Founder of Gamma ($100 ARR AI-powered slide deck software, started in 2020 now valued at $2 billion)

• Insight that led to founding the company: “On Google slides I was spending 80% of time on the look and feeling, 20% of time on content. It should be backwards.” Paul Graham commented “surely the thing that the slide deck is describing is more valuable than the slide itself.” 
• Emphasized the importance of making the first 30 seconds of the product incredible - was the key factor leading to product viral growth and product-market fit.

Marc Andreesen (a leading VC and former entrepreneur):

• "AI is the first monumental rearchitecture of what is a computer in 80 years."
• "The scenario where you have only a few big AI model winners is like if we just had mainframes instead of desktop and laptop and mobile and watch. It would mean basically the top AI models are the best across many things, and cheapest, and most power efficient, and fastest, and easiest to adopt and use for every scenario. And we are going to want AI infused in everything."

Founder of World Labs (Fei Fei Li, a major contributor to AI progress who is focusing on developing AI models that are spatially rather than linguistically intelligent):

• "So much of our intelligence is built upon visual perceptual spatial understanding. Not just language (the foundation for LLM models powering the current wave of AI). I think they are complimentary."
• Use cases for spatial models include: robots, disaster support like putting out a fire ("words alone aren’t enough"), gaming, scientific discoveries (DNA required 3d visualization)
• "Robots are more like self driving cars than LLMs. 20 years from self driving in desert to in the streets, and we’re not even done yet. And self driving cars are much simpler robots - they’re just metal boxes running on 2d surfaces, and the goal is not to touch anything. Robots are 3d things moving in the 3d world and the goal is to touch things."

Slack Founder on good software design:

• "Good software is less about reducing friction and more about reducing people feeling dumb or having to think when using your software.”
• On tilting your umbrella to let someone else go by more easily: “Bezos famously said ‘your margin is my opportunity’. At Slack we said ‘your non-umbrella tilt is my opportunity’. Your failure to really be considerate and empathetic about other people’s experience can be our critical advantage. Slack wouldn’t have grown the way it did without those little conveniences. ‘Tilt your umbrella’ is still on company swag.”

What’s the problem with most schooling today?

• “Educating kids is the most important thing we do as a society, and it hasn’t been changed in 200 years. Teachers are great, but the 200 year old model of one teacher in front of the classroom most of the day is the problem. The teacher in front of the classroom is like the worst way to teach people…"
• “In a time-based model (a certain amount of time allocated for each part of the curriculum  before a test and then moving on regardless of how well you did), only a small set of students can master the material and get a 100%. In a mastery-based model, everybody can. The model is the problem.”

The Opportunity AI Unlocks

• Highly tailored teaching has not been scalable given the teacher/tutor to student ratio it would require. Gen AI has changed the scalability constraint
• AI enables dynamically generating lesson plans precisely fit for each student’s knowledge level and interests. 
  • Example 1: Student doesn’t read much but likes Marvel movies and has friends on their soccer team. Can then create a story for them to read about them and their friends saving the world, and move up the reading comprehension.
  • Example 2: Student finds fractions boring, but likes fashion and shopping for clothes. So teach fractions through fashion rule ratios. 
  • Example 3: Student doesn’t like statistics, but loves baseball. Teach statistics in terms of baseball.
  • Example 4: To learn world history, students generated songs with AI. Students remembered what they learned better.
  • Example 5: Student wasn’t interested in WWI. But liked Taylor Swift. So GenAI created an analogy to when Kanye took microphone from Taylor during Grammys, expanded on that, added names of who was who for WWI. And the kid went from being deeply uninterested to literally saying “ok I am going to remember this forever”

The Pitch

1. 2 hours is all you need. "This whole concept of we need to have kids spend 6 hours a day plus homework for 12 years. It’s just not even close to true. And according to learning science people have cognitive load limits, so they can only learn so many new chunks of information in a day if you want them to retain it. Your kid can learn everything they need to know in 2 hours a day. The amount of time they need to spend is so much less than what we spent or what parents expect. If your kid is 1 year behind, your kid is only 20-30 hours behind at Alpha. You can teach an entire grade level of one subject in about 20-30 hours (compared to the typical 180 school days 1 hour each and homework). It’s 5 to 10 times faster. It just doesn’t take that long. And they all get 5s on their APs."

