar, creating a strong network effect. Many engineers in the accelerator also want to start their own businesses. During a period of unemployment, I found an engineer who needed help—he became my personal tutor and helped me quickly learn front-end engineering in just a few weeks.

3. Hackathons: Even if you’re not working and can’t get into an accelerator, you can still go to hackathons. Every weekend in San Francisco, there’s a hackathon, and the people who show up are usually passionate coders and highly skilled engineers. Some familiar faces show up every weekend, and I got to know many of them and learned from them.

4. Tech Events: San Francisco probably hosts around 20 AI-related tech events every week. Some are demos where engineers show off their projects, while others are company events or lectures. Many outsiders attend large-scale events, but I believe these events are the least effective for learning—they don’t involve hands-on experience, and the discussions are too shallow. At most, you might meet a few people.

Most importantly, though AI can write a lot of code, but it cannot teach you how to think, debug, or understand the philosophy behind a programming language. These are insights only experienced engineers can share with you. For example, why is React’s state management so strange? It’s because most programming languages are imperative—you tell the program what to do step by step, but React is declarative—it doesn’t care about the steps in between, only the end results. These kinds of philosophical insights into coding are what make the craft feel like an art form, and they’re likely to be the most valued skills in the future.

The Network Effects of a Concentrated Talent Pool

Many people think about networking in terms of job opportunities, but in San Francisco, the network effect created by the concentration of engineering talent brings rapid learning. The phenomena I observe are:

1. X (formerly Twitter) has become a fast way to accumulate and filter learning resources and social capital.

2. Shipping code every day (ship everyday) is a badge of pride in San Francisco, and everyone has a side project.

This unique culture of coding every day leads to high expectations for learning and fast iteration. It’s hard to live in San Francisco and not become a top-tier engineer.

X: The Fast Lane for Accumulating Social Capital and Filtering Learning Resources

X, like LinkedIn for San Franciscans, is filled with videos of new tech products, papers, and performance discussions. If you’ve made a notable project, you can go viral overnight and attract investors. For example, when a friend first came to San Francisco, he had only four followers on X. But after a few months of posting about his robotic arms, his follower count skyrocketed. He started a company, and prominent investors were reaching out to him.

San Francisco’s dense talent pool makes it clear that you’re only a few steps away from top-tier tech people. If you know a few key players, you can quickly find anyone you need, get invited to exclusive events, or even Apple’s launch events. In other cities, you have to slowly climb the ranks, but in San Francisco, if you have talent, you’ll get an opportunity. This gives engineers strong learning motivation without the slow process of switching careers or waiting for someone to notice your project.

Shipping Code Every Day is a Badge of Pride for San Franciscans: Everyone Has Their Own Side Project

When talking about the power of community, one must also discuss the culture that the community fosters. In a place where engineers are highly concentrated and AI startups are growing explosively, the criteria for judging people on social media have shifted. It's no longer just about your job title, but about how much effort you’ve put into writing code. The speed at which you can go from writing code to shipping it, or how many days a week you code, has become a way to show off. Many people proudly display how green their GitHub is (how many commits they’ve pushed), and many startup job ads say: "If your GitHub is a sea of green, you’re the talent we’re looking for!" But in Taiwan, no one really cares about your GitHub!

In a city where everyone has an opportunity, shipping fast has become an identity, surpassing job titles and alma maters (since so many people here have dropped out of school). It's the absolute indicator of top-tier talent. While it might seem silly, the best engineers I’ve met, or the fastest-growing startups, all have GitHub profiles full of green. They encourage everyone to code every day, making it as natural as breathing. Now, I too code every day. If I don’t write code, something feels off.

Having a side project is also a byproduct of this culture. During the week, you code at work; on weekends, you continue coding at hackathons. Side projects aren’t just hobbies—they could very well become the next company. If you win a hackathon, you can get investment, an accelerator seat, or attract the attention of well-known tech figures on X. Many people might start using it, which is very different from the typical side project expectations where you build something for fun that no one cares about. Additionally, because hackathons are so frequent, the feedback loop for side projects is extremely short. You can also find others to collaborate with, which accelerates iteration speed and keeps the project moving forward. When you meet people, the greeting often becomes: “What are you building?”

So, if a learning environment values writing more code, where everyone’s main job is coding, their side hustle is coding, and the people around them are discussing code, you’ll find that any talent who arrives in San Francisco will immediately be filled with lines of code.

The Entertainment-Free, Focused Atmosphere of the City

Apart from the network effects, the inherent limitations of the city also help shape good engineers. Because San Francisco has very little entertainment, it’s essentially an engineer’s deserted island.

I don’t have much money, so there’s even less entertainment for me. Unlike Taiwan, where there are tons of great places to eat and things to do, San Francisco isn’t very safe to go out at night. You need to go out in groups, so it's not really possible to have a "night out." Once the sun goes down, most people stay home. Even if you want to explore fields outside of tech, like culture or art, there are few opportunities, and the quality is often lacking.

For example, I took a dance class once and asked my classmates what they were doing.

“I’m a systems engineer.”

“I’m a machine learning engineer.”

“I’m a junior engineer.”

Thank you—everyone was an engineer, engineer, engineer. I suspect the only person who wasn’t an engineer was the dance instructor. The downside is that my exploration outside of the tech field feels limited, but the upside is that there are fewer distractions, allowing for a highly focused atmosphere.

San Francisco's Disruption and Innovation

People often say that San Francisco is where new eras are born, while New York preserves the wealth of the old world. Although San Francisco is truly a city with little entertainment outside of technology, the destruction and innovation it embodies have a deep allure for me. In other cities, if you talk about disrupting the old system and creating the future, you’ll likely be mocked. But in San Francisco, I don’t feel alone anymore. Watching the robot arms at the next table, knowing Sam Altman is giving a talk in the building next door, attending a hackathon and meeting Andrew Ng, seeing so many people around me on Forbes' 30 Under 30—it makes me realize that the world’s top talent isn’t as out of reach as it might seem. I understand that being at the core of a generation steered by AI in my 20s—whether successful or not—this journey will undoubtedly be exciting.

So, if you want to become a top-tier engineer, move to San Francisco!

自從 AI 程式工具出現，學習程式的門檻大幅下降，X 上面就出現很多連 10 歲小孩也能寫軟體的影片。Cursor 等AI 工具的出現，不只是降低的學習軟體工程的進入門檻，更大幅提升了學習體驗的滿足感。

Cursor 的創業團隊說他們在乎的用戶體驗來自 “fun = speed” 是有道理的，AI 提高速度帶來的快感，不只來自於不用再像以前一樣卡在一個 bug 卡很久，求師無門。現在寫程式，也不需要花時間計較編程語法的細節， AI 也可以根據目前代碼庫 (codebase)，了解上下文 (context)，根據同個文檔的資訊寫出需要的代碼。而這個快感也延伸到 AI 的程式的生成速度上，譬如說 Cursor 就會把整個代碼庫暫存 KV cache，因此 AI 就不用重跑變化不大的上下文，你問問題時也不需要給他相關的編碼，他可以自己搜尋。因此，從理解層次（解決問題、學習新的語言）到執行（寫出正確代碼），AI 都能一手包辦。

簡而言之，只要基礎概念足夠清晰，就可以和 AI 協作進行修正，通常讓高級的 AI 可以負責規劃，而基礎的 AI 負責執行，學寫程式變成一個更輕易試錯、迭代的歷程。那在這個人人都可以用 AI 的時代背景下，我們待在什麼城市，有什麼關係呢？

舊金山的工作工作工作文化，寫程式就像呼吸

除了大科技公司的員工多在意 work life balance，其他科技人通常工作及生活，有新創圈的朋友因為極端的工作主義被放上新聞。但即便撇除工作至上主義，很多人可能也都有自己的 Side project。

因此上班、下班其實沒有太大的差別，畢竟上班下班都在寫程式，然後週末去黑客松繼續寫。因為太多黑客松了，每個週末，我可能都得三個黑客松挑一個去，很多黑客松都是大公司像是 Openai、Anthrophic 贊助的，非常高級。

但即便是熱心人士自己在家裡舉辦的居家黑客松，舊金山也因為人才濟濟，會有神秘的名人現身，之前就在某家見到知名機器學習作家 Chip Nguyen 害我說話吞吞吐吐，如果你不能理解，你可以想像他是機器學習界的 BTS，然後你在某個人家家裡撞見他心跳加速的感覺。

因此你可以想像，在這樣生活風格下，可能每天都會寫程式，寫久了，就像是吃飯、喝水一樣令人習慣，也不覺得是在工作了。帶了一年後，寫程式之於我，就如同 Tiktok 之於現代人，是生活的預設的狀態，至於為什麼有這個文化，以下會做細緻的討論。

整個城市都是我的家庭教師

那如果我假設你不在新創工作，只以非常學習取向的方式來討論程式學習，也能夠很輕易地看出為什麼待在舊金山比在學校好得太多了，而以下會以人際網路的方式來探討為什麼舊金山的學習工程的速度是其他地方不能比的。

第一次學程式的困境：只有半個老師

就拿大學時，非工程學習，但修了一堂資訊課的我來說吧。說實話，我並不覺得自己有太多工程上的天份，更直白點說，應該是完全沒有天份才對。在台大上個給商學院學生的程式課學得亂七八糟。後來還主動學了四五次，都因為遇到卡關而作罷。

當時最大的困難，不只在於不明白程式語言、工程思維、連英文能力很差的情況下，連查個資料都很有難度。老師總是喜歡說「不會，就Google。」但我可是連把自己的問題，打成一個 Google 可以準確搜尋的關鍵句子都非常勉強。

