Skip to main content

· 6 min read
Carl Liu
context

Recently accepted to Stanford's M.S. in Learning Design and Technology 🎉, having developed educational technology projects and ventures (see my GitHub), this program represents a perfect opportunity to advance my work! I'm sharing insights from my application journey.

Stanford's graduate programs are highly selective, often accepting only 20 students per cohort. This essay will be a summary guide for how to approach a grad school application at Stanford, and provide some tips on writing a clear and concise SOP.

Note: This guide focuses on Stanford Master of Science applications, though requirements vary for MBA, Medicine, Law, and PhD programs.

Below, I'll outline crucial components of successful Stanford applications, emphasizing the Statement of Purpose—typically the most challenging and important element.

How to Prepare

For master's program admissions, the importance of different components varies by school and program, but generally, the Statement of Purpose (SOP) and Letters of Recommendation (LORs) play the most significant roles. Work experience, GPA, and test scores are less important and are substitutable.

Application Materials Importance Hierarchy


· 6 min read
Carl Liu
Kazem Jahanbakhsh

Goal

In this analysis, we explore how Large Language Models (LLMs) and AI technologies have fundamentally transformed the internet landscape since ChatGPT's debut. By examining market dynamics and how major players are adapting, we present four key predictions for the AI industry through 2025. Our forecast draws from current trends, market data, and emerging patterns in how businesses and users interact with AI technologies:

  1. New Frontiers in Product Design
  2. The Search Will Shift
  3. LLMs as a Market Disruptor
  4. Future of LLM Development

New Frontiers in Product Design

Shifting Metrics at Airbnb

A conversation with a senior design manager at Airbnb revealed a paradigm shift: instead of focusing on Page Performance Scores, the company now measures “Resolution Time”—how quickly an AI agent can meet user needs, whether it’s booking a stay or finding relevant info.

  • Implication: Traditional UI metrics, such as load times and animations, may soon take a back seat to AI responsiveness and contextual understanding.

· 8 min read
Carl Liu

Definition of Liberal Arts: Liberal Arts is intended to provide chiefly general knowledge and to develop general intellectual capacities (such as reason and judgment) as opposed to professional or vocational skills.

Growing up in two different countries, China and Canada, I encountered a common trend: faculties were often divided into the Faculty of Arts and Faculty of Science. However, my experience working as an engineer at Presence, a pioneering AR tech startup, has taught me that what is often underestimated in the tech industry is the value of liberal arts education.

In some extreme cases, engineers believe that hard skills like coding are the only skills that matter, while liberal arts education is dismissed as irrelevant or impractical. However, I argue that this is a flawed perspective. In fact, liberal arts education can be just as valuable as hard skills for engineers working in the tech industry.

Limitation of Engineering Education

Throughout my academic and professional experience in the technology industry, I have come to recognize three major issues that are rarely discussed.

1. Fixed Reward Mechanism

In academia, technical interviews, and in the industry, the standards for evaluating and rewarding engineers are often fixed. Engineers tend to obsess over code cleanliness, optimization of memory and computation usage, and test coverage. While these standards may contribute to the development of better engineers, they may result in less creative problem-solvers overall. In fact, some experts in the field, like Dan Abramov, have highlighted how an obsession with clean code can be problematic. Although there is value in these standards, they prioritize certain skills over others, and consequently, limit engineers' capacity to be well-rounded creators.

Examples of these reward mechanisms include getting an A in a course because your exam answers were elegant, or landing a job offer because you wrote a perfect algorithm that solved a Hackerrank problem faster than anyone else. Additionally, building a better API product than Stripe does not necessarily mean that people will abandon Stripe and use your product.

· 2 min read
Carl Liu

This is an ongoing blog...

Observations

  • Humans unite when there is a common enemy
  • Core values and organization cultures can increase productivity of a team
  • When people are not confident, they tend to use the bigger identity to hide themselves:
    • I work for a big corp
    • I am from a developed city xxx/country xxx/continent xxx
    • I went to xxx University

About Ego

Understanding that our identity is not fixed but is rather a collection of ideas and interpretations about who we are can give us the capacity to, when necessary, think critically about what our ego is telling us particularly when the instinct is there to preserve our identity, or ego, at all costs. The ego has gotten a lot of bad rap, but the truth is that the ego is neither negative nor positive. It is merely a sense of self that allows us to operate and function in the world, giving us an idea of who we are and where we belong. The ego is only problematic when it takes control over us rather than our true selves being in control, and when the things it tells us to do or how it is interpreting events becomes damaging to us and/or others.

References

· 5 min read
Carl Liu

For the past year, I have met a lot of entrepreneurs and engineers. I found a common pattern: these knowledgable people not only read more than me but are very smart in choosing the right information to read and digest.

In terms of intelligence, everyone is similar. What matters more is whether a person wants to learn/read and from which source the person learns (the information input). Just like what people always say in the gym -- 'You Are What You Eat', this applies to information and learning.

Information Hygiene 👩🏻‍⚕️

The first thing to practice is to identify the quality of information. For the modern era, a lot of online media sites and content creators tend to tune their content to maximize the impression rate or view count (if you want to learn how to write news to catch eyeballs 👁 here is a link) a better term to describe the quality of information is Information Hygiene

The following chart is my understanding of information distribution:

· 2 min read
Carl Liu

Definition of Goals

  • Subjective vs Objective
  • Goals can be gamed by blurring its measure of success:
    • ❌Wrong: I made a good product because I think people will like it
    • ✔️Correct:
      • Physical Fact: Humans landed on the moon
      • Economic Fact: People like to pay millions of dollars for a software license per year

First Principal thinking

  • Get to the fundamentals than learning the buzz words
    • ❌Wrong: Learning definitions of Cloud, SaaS, PaaS, Metaverse...
    • ✔️Correct: Understanding FSM "finite state machine", Turning Completeness, Gödel's incomplete theorem... (This is why getting a degree in Business is so less valuable than getting a degree in Engineering. If I give you a book that has the same business strategy as Airbnb, you probably won't build another Airbnb; Whereas, if I give you a snippet of code, whether its 0 or 1, you can justify it in hours)

· 6 min read

Why China?

At the end of 2021, I decided to relocate back to China due to both family and personal reasons. Extended WFH was a huge one: I was losing a ton of passion and efficiency. After evaluating a few opportunities, I decided to join Airbnb, and work out of their Beijing office!

Airbnb Beijing

· 5 min read

I was talking to my friend who does meditation on a routine the other day. I have always wondered why someone enjoys spending the time to meditate. His theory is that the idea of meditation is to focus on living experience because not everyone is experiencing living. He gave me a few examples:

  • When you are enjoying an extremely fancy dinner that costs you 300$, but you know that you are having a technical interview tomorrow morning, you have to go back and sleep early. Are you enjoying the dinner, and tasting every single spice in your food?
  • When you are drinking an amazing espresso shot, and someone is knocking on the door which draws your attention away from the espresso. Are you tasting the coffee?
  • You can stay in the library for 12 hours and chatting with friends on Facebook. Does that count as studying if your library experience is not studying?

· 3 min read
Carl Liu
info

This blog will explain React Fiber Internal Algorithms, we will:

  • Revisit how React diff works
  • Problems with Tree Traversal
  • Morris Traversal

It's recommended to have a basic understanding of React, virtualDOM, and diff before reading this blog.

You can first experience the difference between React Fiber (v17+) and pre-Fiber (v16) by playing with the playground below:

You can tell a huge difference between the two versions, the old reconcilor is very slow, but the fiber one is very smooth.