2. Your kid can love school more than being on vacation. “The top commitment of Alpha is kids will love school. Students often don’t like school, the school experience is not fun. We’re going to be able to transform their years in school because they are not gonna sit in class all day. Motivation is 90% of the solution.”

The Model

1. Teachers/Guides: 5 students per guide ratio at the lower levels; higher ratio as students get older to allow more space per student 
2. 2 hours daily of AI-generated course material (there is also an AI vision model used to analyze student interaction with their generated lessons and provide feedback)
3. 4 hours of time post-course material for "workshops" (examples have included students managing a food truck for a year and learning about operations, profit and loss statements, and marketing; writing a musical; working on robotics; practicing public speaking and storytelling in front of real sports journalists)

Double Click - Avoid a Swiss Cheese Student

• Much of knowledge is built on top of each other (addition -> multiplication -> algebra). 
• Without mastering each layer it’s like Swiss cheese: "you’re building this foundation with all these holes, and it’s going to collapse as you get high enough."
• At most schools: “A teacher in front of a 6th grade classroom has to deliver 6th grade material. But many students will not have mastered prior material and need a 4th grade lesson, but they won't get it again. And if they get 80% of the next lesson they will move on even though they haven't mastered the material - the standards are just too low. And then as kids get older the biggest issue isn't that they're struggling with the current material but with the prior material that the current material builds on.”
• Difference at Alpha School: " A student got a 740 out of 800 on the math SATs (a top percentile score). But from that test we saw she didn’t have fluency with a component of division. So we literally sent her back to 3rd grade level math. We’re the only school in the world that will take a student with a 740 and send them back to 3rd grade math. And she then got a 790."

What's Next?

• Scale the software over the next year to support more schools
• Federated schools (such as sports or music academy that use the 4 hour workshops to specialize, or for more academics for kids who love that)
• Like the iPhone (and unlike much of the education model that has remained fairly stagnant), improve the model to get better every year

Level of commitment

• Did the same job almost every single day for 75 years
• Up by 5am and back from restaurant 10pm every day while his kids were growing up 
• Doesn’t like holidays
• "All I think about is sushi...I would make sushi in my dreams” (Good luck trying to outdo someone who is working in their dreams!)

Steve Jobs said when making the Mac: “The [existing computer line at Apple] wanted to build something great; the Mac team wanted to build something insanely great. The difference showed”. Jiro wanted to make insanely great sushi:

• Each of his vendors has to be the best vendor for that specific ingredient. And each vendors  is a specialist (tuna, rice, etc).
• Octopus: used to massage for 30 mins before boiling, now I massage for 40-50 minutes
• Rice: we have a rice supplier who only sells to us (wouldn’t sell to the nicest hotel in Japan). We put a heavy lid on the rice that adds pressure and takes two people to lift. And we serve it and devised techniques to serve it at the right temperature. 
• Apprentice for Jiro: “10 years on fish, then they let you cook the eggs. I made over 200 eggs. Every single one was rejected until I finally made one Jiro liked. He tried all of them”

Miscellaneous

• Didn’t start his own restaurant until he was 39
• "Jiro's restaurant was immaculately clean. It may be the cleanest restaurant in the world"
• When his son opened his own restaurant, Jiro said “now you have no home to come back to. He said that I would be buried in the place where the restaurant is. Failure was not an option. There is no turning back”. When Jiro left home at the age of 9, he was told “you have no home to come back to. I had to work to survive. That has never left me.”