當時唯一的學習管道，算是朋友 Reggie 吧！但我們的關係為非正式、非義務性、也沒有明確學習路徑的朋友關係，也沒有機會接觸資工相關的人際網路，有一搭沒一搭的學，學習動機容易隨著卡關下滑，自然沒多少進步。

轉戰資工系：也只有一個社群

那你可能會問，那如果是資工系的學習呢？畢竟加入系所，就是正式、義務的學習模式，也有同學相互切磋了吧。

在 Minerva 念 CS 的時候，確實是有一群同學和教授，有明確的學習路徑、學習時數、完成作業的義務。但即便如此，大部分的時候也只在課堂互動，或限於作業的框架範圍。教授為主要的資訊傳遞者，但也大概只有一週兩次下課加上 Office hour 的時候才能夠更深入的學習。如果想要擴增學習領域，從基礎演算法框增到學習 Duolingo 推播的演算法，還是要自己想辦法的。學習網路是有的，但只有一個，而且身邊同學的實力大同小異，大家常常卡在一起。

舊金山的學習：整個城市都是我的家庭教師

在舊金山有機的學習網路就比其他地方多上許多，我分成四個學習，工作、加速器、黑客松、科技活動來探討。

1. 工作環境：當然會有資深工程師的指導，新創講究速度，非常有利於學習曲線的指數性成長，但如果我們現在假設你不在新創工作，那你還有接下來三個社群。

2. 加速器環境：我的公司和前公司都在新創社群，簡言之，就是一個讓很多新創進駐工作的加速器，平常人如果找得到管道也可以進來。因此每天除了和公司內部人才的學習，也和身邊的其他新創學習，粗淺的玩其他公司做的機器人，深入的則是合作發展 SEO 策略，軟體新創的問題大同小異，有非常強的聚集效應。此外，會來的工程師很多也想自己創業，我在失業期就找到需要幫忙的工程師，他成為我的個人家教，讓我在幾週內快速學會前端工程。

3. 黑客松：即便你沒工作，混不進加速器，那還是可以去黑客松，舊金山每個週末都有黑客松，而週末還會來寫程式的人通常也是對於寫程式充滿熱情，神乎其技的工程師。有些熟面孔每週都會看到，最後也都認識了，向大家學習。

4. 科技活動：舊金山一週大該有 20 場 AI 相關的科技活動了，有些是工程師們去 Demo 自己的專案，有些是公司性的推銷活動或是講座。大部分不懂舊金山的外人會來參加大型活動，但我個人認為這些活動沒有實作、講的粗淺、也只是幾分種粗淺的認識人，學習效果最差就是了。

此外，AI 無法取代人的地方，是見解和哲學。AI 或許可以寫很多程式，但沒有辦法教你如何思考 debug 和某個語言的哲學，這只有遇見好的工程師會告訴你的，譬如説，為什麼 React 的 state management 這麼奇怪，其實就是因為 大部分的程式語言是 imperative ，你告訴程式第一步到第十步，但 React 是 declarative 語言，不在乎中間的流程，只在乎最前與最後的結果。這些更像藝術與哲學的思維與想像，我想都會未來最被重視的技能。他們的哲學，也讓我從害怕程式到出現真實的興趣，可以一輩子帶著走。

聚集經濟帶來的網路效應

大部分的人思考人際網路，可能會想到未來可以找工作的人脈等等。不過我想談的是舊金山工程人才密集帶來的網路效應 (network effect) 帶來的高速學習。

我觀察到的現象有 (1) X 變成快速累積篩選學習資源與社會資本的地方 (2) 天天交付程式 (ship everyday) 是舊金山人的驕傲，所有人都有自己的 Side project。

這樣獨特的文化推崇天天寫程式，也讓學習有很高的收穫期望值和快速的學習迭代，住在舊金山要不成為頂尖的工程師也挺難的吧。

X 變成快速累積社會資本與篩選學習資源的地方

社群網路 X 是舊金山人的 Linkedin，我的 X 充滿著新科技產品的影片、論文、和效能討論。實際上是，如果你有做出一些像樣的專案，你是可以一夕爆紅的，紅到吸引投資人。像是朋友剛到舊金山的時候，只有 4 個追蹤者，發了幾個月的機器人手臂之後，追蹤者暴漲，創了公司，知名的投資人也去密他。

舊金山的人才高度密集，讓你知道你和知名的科技人在 X 上真的只有幾步之遙，如果認識一些網路核心 (network hub) ，想要找誰其實都很快，也會被邀請進有門禁的特殊活動，或是 Apple 的發表會等。這和一般城市要慢慢向上爬完全不同，舊金山相信年輕人，有才就有機會，帶給工程師們非常強烈的學習動機，不會有花很多時間學習慢慢轉職成功，或是花很多時間終於有人看見你的專案等，收割遲緩的問題。

既然好的新科技很快就會在 X 上出名，平常見到的人也很快就會告訴你，篩選出好的學習資源自然容易，畢竟整個城市都幫你篩選好了。學習新科技遇到的問題常常是工具的選擇，選了一個不好用或是難學的工具，也找不到人求教。舊金山的集體智能都幫你選好了，好的工具在 X 上曝光，在線下的認識的人認可，是會聽到很多次，看到很多熟人轉發的，想不知道都有困難。

這就和菜市場不知道要買哪一攤的菜一樣，不需要每家都嘗試，認真審查，只要聽三姑六婆咬耳朵就差不多可以相信了。目前學工程的大問題其實不時資源不足，而是資源雜亂，不知從何下手，但如果社群已經幫你做出學習選擇，不但省時、省力，踩雷機率小，還可以在透過身邊的人的解釋大致了解工具的運用方式，在 X 上看效能，其實學習地圖自然而然就在腦海裡成行了，根本不需要大張旗鼓地做研究。也降低了學習卡關的阻力。

總結而言，X 與現實交錯互動帶來的外部肯定，與對學習資源的篩選效果都帶來快速學習的動機與成效。

天天交付程式 (ship everyday) 是舊金山人的驕傲，所有人都有自己的 Side project

如果談論社群的力量，就得談論社群所造就的文化。當工程師高度密集，AI 新創爆發性成長，大家在社群媒體評斷人的標準就已經不太是職位了，而是已努力寫程式。從開始寫到上架的速度，一週寫幾天都是賣弄的內容。許多人會炫耀自己的 Github 有多綠（推送了多少程式），很多新創招人廣告也會說：「如果你的 Github 是一片青綠，你就是我們想要的人才！」但在台灣根本沒多少人會管你的 github！

在這個人人都有機會的城市裡，Shipping fast 成為身份認同，超越職稱和畢業學校（畢竟這裡超多人輟學），是優秀人才的絕對指標。雖然看似愚蠢，但其實我所見過的優秀工程師，或是高度成長的新創，大家的 Github 確實是一片青綠，鼓勵大家天天寫程式，讓寫程式變成像呼吸空氣一樣自在，我現在也真的是天天都會寫，沒寫就覺得哪裡不對勁。

至於人人都有 Side Project 也是這樣文化下的產物，週間上班寫程式，週末在黑客松繼續寫。Side project 不只是愛好，他們也可能變成下一個公司，如果在黑客松贏了的話，很多是可以拿到投資或是新創加速器的座位，或是得到 X 上知名科技人的青睞，或是很多人使用，這和一般人想像自己做好玩，沒有人在乎的 Side Project 期望值完全不同。此外，因為高度密集的黑客松，Side Project 的回饋週期非常短，也可以找到其他人一起做，因此迭代速度可以非常的快速，專案的發展自然有好的動力。大家見面打招呼也變成 “What are you building?” 。

那如果一個學習環境以寫更多程式為榮，大家正業是寫程式，休閒副業還是寫程式，身邊見到的人也討論程式，任何一張白紙人才一送來舊金山馬上就會填上代碼的。

沒有的娛樂，專心致志的城市氛圍

除了網路效應外，先天的程式限制也會讓人成為好的工成師。因為舊金山沒、有、娛、樂，完全是工程師無人島。

我沒錢，所以更沒有娛樂。不像是在台灣，到處都有好吃的好玩的。而且舊金山晚上出門並不安全，還得很多人一起才敢出門，根本不可能有什麼夜衝，太陽下山了，大部分的時間就待在家了。即便想要探索科技以外的領域，像是文化與藝術，不但機會不多，品質也不高。

像是我去上了跳舞課，問了同學他們在做什麼。

「我是系統工程師。」

「我是機器學習工程師。」

「我是初級工程師。」

謝囉，全部都是工程師、工程師、工程師，我懷疑只有跳舞老師不是工程師。壞處是我感覺到自己科技領域外的探索有所受限，好處是沒有太多分心的事物，專心致志。

舊金山的破壞與創新

大家常說，舊金山開創新的紀元，紐約傳承舊世代的繁華。雖然舊金山除了科技以外，真的娛樂與享受都不多的城市，但它承載的破壞與創新，對我有著深深的魅力。在其他城市口出狂言，表達自己想要破壞的舊制度，想要創造的未來，可是要被取笑的，但在舊金山，我不再感到孤獨。看著隔壁桌的機器人手臂、知道 Sam Altman 就在隔壁棟演講、去黑客松也能見到 Andrew Ng，看見身邊一堆人進了富比世的 30 under 30，就知道自己離世界頂尖，並沒有想像中的遙不可及。也能明白能在人生二十來歲，在一個掌舵駛向 AI 帶來的世代的核心，不論是成是敗，這航道都一定精彩可期。

所以如果想成為頂尖的工程師，搬來舊金山吧！

You might ask me, why don’t you start with an MVP? The thing is, I’m already quite fried from coding at work. Second, I realized one thing: even without an MVP, a pitch deck can still be an entry point to secure funding for your idea. Last but not least, honing your idea is still a good call!

I realized I feel less scared about the 200 days commitment if I’m just reading books and drawing random things in my mind! And also gathering some output weekly feels much less daunting and also quite comprehensive.

My Practice on making pitch deck

This is my first idea sketch.

How to tell a good story

Pitch deck essentially is about story telling, a skill I can practice and exercise on. Hence, I started from How to Tell a Story, a book recommended by a talented video maker Joshua from our incubator.

Let me share with you on my favorite story so far in the book.

She locks me in a gaze and says, "You're cursed. You're cursed! You'll never find love. This is the worst curse I have ever seen.

Your whole family this is generational. I am floored by the level of curse that's hanging on you." And I'm standing there. And I

gotta tell you, I'm pretty excited about this. Because all these years, I thought it was me! I thought I was too loud, too fat, too Black, too whatever. But if I'm just cursed? Hallelujah! Yes, that means we can fix it. All right, here we go.

Personal stories are usually the easiest, here are some to start with.

• Felt an emotion: doubled over with laughter, burst into tears, or lost your cool. Did something you never thought you'd do.

• Tried to be something or someone you aren't.

• Discovered something about yourself, your environment, your family, or the world

• Changed your relationship with someone for better or worse, a little or a lot.

• Had a secret revealed—by you or someone else.

• Stood to gain or lose something that mattered to you.

• Made a tough choice for the right (or wrong) reason.

• Found yourself saying, I do! I won't! Hell no! I dare you. You couldn't pay me to. It would be my greatest honor.

But how can I translate personal stories into a professional setting?

I found this a very funny but inspiring story!

Think through the stories in your arsenal before the interview and have one or two prepped In an interview for a TV producing job, I once managed to work in a story about me surviving a runaway camel in Morocco without dropping the expensive camera I was holding. (I got the job!) But one of my favorite on-the-job stories is told by my friend Tim Bartlett. He once spent a year traveling solo through rural England, shooting a documentary. He took buses from town to town, and stayed overnight in the small villages where he was filming. One cold winter night, he missed the last bus that would take him to the next town. There were no taxis or hotels in the tiny village where he was shooting. In the end, he ordered a pizza, asked if it could be delivered to the place in the next town where he was staying, and then said, "Uh... can I go with it?" The delivery guy took Tim and his pizza to the hotel in the next town. The story demonstrates that Tim is a creative problem solver, with an adventurous spirit and great sense of humor. I'd certainly hire him!

This week from the book, I learn this:

Is there anything decision making or turning point I can put in my pitch deck?

Learn from existing good stories

I studied several tech pitch deck I can find

• Tinder

• Uber

• Wework

I found a lot of comfort that all the pitch decks are extremely ugly. But it helps me to kick things off too. So I made this second version of pitch deck.

Tbh, I have no idea if this makes sense at all, but that’s the point of practicing!

People of the week

I hung out with one of the founders from Quizard AI. He gave me some ideas of approaching my vision. He said “Let’s think about who is the person on this planet that is at the frontier of the vision.” It might be Sal Khan from Khan Academy.

What is his path? He got lots of donation after he became the influencer. Is it replicable? Yes! I can be an influencer but will take me 3+ years. Can I skip the steps?

Maybe, he funded some other ideas, like schoolhouse. Can Sal Khan also be my sugar daddy too? Maybe.

I still don’t know how to make this work at all, but he gave me another potential route to explore.

Next week goal

Pitch to 3 random friends and get feedback about it! Well, if you want to give me feedbacks now, that works too hehe.

He told me that “go big or go home.” But essentially, it’s just some weekly fun call that we take about shit and discuss about ideas and problems.

Practical thoughts on building my vision

After learning about San Francisco this year, I realized there are several routes for me to work on my learning problem.

• influencer route: The tokenization project I made is pretty solid with learners so I know that if I continuously doing it. I can be the next 3blue1brown. I can start from accumulating a big audience first.

• company building route: SF has lots of incubators, and I started to accumulate connection, plus our incubator does have program to coach and support early founders already.

Within the company building routes, there are also two more routes I can take

• Incubator routes (like YCombinator): As far as I know, there are two routes that I can potentially take to get into YC.

  • Great credential with mellow product

  • Great products with less credential

  • As we are the top 5% of our YC application in the last batch even without any user, I know that if I can find a good co-founder with big tech background, we have very high chances. But the downside is that product and vision misalignment causing cofounder breakup might happen in the middle. As for the second route, I will need a half baked product with some sort of revenue stream ~5K.

• Finding investors: It is really possible to pitch and directly raise in SF too, I know lots of founders that can give me intros. It feels a bit scary and stressful so I feel I resist this idea A LOT.

I feel so happy when I’m writing this down, because all the insights come from all my failures and experiments last year!

The biggest challenges - 28 weeks consistency

You thought my biggest challenges is building a company? No, haven’t even there yet! My biggest challenges is doing this new project for 200 days! How is that even possible? Do you mean doing something else EVERYDAY while I’m working on the busy startup from Monday to Sunday? That’s insane!

200 days is ~28 weeks, so I came up with a 28 weeks curriculum for myself. Let me show you the first week curriculum in my mind.

• Week 1: Understanding Startups and Identifying Problems

  • Learning Tasks:

    • Familiarize yourself with the startup ecosystem and YC's philosophy.

    • Learn about the Lean Startup methodology.

  • Output Tasks:

    • Generate a list of 10-15 problems you're passionate about solving.

  • Recommended Resources:

    • Book: The Lean Startup by Eric Ries.

    • YC Startup Library: YC's Essential Startup Advice.

    • Video Series: How to Start a Startup.

  • Practical Exercises:

    • Problem Brainstorming: Spend time each day noting down problems you encounter or observe.

    • Discussion: Talk to friends, family, and colleagues about challenges they face.

Though 28 weeks sounds so scary, but “spend time each day noting down problems you encounter or observe” sounds like a fun one!

What I learn this week - Criteria for my solutions

Today, friend and I brainstormed the thing I want to build. And I learned that the things I wanted to build contains these 5 elements.

• in-depth learning: Quantum country, How might we learn

• creativity & fun: pika, speedrun companies (An incubator focusing on game or gamification, brilliant, trees Esther made

• extendibility & flexibility: heptabase, the curious interface, one million checkbox from eieio

• decentralized learning (lack of predefined curriculum): buildspace

• potential to replace formal education system: buildspace, Minerva

And then I realized I really cannot imagine anything that hit all of these criteria till today. But at least I figured out the criteria.

Idea for the week — Black mirror for learning

This week, we came up with a new learning idea—Black Mirror for learning.

Traditional curriculums can feel limiting, so what if we gave learners more flexibility? Imagine the learning content as a "black mirror," where everyone receives different experiences tailored to their learning style, much like how astrology categorizes people.

For example, when learning engineering, we all follow the same standard Jupyter notebook—predictable and uninspiring. But what if, instead, each learner was guided through a unique storyline in their project? After completing a problem, they would be assigned a new narrative or problems to solve based on their learning style, creating a personalized and dynamic learning journey, just like in Black Mirror.

Rating for this idea

• in-depth learning: Yes!

• creativity & fun: Yes!

• extendibility & flexibility: Have some possibilities

• decentralized learning (lack of predefined curriculum): A bit

• potential to replace formal education system: Not sure

I still think 100 days is my max consistency ahhhh. Weirdo friends, 28 weeks consistency is a bit too hard for me so please don’t hold me accountable haha

Or is there anyway that I can make this work? Like creating a gamified or beautiful visualization for consistency? Or request a winter vacation from my friend?

Maybe I’ll request some vacation.

I need you to be my listener and brainstorm buddy today. My founder friend told me to start by writing my vision on a slide deck, but I cannot. None of the ideas feels good enough. Or maybe it’s my vision that isn’t clear enough.

I’ve gradually realized that I’ve hated all the existing and new education and products even the new AI technology is blooming. But what do I want?

I hope I can use this post to explore and listen to your idea brainstorm.

What do I dislike about the existing education system?

A major flaw in the current education system is its failure to recognize the critical importance of experiential learning. We spend years studying subjects that others deem essential for our future, but much of what we learn often feels disconnected from the real world. Even when the content is valuable, students struggle to engage deeply with it because they haven’t encountered real-world problems that demand such knowledge. Without practical experience or a clear connection to real-life challenges, students lack the motivation to fully invest in mastering difficult subjects.

For example, during business school, I was buried in case studies, multiple-choice questions, and marketing theories long before applying any of it. But what if we experienced these scenarios first, and then learned the concepts? It’s like someone talking endlessly about relationship advice without ever having been in one. Experience should come first, and teaching should grow naturally from that experience.

What do I dislike about existing education products?

Many existing education products offer only incremental improvements rather than the transformative changes we need. They focus on optimizing tasks like grading, exam preparation, and lesson delivery—essentially improving the mechanics of a flawed system. But in the year 2030, should we still rely on the traditional concepts of teaching, grading, and exams? These products perpetuate an outdated model rather than challenge the fundamental assumptions of how learning should happen. The question we should be asking is not how to make incremental improvements, but whether these structures should exist at all in the future.

The Purpose of Education

My friend and I have identified three foundational pillars that education should cultivate:

1. Values: Instilling a strong ethical compass and guiding principles that shape decision-making and behavior.

2. Abilities: Developing cognitive frameworks, such as learning how to approach new tasks—like acquiring a language or solving complex problems.

3. Skills: Building specific proficiencies, such as fluency in languages like German, Spanish, and English, or mastering technical disciplines.

To invent the future, we need to have a clear vision of what that future looks like. For instance, what should education in 2040 aim to achieve? When I try to imagine the kind of company or product I want to build, I realize I'm confined by the present. If we want to create something groundbreaking, we have to start by imagining society in 2040.

Critical Questions for Education’s Future

• What will technology look like in 2040? Will we live in a fully immersive digital world, perhaps the metaverse?

• What skills will be needed? Will memorizing information still be relevant when AI can retrieve and process knowledge instantly?

• Will traditional school structures—elementary, middle, and high school—still exist? Or will these rigid distinctions become irrelevant in a world where learning is continuous and adaptive?

• Will subjects as we know them still matter? Could we move away from predefined disciplines and focus instead on solving real-world problems?

As I ponder these questions, it becomes clear that we are currently underestimating the extent to which education must evolve. The goal of education shouldn’t be to simply improve existing systems, but to rethink them from the ground up. If our future remains tethered to traditional concepts, we will fall behind.

Foundational Assumptions About the Future

1. Rapid Societal Change Will Demand Faster Learning Cycles
   The pace of technological and societal change will be so fast that traditional, multi-year education programs will be too slow. By the time students graduate, much of what they learned may be outdated. This means that education must become more agile, with quicker feedback loops between learning and application. Students must be able to pivot and adapt continuously, based on real-world changes.

2. Technology Will Redefine the Core of Education
   Technology, especially AI, will dramatically reshape what is considered essential knowledge. For instance, AI tools like ChatGPT have already diminished the importance of memorizing facts and publicly available information. As a society, we’ve long considered knowledge retention to be a core indicator of intelligence. But in a future where AI can instantly retrieve and apply this knowledge, memorization will become irrelevant. Instead, collaboration, creativity, and problem-solving will become the cornerstones of education.

   By 2040, if machines take on many routine tasks, the skills we’ll need will shift dramatically:

   • Collaboration: Not just with people, but with machines. Learning how to work harmoniously with AI and other advanced systems will be crucial, as will the ability to keep pace with technological advancements across different sectors.

   • Control: Mastery over one’s habits, especially regarding technology (e.g., self-control in the age of social media), will be increasingly valuable. Ethical considerations and responsible use of technology will become as critical as technical knowledge.

   • Communication: Advances in AI will change how we interact with machines and with each other. With so much more information at our disposal, communication will need to evolve to handle greater complexity, nuance, and the exchange of ideas.

   • Invention: The ease of turning ideas into reality will make creativity and innovation central to success. For example, AI might enable individuals to create new products, companies, or systems without traditional expertise in engineering, marketing, or design. The emphasis will shift from rigid qualifications (such as work on C++ for 5 years) to the ability to ideate and execute effectively on specific problems. Good ideas will matter.

What Should Education in 2040 Look Like?

Problem-Centric Learning, Not Institution-Centric Learning

By 2040, formal schools as we know them today will likely be obsolete. Schools were created to bridge a skills gap between people and the workforce, but now they often perpetuate outdated models that no longer address the real challenges of society. Schools project prestige through branding, but they often fail to equip students with practical, relevant skills.

Project-based learning models exist today, but they are often theoretical or based on small-scale "toy" projects. In the future, the focus must be on real-world problem-solving from day one. With the accelerating pace of technological development, students will have the tools to build and test actual products, not just propose hypothetical solutions. This shift would create an education system based on direct experience and feedback.

Instead of having fixed institutions, I imagine a dynamic, problem-solving network that partners with companies, NGOs, and government organizations. Students would rotate through different real-world challenges before deciding what they want to focus on. Education would be deeply embedded within problem-solving institutions, allowing for continuous learning and practical application.

Instead of having a school, I’m imagining and wild problem rotation programs for people before they deciding on something they want to solve. The function of a school will be merged into those entities like Uber x Waterloo University, Startup incubator x Minerva university. These "schools" wouldn’t exist as independent entities anymore, but as collaborative spaces for learning, sponsored by governments, companies, and learner contributions. For simplicity, I'll still refer to them as schools to help illustrate the idea.

Imagine a “school” building next to the Uber office. Students would take classes on real problems Uber is facing, learning theories from fields like computer science, economics, and data science. Then, they would spend six months working directly at Uber, applying what they’ve learned to real challenges. Afterward, they could either continue taking classes related to the problems they encountered or rotate to another problem area, depending on their interests.

To make the idea even clearer, imagine you want to become a Wall Street financial analyst. You would live alongside professionals in that field, observing their daily routines, how they dress, and how they network. This immersive experience would give you a firsthand understanding of the lifestyle and culture, helping you quickly determine if it’s the right fit for you.

Implementation of ideas throughout the process will be encouraged, either through the support of those entities, or “the school” will assist the experiments through personal launch due to the possibility of technical advancement.

What would students learn in these schools? While implementing solutions focuses on details, schools would extract lessons from the challenges faced during implementation and provide guidance. For example, working on Uber’s real-time data streaming platform might teach students both the technical challenges of real-time architecture and the economic principles of supply and demand fluctuations.

As for community and relationships, these would still form through cohort-based learning models. The difference is that segregation would no longer be based on age but on passion and shared interests.

Education should help students discover not only their interests but also their values, life goals, and desires. Through real-world problem testing, students will learn what truly matters to them—whether it’s money, stability, the problem itself, or the vision behind it. This would reduce the frustration of spending years in traditional schools only to find out after graduating that a chosen career path isn’t the right fit.

Currently, education only allows us to peek at problems through reading or classroom discussions, but in 2040, we will be able to engage with real problems, deploy solutions, and receive immediate feedback from society or experts within problem-solving institutes.

The interaction between college professors who’ve never faced real problems and students who lack experience won’t be enough. Problems create pain, and pain drives motivation and learning. Young adults don’t want to be treated like children—they want to explore and contribute. With the right playground for experimentation and guidance, students will no longer be passive learners—they will be active creators, contributing meaningfully to society from day one.

Well, this is a new idea produced in my brain today. I’m curious about how you think!

We decided to visualize the progress through the metaphor of tree growth. This is the previous version I shipped. Essentially, when people answer questions, they will see their tree grow every 10 questions.

This version of the tree growth feature I released increased user engagement by approximately 40%, and today I’m launching a new version!

Your tree will grow alongside your progress!

If the learner answers the question correctly, they will be able to grow more trees and see their progress!

I loved this tree so much that I spent two whole weeks working on it, except for the day I was moving a house.

Well, to be fair, I was still working that day, but I didn’t have a desk in the new house, so I stacked all the paper boxes to use as a desk, haha.

Since the start of the school year, we’ve been consistently reaching all-time highs, which is a really good sign that our product decisions and marketing strategy is working well so far.

But the data isn’t the only thing that makes me happy. I feel super excited about this new tree launch I made because it can only be unlocked in the digital experience. I feel that I start to find some uniqueness in digital learning experience that I can explore. Plus, I disliked many digital learning products because they tried to make learning easy for people without really thinking of ways to help them go through the process of struggling. For this tree, the difficulty and the number of questions haven’t changed, but it can potentially greatly increase people’s willingness to try a bit more.

I find this to be a potential sweet spot to explore how technology can genuinely help humans learn in a creative way. Excited!

最近在加速器裡認識不少 19、20 歲的同學和高中人，有些高中剛畢業甚至已經賺超多 1M 了，但從國外空降。我問其中一個高中生，「你加拿大的高中有人做 AI 嗎？不然你怎麼有辦法找到這裡來？」他回答：「當然沒有人知道，不過我就在推特上發自己做的東西、到處逛，慢慢就找到這個 community 了。」

大部分的台灣人可見的線上資源，也多著重於怎麼進大型科技公司的計畫，我們從小到大熟悉的故事也是，念完高中和大學，工作幾年賺了錢，碩士之後就可以出國學習了。但是這是唯一的路徑嗎？

因此我想推薦一些待在成本較低，甚至待在台灣也可以嘗試的學習資源，如果能運用得當，不但可以認識很多厲害的人，甚至會有不用花大錢也能出國的機會。

第一步：BuildSpace

• 適合人選：推薦給還在探索，不知道自己想要做什麼的人

• 簡單介紹：目前我看過辦的最好的全球性線上 program，只要有個想嘗試點子，應該是有申請就會上，大家連續六週在推特上互相監督，點子不限於科技產品，有些人做 podcast 訪問自己想訪問的人，有些人寫歌，練習如何當 DJ，有些人辦活動、組織和陌生人吃晚餐，有些人做自己喜歡的 AI project。六週結束後，有做完的人能到舊金山或杜拜參加 Builder Demo 認識其他人。因為是自付，所以在台灣可能不方便去，但光是線上的 community 就已經非常好了。Builspace 有機制讓你連接和和你專案相似的同好，大家也都會在推特上給彼此每週的小 demo 建議，對於還不知道自己想要什麼，要不要投資的人來說，是成本極低，探索廣泛性強的 program。 這是最近 Buildspace 評選大家每週 demo 的連結：https://x.com/i/broadcasts/1LyxBnXjVRyxN。

• 官網連結：https://buildspace.so/

第二步：The Residency

• 適合人選：已經在自己興趣的點子上做出點東西，但還不確定要不要當成職業的人

• 簡單介紹：The Reisdency 每幾個月就會選十幾個 Residence 來舊金山住，最近辦了給大學生的六週 program（也有高中生），會需要一些科技相關的背景。他們和新創加速器有合作，讓大學生到加速器做自己的專案，也有畢業生參加他們六週的硬體 program 和四個月的新創 program，一樣是探索並做自己的點子。我推薦 The Residency 的原因是因為裡面的人聰明謙虛，也互相支持，community 的感情很好。有人開發用真的神經開發 AI 運算的實驗、有人做衣服穿搭 app 的 side project，最後變成一加公司、有人做了機器人手臂而一砲而紅。當然，雖然有人成立公司，也不是所有人都在 program 後繼續做自己的點子。但這是個認識矽谷人脈和資源、並更深入探索自己興趣的好機會。如果想瞭解大家的 project demo 和生活樣態：https://x.com/_TheResidency

• 另外，灣區有也很多的 hackerhouse community 像是 Arcadia，德國朋友大學暑假就來住幾個月一邊做自己的點子，一邊體驗灣區生活，也是個成本較低的探索方式。

• The Residency 官網連結：https://www.livetheresidency.com

• Arcadia 官網：https://joinarcadia.org/

第三步：Neo accelerator, Entrepreneur First

• 適合人選：打算全心投入一個點子，在加速器裡得到支持或融資的人。

• 簡單介紹：大部分的人應該都只知道大間的加速器，像是 Y combinator，但其實還有很多很好的隱藏版加速器。Neo Accelerator 是為期三個月的加速器，也有 scholar program ，媒合 scholar 的人到裡面的新創工作，做得好的話，他們的開的條件為 $600K（1.5% equity），因為條件比 YC 好多了，所以朋友放棄 YC 去 Neo。也朋友推薦歐洲緣起，最近跨足美國的 Entrepreneur First，是為期三個月，如果做得好一樣會得到投資。

• Neo 官網：https://neo.com/

• EF 官網：https://www.joinef.com/

最後，我發現一個隱藏版的重要社群是推特。大部分的台灣人雖然不用，但最稀奇、最先進的科技產品都在這裡，大家也會發自己的做的點子，和各種學習資源。上面推薦的 program 都有不錯的推特，稍微滑一下就可以看到有趣的人和專案，如果不知道如何開始，可以從很GenZ 的 BuildSpace 開始看起。

I often receive messages from students who are dissatisfied with the current state of learning in Taiwan. The most common question is: If I can’t go abroad or to Minerva, what can I do to continue improving in Taiwan?

Recently, I’ve met quite a few 19- and 20-year-old students and high schoolers in an accelerator. Some have just graduated from high school and are already making over 1 million dollars, but they came from abroad. I asked one high schooler, “Does anyone in your Canadian high school work on AI? How did you find this place?” He replied, “Of course, no one knows about it, but I just posted my work on Twitter and browsed around until I found this community.”

Most online resources available to Taiwanese people focus on how to get into big tech companies. The stories we’re familiar with from childhood are about finishing high school and university, working for a few years to earn money, and then going abroad for a master’s degree. But is this the only path?

Therefore, I want to recommend some learning resources that you can try while staying in a lower-cost environment, even staying in Taiwan. If used properly, you can meet many talented people and even have opportunities to go abroad without spending a lot of money.

Step 1: BuildSpace

• Suitable for: Those still exploring and unsure of what they want to do.

• Brief introduction: This is the best global online program I’ve seen so far. If you have an idea you’d like to try, you should be able to get accepted. Everyone supervises each other on Twitter for six weeks. The ideas aren’t limited to tech products; some people create podcasts, write songs, practice DJing, organize events, or work on their favorite AI projects. At the end of six weeks, those who complete their projects can attend Builder Demo in San Francisco or Dubai to meet others. It’s self-funded, so it might not be convenient for those in Taiwan to go, but the online community is already excellent. BuildSpace has mechanisms to connect you with people who have similar projects, and everyone gives each other weekly demo feedback on Twitter. It’s a low-cost, wide-ranging exploration program. Here’s a link to recent BuildSpace weekly demo selections: https://x.com/i/broadcasts/1LyxBnXjVRyxN.

• Official website: https://buildspace.so/

Step 2: The Residency

• Suitable for: Those who have already made some progress on their interest but are unsure whether to pursue it as a career.

• Brief introduction: The Residency selects about a dozen residents every few months to live in San Francisco. They recently held a six-week program for college students (and some high school students), requiring some tech-related background. They collaborate with startup accelerators, allowing students to work on their projects at accelerators. They also have graduates who participated in their six-week hardware program and four-month startup program, exploring and working on their ideas. I recommend The Residency because the people there are smart, humble, and supportive, forming a strong community bond. Some develop AI computing using real neural networks, some create clothing apps that turn into companies, and some build robotic arms that become popular. Although not everyone continues with their ideas after the program, it’s a great opportunity to connect with Silicon Valley’s networks and resources and explore your interests more deeply. Here’s a link to their project demos and lifestyles: https://x.com/_TheResidency. Additionally, the Bay Area has many hackerhouse communities like Arcadia, where a German friend stayed for a few months during summer break, working on their ideas and experiencing Bay Area life at a lower cost.

• The Residency official website: https://www.livetheresidency.com

• Arcadia official website: https://joinarcadia.org/

Step 3: Neo accelerator, Entrepreneur First

• Suitable for: Those planning to fully commit to an idea and seek support or funding from an accelerator.

• Brief introduction: Most people only know about big accelerators like Y Combinator, but there are many other excellent hidden accelerators. Neo Accelerator is a three-month accelerator with a scholar program that matches scholars with startups. If you perform well, they offer terms of $600K (1.5% equity), which is much better than YC, so some friends chose Neo over YC. Another recommendation is Entrepreneur First, originating in Europe and recently expanding to the U.S. It’s a three-month program, and if you do well, you can also get investment.

• Neo official website: https://neo.com/

• EF official website: https://www.joinef.com/

Finally, I found an important hidden community on Twitter. Although most Taiwanese people don’t use it, the most innovative and advanced tech products are there. People also post their ideas and various learning resources. The recommended programs above have good Twitter presences, and a quick browse can reveal interesting people and projects. If you don’t know where to start, you can begin with the very Gen Z BuildSpace.

I remembered all the past conversations on “Wow, you got into a xyz company. Please take me to the moon with you!” And I wonder what’s the point. Tbh, I find it incredibly hard to work with those who considered themselves as “elite” because they are so egocentric. Everything needs to be about them. If you disagree with their ideas, they think you are disagreeing with their elite identity. Most of the time, I get nothing done with them.

If you really want to compare social status, almost everyone around me has worked in Google, Meta, and Amazon, and doing even better now, but I seldom meet people who think they are elite. How can these people even be so arrogant?

Even till today, lots of Taiwanese reached out to me, there is a pretty high ratio that the experience are shitty. I find them unpassionate in the topic they are talking about while eager to place me inside their network.

But enough of my complaint. If I meet 1 elite out of 100 people, then it’s that person’s problem. But out of random sampling, if I meet 80% of people think they are elite, there is no point for me to be angry at them anymore. Because it’s obvious that the phenomenon aren’t individual, but collective. And here is my thoughts on the societal elitism in Taiwan.

Taiwanese Elitism is intertwined with insecurity

Many people perceive the elite as always successful and, consequently, arrogant. However, my experience with the Taiwanese elite has led me to a different understanding.

Elitism, I've realized, often stems not from confidence but from a deep-seated insecurity driven by intense competition. After all, if you are genuinely confident in your greatness, why the constant need to assert and showcase it?

This competition, some argue, is a fundamental aspect of human nature. But does this hold true everywhere? For instance in our school, Minerva, where we are famous for our supportive community; toxic competition is rare. Although you might still encounter kids say “X starts a company. Why I haven’t!” “You are doing so well, I feel so behind!” such expressions more often highlight their own insecurities rather than a genuine competitive drive.

However, when elitism ingrains itself as a culture, these insecurities become collective. In Taiwan, I observed two main aspects of this phenomenon. First, elitism is a status perpetually under threat from constant competition, asking individuals to continually assert their position. Second, it is locally bounded: individuals may feel accomplished in Taiwan yet believe they fall short on an international scale.

Where does that collective insecurity coming from?

Believing in your own capabilities within Taiwan is straightforward, but envisioning yourself competing internationally sounds impossible. This was my mindset until I studied abroad in the Netherlands. The inferiority complex I experienced, I've come to realize, is deeply influenced by our education system's design. This includes the scarcity of resources in individual schools, the low confidence among educators, and a toxic peer culture.

Our education system: Simplifying Success to a Single Metric

I argue that the biggest problem of this culture is the grading system in Taiwan education. Why? Because giving people a grade, is a very simple and single metric for success. This single-dimensional metric simplifies success into a numeric value, leading to inevitable comparisons among students.

On the surface, it ensures fairness, but as it completely ignores the individuality, then we ignore our own individuality too. Even though after university, this measurement disappears, but as everyone is so used to seek that 1 and only successful metric, the society create that for us, like that 1 spot in Mckinsey every year. This perpetuates a cycle where society continuously seeks to define success through narrow lenses. Elites are those who succeed in that single metric.

Imagine, instead, a society where success is defined by both academic performance and community involvement. In this world, children are recognized every three months not only for their academic achievements but also for their contributions to the community. Picture a school assembly where students are celebrated on stage. A teacher might announce, “Sarah has the highest grades this term. But let’s not forget Ken, who has made outstanding contributions to our community. Everyone, please clap for Ken and strive to follow his hardworking example next term!”

In Taiwan, the top school isn't just NTU; it’s also NCU – National Community University (my fictional creation for this story). Parents boast to their neighbors, not just about their children's academic prowess, but about their enrollment in NCU. These students are celebrated on the "community cram school" boards for learning the importance of creating a positive impact in their communities. NCU students are sought after as friends and partners because of their dedication to community service. Instead of seeing headlines like “How to Ace Your Exams,” news and magazines feature stories like “Look! She Grew a Green Garden for Her Entire Neighborhood!”

How will our society look like?

This is just an imagination, but it’s a simple example proposes that broadening evaluation criteria can cultivate a more inclusive environment that appreciates diverse talents and contributions. When we move beyond the single metric of grades and include others, it reduces direct competition by complicating the metrics of comparison. This diversity in evaluation criteria encourages a wider range of skills and interests, fostering curiosity and passion. 1

Even if you insists that the fairness matters so much that we can only have 1 metric, the current metric, grade, is still quite meaningless. It’s very odd to assume that academic success ties strongly with professional success while the curriculum itself has little tie to the skills in workspace in 21st centuries. We need to open the door to a more holistic and meaningful assessment of an individual's achievements.

University: Competition of scarce resources

Education resources are poorly allocated in universities. Without exams after the universities, society constructs new benchmarks for success. For example, in my experience, some consulting class or clubs on right on the metrics for business school. Yet, there are so many students, but so little spots, what can you do about it?

Hence the narrative becomes — You compete, dirtily or not, you need to. The successful ones get into the elite circle, while the rest of students feel left out and anxious.

Due to the policy that ensure fairness and ,accessibility Tawianese university are so cheap. But the resources become so scarce for each that I can hardly call it “an education” but spending your four years down the palm trees. This low quality education fosters a sense of unfulfillment and existential crises among students, who miss the clear direction previously provided by high school grading (Our existential crisis 😉).

The intense competition for quality educational resources amplifies these insecurities, creating a harmful zero-sum game. Those who fail to secure resources experience increased anxiety, exacerbating feelings of inadequacy and uncertainty about their futures. Even for those who do manage to get these resources, there's the persistent fear that they might not be so lucky next time.

This uncertainty drives a need for maintaining an elitist status as a protective measure, ensuring their place in the hierarchy. This cycle reinforces a toxic educational and societal environment where “elitism” become a an in-groups to protect their resources and status through maintaining a certain reputation. 2

Educators: Negative motivation transmits insecurities

Our educators also contribute to the problem. They often carry their own insecurities and a sense of elitism, which they pass on to us. Additionally, our culture tends to emphasize negative motivation, further complicating the problem.

For example, NTU professors often told us, "Don’t make mistakes, because your boss will not tolerate them!" or…

But how can a human, realistically, not make mistakes. And why do our professors think we portray us to be so incapable before we even step into the real world? If we haven’t yet meet the real world, and people who do say we cannot survive already, how can we even believe that we can?

Contrast this with my experience in the Netherlands, where teachers refer to us not merely as students, but as "future researchers." Professors at Minerva similarly uplift us, saying that we are special. They tell us that we will not only excel in our workplace, but we are the change makers in the world. We are challenged to think “big question” (e.g. how to solve world hunger?) in school, encouraging us to think broadly and ambitively.

Setting aside the actual quality of the educational content, the trust and belief conveyed through encouraging communication and a thoughtfully selected curriculum make a significant difference in building personal confidence and future vision. One environment tells me I might not even survive a job; the other assures me I can change the world.

Now, I start carrying that sense of confidence to make a meaningful impact on society. How many educators in Taiwan truly have that same confidence in themselves?

Peers: Elitism as self protection mechanism

When I was in job search anxiety in my freshman year, seniors told me, “If you are a Minervan, you are more than fine. Look at all those seniors. If you have problems, let me help you prepare interviews.” After I got in Meta, I gave back, helping most juniors interviewing for Meta too.

Yet, in NTU, I constantly heard story of blackmailing and gossiping each other as the high quality resources are so rare. Millions of toxic gossips on work and relationships.

Once, someone told me, “Someone say something bad about you. Do you want to know?” Why on earth will a person in this community give back?

This culture taught us to not trust and not help each other. Instead, we need to brand ourselves as a “elite” so we will not be looked down on. Being considered an elite doesn't guarantee happiness; it merely serves as a shield against being perceived as a loser—a concept deeply rooted in the single metric of success. This external pressure, combined with fierce competition for scarce resources, perpetuates a message: you aren’t good enough, you are incapable. Those who really want to excel but cannot really believe themselves become an elite. A elite that needs to defend themselves for their deep-seated insecurities.

I find this a real tragedy in our education system.

And I said, fxxk this, because…

What if the collective insecurity is false?

I think this collective insecurity is entirely misplaced.

My older doctor cousin once told me, “Why do you think you can go to America? Why do you think you can compete with kids in Harvard?” I might not possess the traditional "smartness" attributed to a doctor, but her comments ended up to be quite wrong.

Walking around international community, I learn that, sure, we might not as hard working as Chinese kids, but we are still hardworking, creative, and friendly. We are chill and playful. We easily win friends and we come up with creative solutions. Do you think problem solving only requires skills and hard work, or do you think network and creative solutions matter too? Is success genuinely dependent on a single metric?

Rethink my anger toward elitism

Initially, I felt so much anger and annoyance towards the elite. (This post was initially a complaining post haha) Why can’t they be nice and collaborative? But I've come to understand that they are merely products of systemic problems. If life is already so hard, it’s actually quite sad that they need to handle the individual and insecurity complex.

However, most elites will probably remain the same as molded by an educational system that cements their status early on. Minerva has significantly transformed many "elite kids" from Taiwan, but this transformation takes four years, a fun and talented international community, and a committed educational framework. If we can't offer these to Taiwanese kids, what hope do we have?

I wish to see a society that our local Taiwanese kids have genuine confident to thrive internationally, with the kindness to help out their own community.

What really is education?

I talked to our Minerva founder if it is possible to have a really good technological solution to disrupt higher ed. He said no, explaining “because education is about changing the mind of a person.”

And I realized, when we thought of education, we always think of knowledge and skills only. But in my entire Taiwan education journey, I’m still, being educated about ego, competition, and insecurity —elements that shape the elitism and nearly shaped me.

True education should teach us the ability and skills to thrive internationally, the kindness to support each other, the inclusivity to appreciate different talents, and the passion to pursue our curiosity. That change in individuals, is the real education I want to see.3

"Minerva is a reformist. Reform is a very slow way to make changes. Also, it takes years to gather data to present to people. Even then, they’ll point to confounding variables, saying Minerva graduates succeed because of factors X, Y, and Z," he explained.

I then asked, "How can I disrupt higher education?"

I felt comfortable sharing my vision with him because Minerva itself is a crazy idea.

How to disrupt higher ed?

As a businessman, he approached the question from a business perspective.

"Find an underserved market and capture a larger share there," he advised.

What? I'm a CS major. It’s actually wtf to me.

He elaborated, "Target the top 1% of performers (sophomores and juniors) from B-list schools as the new Ivy Leagues and compete with the Ivy Leagues. Provide them with a high-quality education and support them with employment guarantees. These students are equally smart but lack the arrogance of Ivy Leagues. If you gain enough traction with employers, the old system will have to adapt."

Though there were many hypotheses to test, this idea intrigued me. It seemed like a simpler solution than building something like Minerva, as it bypasses the extensive student screening process and the creation of a four-year curriculum and experience.

I then asked how he started Minerva.

How did you start Minerva?

Ben revealed that he knew nothing about accredited programs or formal education when he began. He simply aimed to design a better curriculum for his school, an idea he paused for 13 years before pursuing it again.

"So, what was the first step you took?" I inquired.

"A spreadsheet," he replied.

What? A spreadsheet? Like Excel from the last century?

"Yes, a spreadsheet for a business model: estimating customer payments, acquisition costs, and funding needs," he explained.

Oh, I see. Hmm…Not product first? This confused me because tech people in San Francisco usually prioritize product development.

"What did you do next?" I asked.

"Raised money."

"But that was your 4th company, so you had a great network to start with, right?"

"Yes, I had a great network, but the idea was so crazy that no one believed in me. It took a very long time," he laughed.

I laughed too. It seems everything is difficult anyway!

Learning Points

• On finding like-minded people: I initially thought my idea was too crazy to find others who shared my vision. However, I learned from Ben that by putting myself out there, it's possible to attract people with similar vision, as he didn’t know anyone in his founding team initially.

• On the timeline it will take: I realized my vision might take more than a decade to materialize. I might need to create and fail multiple times to understand what to do and how to do it.

• On the difficulty of starting my own solutions: Regardless of my starting point, I understand now that it will be challenging. Everyone has a hard time anyway!

• On what to do now: Talk to more people and put out my work more. I need to attract like minded people.

• On feasibility: The idea Ben gave me sounds ambitious and difficult, but considering my credentials and background, I believe it is possible able to implement it in Taiwan. My biggest challenge is my preference for technical work over business meetings. Is there a way to navigate around that?

I interviewed education leaders and successful edtech entrepreneurs.

• Why is advanced technology and the information explosion still unable to transform our current system?

• Even though Minerva sets an example of a good edtech model, why doesn’t the system change?

• Is “breaking” the current system the only way to force changes?

• What are the ethics of disrupting the current system?

Honestly, I don’t know if those professionals take me seriously or think I’m just a crazy idealist, but I don’t care. This is the problem I want to work on. I have started to get some clues. Someone told me that cost might be a breaking point for the US education system because it’s so expensive. It means that if I can leverage very good tech, making my product/institution scalable, there is an opportunity.

Here are some ingredients I gather.

• I need some elements from Buildspace, where people commit with strong internal motivation, and top performers will be so good that the entire world will look at us.

• I need some elements from Minerva, where real social learning and community still happen.

• I need some elements from Final Round AI. When everyone wants to use the AI in their interviews, it will force the old system to break.

So first things first, now I only have passion. How can I get the world to look at me? Wait I might actually know. Based on writing this blog and seeing fame in San Francisco, I know the right way to do it is consistent experiment for either products or personal brand. So, it seems that I just need to keep building learning products like the way I’m doing now.

Damn…I feel that Esther is going to be so poor for the next few years already. Also, I’m praying for like-minded people to drop in front of my door every day.

But my friend claims our technology should allow us to play with curiosity and multiple uses of it. It should allow us to explore but with a designated route.

Making the design unclear sounds dangerous, but it is an interesting contrarian idea to explore. Products like Padlet, Heptabase, Obsidian, mmm are some examples that present her ideas. They provide some mediums (like nodes, edges, website interactive elements).

Though I’m not a person who usually uses apps to look for pleasant experiences (I only played PSP during middle school and that’s all), her idea is intriguing. It is because if my ultimate goal is to make learning an experience itself, creating better digital experiences does matter.

As I’m building interactive elements to help people learn concepts better, I do learn the power to engage multiple forms of stimulus, colors, visuals, interactive responses, sounds, can be very powerful. Both for digesting knowledge and learning with better motivation.

Current learning mediums we have today

Here, we need to talk about medium, and I will only spend time discussing digital forms of mediums. Most of the learning mediums we encounter online are either texts or videos. Texts from books or blogs, and videos from recorded lectures and YouTube. If we discuss programming education, we also have running the code online as a medium.

Some traditional mediums can be good if carefully designed, like 3Blue1Brown, which is able to engage sounds and very good visual breakdowns.

Some companies like brilliant.org are able to engage user interaction to learn and play on devices in a creative and crazy way.

Because of my friend’s question, I start to wonder if there is a possibility to combine the two: some tool that we can give people the power to create some tailored interactive learning experience?

Padlet, Heptabase, Obsidian, Notion are an empty playground, but their interactivity mostly focuses on creating a new note page and how to connect the knowledge among each other. Can we make it more interactive like Brilliant.org? When knowledge is being presented as an experience, not as an input?

Brilliant.org is so good at coming up with interactive components to help people "witness" a concept. Can we make it more general like a note page that people can create and integrate?

Yeah haha... I know this is a very hard problem because generalization might ruin customization. But the thing is if AI can generate images, codes, and respond to people, the generalization might not be “generalized.” For instance, when I’m building my tokenizer app, most of my code is generated by AI. This suggests there are indeed opportunities to allow AI to use texts, videos, and code to create a better digital learning experience.

Digital learning idea today

Today, my friend and I were ideating on how this alternative medium might look like. For instance, what if I taught a class on how to prepare for interviews on algorithms and data structures?

Every day, people who take the class will need to go to the same cafe (habitual environmental setup).

1. They will send a photo of the coffee to me or they can drink the coffee animation.

2. They will prepare a pen and pencil for the problem, with an example of my notebook on how I did it.

3. They enter a problem with AI-generated solutions in the background.

4. They get 20 minutes to think and write down their ideas and solutions.

5. They will submit their solutions with AI or my feedback.

6. They will get some habit tracking visualization to see their progress.

What I like about this idea is that it maximizes the fun experience for learning on a subject people usually perceive negatively.

However, I’m not quite satisfied with the specificity since it teaches nothing on how we can think about solving algorithm problems because it’s not in enough depth of the subject (My dillema on Why most AI curriculums suck).

Compared to the tokenizer website I show you in the beginning, they are both visualizations with interactivity, but one allows learners to interact with the knowledge, another with the experience. Since the idea is early on, I still allow the idea to have space to iterate and grow because the overarching question: How can we rethink learners’ digital learning experience? is quite meaningful to me. And I think both the progress and the experience of learning both matter.

Real learning is like going to a gym

Andrej Karpathy said learning should be like going to the gym, and all these TikTokers who make learning videos only make people feel good without learning in depth. I strongly agree with it.

Learning is not supposed to be fun. It doesn't have to be actively not fun either, but the primary feeling should be that of effort. It should look a lot less like that "10 minute full body" workout from your local digital media creator and a lot more like a serious session at the gym. You want the mental equivalent of sweating. It's not that the quickie doesn't do anything, it's just that it is wildly suboptimal if you actually care to learn. — Andrej Karpathy

That’s why I criticize my own idea so hard. Viewing learning as entertainment like a 20-second TikTok is quite different from the joy of figuring out difficult and complicated concepts through some sort of struggling. Their entry barrier to start is also very different (Andrej’s video has a very high entry barrier but very high quality).

Is there any way we are able to inject that effective learning and sweating process with a simple, fun, and easy-to-create medium?

Worst of all, many of the AI edtech products are creating a new workflow or behavior that demands extra work for learners and teachers. Though they built feature after feature, I find most of them, tbh, terrible.

After I went to TCEA, and observed how great products like Curipod are, I started to learn what means to be “user-centric” in the AI field. Though they also generated a curriculum, the curriculum is carefully matched with the state’s requirement, such as the State of Texas Assessments of Academic Readiness (STAAR), that most teachers resonate with since the schools are promoting it.

The popular AI live grading is built for big classes but still allows teachers to check before sending it out. Spanish feedback is integrated because of the large Mexican population in Texas. In the Kahoot game, student’s names will be masked to protect students and have fun seeing their newly assigned names like an “imagined giraffe” that students can discuss with each other.

I’ve seen lots of great engineers and designers in YC. Their products are much more fancier, but they aren’t down to the ground even if they think they do. Tbh, I think that my ideas aren’t down to the ground after seeing Curipod too.

Hence, I decided to build things in the reversed order. I started without AI. I built this curriculum to teach tokenizers through Andrej Karpathy’s video.

My idea was simple: The best content online is usually 2+ hr videos or long-ass technical blog posts that freaked people out to even get started. But if we can engineer it a little bit, we can still introduce the content in-depth but with a friendly format like waitbutwhy. 

Therefore, I built a prototype to experiment with this idea: Build a tokenizer from scratch Part 1. 

I started with the minimal AI possible. I handwrote the curriculum and picked important concepts to expand and discuss. I used my tone to write and only asked AI to generate some code for interactive components.

And then I asked how people think about it. I presented this to a house full of ML researchers, in a talk in front of first-year students with only a tiny understanding of CS, in a deep learning class with undergrad students, on my Twitter, to regular software engineers, and asked multiple professors for feedback.

The feedback and ideas vary. People say,

“It's like W3 but not only for web development!”

“The ASCII code computation is confusing.”

“I love that I don’t need a separate notebook to learn. Maybe you should find a way to make all the code blocks build on top of each other.”

“I think it really chops concepts into small tiny parts. You give us mini puzzles to think through.”

“I don’t think first-year students understand the tokenizer. You should explain it more before doing the interactive part.”

“It’s playful and informal.”

I don’t call this process “gathering user feedback.” It is a process of imagining how we want to learn with more joy, curiosity, and imagination. We are collaborating on this imagination.

I ask so many people because I want to reach the most optimal format and content first (and of course, also to stroke my ego), and only think about what AI can help with later. I believe learning should be centered on learners first and gradually see which part AI can replace and do faster and better like Curipod does, not having AI first and trying to get it close to the learners. How do I want to learn? How do you want to learn? AI-first might be more convenient for engineers like us to do, but it easily loses the core purpose of learning.

People ask me, “How far can AI write an article like this one you do?” I realized it’s still very, very far away. It can say and scratch the surface of some knowledge, but GPT has a uniformly fun personality (like, let’s delve into!), and is unable to prioritize important concepts based on learners’ level of understanding and potential confusion spots. That’s why most AI-generated curricula suck. I hate the way we are learning now through 2-hour stressful YouTube videos, lectures, and long, difficult Medium articles already. I desire high-quality, fun, and challenging ways to learn. (AI can generate simple codes well though, which is very beneficial for me to build interactive components.)

But of course, it’s not fair to say that AI curriculum sucks. I spent a few hours learning and seeing my confusion spots, spent 20 hours writing and polishing my tokenizers with jokes that learners will resonate with, and revised, revised, revised after feedback. AI only has such limited information and understanding to start with compared to me. It is for sure much worse than I am. Therefore, the next step for me in this project is to find annoying tasks and outsource them to AI.

Learning is like camping in a forest. Sometimes it’s delightful when I discover that the visitor center sells S'mores. Sometimes it’s stressful when I realize our tent isn’t waterproof during heavy rain. Sometimes, it’s confusing as to why the fire doesn’t ignite. Sometimes, I bow in awe when I see a majestic deer. But they are all part of the learning journey.

Indeed, they aren’t fancy tech people you will meet in San Francisco, nor aren’t they fashionable designers you can appreciate in New York, but they are really really kind and real. They thought and complained about kids as kids, not some far away “end users” that the Silicon Valley think. As I was always annoyed by ambitious, intelligent, but low EQ men in San Francisco, it feels safe and relieved surrounding by energetic old ladies.

“The amount of women I saw today was more than the sum of them I saw since living in SF.” It doesn’t feel real.

When I saw all these teachers talking about how they think about using AI product in the classroom, many sf startups look so dumb when they post “we have a users use our product for 70 days!” “My ARR is bla bla bla!” That difference of magnitude in impact is no where to be compared. This is also my biggest critiques on sf. When they talk about users, they are mostly someone far away on the screen or as a number that it’s difficult to care about their wellbeing genuinely. But when you are making products to help teachers and talk to them walking towards you, I doubt you feel morally comfortable if you only maximize your revenue and “do research” on your impact.

Another thing that shook me was seeing the passion from all walks of life. You know when we think of passion, the newspaper only talked about powerful and famous people, like successful entrepreneurs, researchers and politicians, Bill Gates, Einstein, and Obama. But if you see those grey hair teachers sitting together trying to use their Sunday morning learning about how to use AI for their classroom, you will feel speechless too.

They might not be as educated as I am, they might have terrible taste of dressing, and they might never get famous or respected, but they also care about their work as much as I am, as much as any other famous people are.

I talked to a Taiwanese-American VC about my frustration about education scene in SF (or in general). It feels small and no one cares or rigid, stupid, and unchangeable. It just felt impossible for me to work in edtech and be great engineer at the same time. I told her I wondered if I should give up for now and thought about it later. She told me not to, and told me her story from Stanford business school to be a 4th grade school teacher (unbelievable for a taiwanese family), and now, an edtech VC.

From the teachers and the VC, I feel humbled, honest, and hopeful. I grew lots of respect for those nobodies who quietly contribute the society, my arrogance of my intelligence and achievement has softened, and I learn it might be possible to build a career that is impactful for others while being fulfilling for me.

As I always have a hunch that I might be the chosen one, I’m definitely on the right way.

I often think I cannot learn a lot of things because I think I’m not smart enough or don’t have enough background knowledge to do it. But my algorithm and data structure professor proved me wrong, and my high-level math professor did it too. Hence, I wonder if is it really because they are too hard to understand or if it’s just because I haven’t put my heart inside.

In a civil lifestyle, if I find a book hard to learn, I can go work on something else. And I’m encouraged to give up because of all the distractions present in my life. But if I’m in an environment where I have no other choice (no phone, no internet, no hangout with friends), will I be able to do it? And reading AI papers during a camping trip seems to be a romantic and pretentious idea anyway. So I’m for sure going to give it a try.

I bought two NeurIPS award papers on AI, The Art of Doing Science and Engineering by Richard Hamming, Invention and Innovation by Vaclav Smil, and The Innovator's Dilemma by Clayton Christensen.

And then, the reality is so much less romantic than I thought, because I was super busy surviving. In the night, it was freezing cold with rain penentrating over our tent. We needed to wake up in the middle of the night to move away tent from the rain pond in the dark, thinking if we should go home tmr. The moment I had some spare time, my brain was too tired to digest anything…

Well…less romantic than I imagine but still inspirational. I was able to finish the two AI papers and even understand most of the math. I also find connections between The Art of Doing Science and Engineering and the AI papers. I was like: am I this smart???

Look at me.

I read in the tent with tiny bugs, learn in front of the ocean, take notes inside the car, and even concentrate in front of the campfire with my headlamp (?)

Hey hey, weirdo friends. I know you are thinking that, wow you are such a nerd! No! I met the biggest rainbow in my life, meditated in the sounds of waves, watched a sky of constellations, and said hi to banana slugs eating grass.

I learn about my potential to learn on this camping trip. Most of the time, I don't want to read the math section in the paper just because I lack patience. I easily find a book boring because I want to feel stimulated as quickly as I can. However, when I wasn't given a choice, especially in the cold tent where you don't really want to do anything else, I cultivated the patience to digest, took the time to read the same section again and again, and freed my mind from other daily worries that distract me. And then, bam, a miracle happens; I become smarter! (aka deeper learning happens)

This is my story about learning AI without a machine hehe. I learn more deep learning when I am able to learn deeper without a choice. (sorry about my dad's joke)

What do changes mean to me?

I want to redefine why we learn. So far, most of us associate learning with assignments, grades, job searches, and promotions. Society views learning as a means to an end (survival, reputation, success, obligation). I think that's completely wrong. Learning is an experience worth pursuing, much like the way people go to an amusement park just for the experience.

However, most people pursue learning as a transaction, not an experience. To learn is to show that you are educated and smart. To learn is to secure the next job opportunity. To learn is to pass the next exam and meet the expectations of our parents. Learning should be valued as an experience, not a transaction, and the world is filled with violations of this principle.

When I witness the violation of this principle, I don't see it as an opportunity. I don't say, "Interesting, let's gamify the product so that people will be willing to learn for five more minutes. Let's build a new product to boost learning efficiency." Learning is very precious to me. I'm not excited about creating smarter and faster humans. I'm not interested in encouraging people to learn more or for longer durations. I'm here to unleash that innate desire to learn.

When I see the dying of the internal desire to learn, it hurts. It's the death of another pair of sparkling eyes. I see a tragedy. The joy of learning should be a universal human right that everyone has access to. Learning is very precious to me.

What would the paradigm shift look like?

When I was in my first year of Minerva studies, I was puzzled by an optional math question. I opened the door of my classmate's dorm and asked her if she could explain it to me. She said, “Yes! I’m interested to know too. Let’s think about it step by step.” And then, as we finished working through it on paper, she held it in front of her face and showed it to me. She stood before me, with sunlight filling the room, and she beamed with joy. “Math is very beautiful,” she said. That’s the most beautiful moment I’ve seen in my life. The world should look like this.

Growing up in classrooms filled with mathphobia and hatred, was like living in 1984; you were not allowed to say you loved math. But she was the first person who shattered my perception because I was finally allowed to think and say that math is very beautiful. We should be able to appreciate the beauty of knowledge & wisdom as it is, not for any negative social perception or utility. I imagine a society that allows genuine love and appreciation of learning for all curious beings to finally feel free.

Inspired by Bret Victor - Inventing on Principle

If it is fixed, why do kids ask so many questions about the world, while adults don’t? It might seem obvious that the distinction is that our curiosity is mostly dead, but not theirs. And schools are responsible for that. But that’s more of a self-imposed constraint than reality.

One of my friends told me that she cannot motivate herself like I do because I am a person who can function well with internal motivation, while she relies on external one. Hence, she always needs accountability and social learning to move forward. This is the constraining thought. This thought is almost identical to thinking, “I’m too lazy to self-learn.” Relying on it gives you an excuse to free yourself from the responsibility to explore your own curiosity.

Instead, ask yourself, “Given that humans are innately lazy, how can I still motivate myself to self-learn?” This is the moment when your curiosity is able to breathe, and you welcome it into your life.

Curiosity is a muscle that you will need to actively cultivate — a lifestyle that you adopt by discarding others, including your self-expectations and your sense of control, such as your desire to appear intelligent and feel successful.

Many intelligent people like to brand themselves as curious when they are asking questions at a dinner party. But few of them are willing to give up what they already have and only follow that trace of curiosity. The sensitivity of that trace is the core practice of exercising the curiosity muscle.

As for how to cultivate and grow your curiosity muscle, there are many ingredients, such as having enough free time, exploring different tricks to motivate yourself, being in a solitary environment, and allowing yourself to fail and feel dumb.

I never really thought of starting my “self-learning” journey. Four years ago, I was just a bored exchange student because I had finished all the enjoyable things I could do (I only had two hours of class a week). I started to self-learn for a week, got distracted for a month, felt like coming back to experiment, and failed again. Yet, over time, I started to listen to my instincts about how to learn and what to learn. With enough boredom and loneliness, I constantly exercise curiosity muscle and experiment with different self learning tricks. Four years later, people are amazed by my motivation to keep writing and doing random projects, it’s just something I’m used to because I can bring out all the tricks and easily recall that curious lifestyle again. That curiosity trace has become a well-supported web.

Some people told me that following curiosity is too hard, so they need to start with goal setting. But then you probably start the journey wrong. In fact, most of us care a lot about goals and efficiency consciously and subconsciously, so even if you aggressively seek curiosity, a sense of progression will still be there. Hence, to really learn what you like and want, try to let go of that expectation as much as you can, so your curiosity has enough nutrients to grow. More often than not, your curiosity has better guidance on how and what to learn than your own expectations.

Successful people, like Paul Graham, like to market the idea that curiosity will lead you to perform great work. Therefore, curiosity is a charm among the well-educated circle. However, it’s not an ingredient that will guarantee success, because curiosity might have a direction or not. It’s all about how to interact with and harness it for your use. Besides, if your goal is to laser focus on achieving great work and curiosity is a tool to achieve it, then you may have already lost that curiosity in the first place. Great work is just a by-product of curiosity because how your curious instinct guides you is nowhere to be predicted. But that’s also why many new ideas and unexpected progress are discovered when people dive deep into their curiosity. Yet, whether you discover or create anything great or not in the end, you will feel content, passionate, and happy, no matter what you are pursuing, with an in-depth understanding of yourself and your career.

Hi weirdo friends! Today I’m really bored so I pretended I’m Paul Graham and wrote this post. How do you feel about this authoritative daddy tone? I feel I was 70 years old when I was writing. And end up feeling like I at least need to draw something or I will die. I love learning how to write in compact yet logical way like he does, but please bring some youth to me.