←
Sue2,087 posts
Sue avatar✓

Sue ✓

@suekhim

Co-founder & CEO of @brilliantorg. Building better ways to learn in the age of AI. Writing about learning, parenting, and the systems shaping childhood.

◷ Joined February 2010▧ 61 Photos and videos
Sue avatar
Sue@suekhim·May 29

AI is making kids dumber. It should be making them geniuses. Introducing Koji, the first AI tutor that gets kids to actually think. 👇

2:15
Vybhav Ramachandran avatar
⟳ Sue retweeted
Vybhav Ramachandran@Vybhav·Nov 10

As I've been saying, Learning has now changed! With @claudeai Fable 5.2, my 10 year old son was able to: 1. Virtually build a RC car, with the components that he wanted 2. Train a neural network (by driving it himself for 15 minutes across various track types, and observing loss curves) and test it on the actual track. Here's his car, driving a course. He's super excited and even asked me if the model is "overfitting", which was surprising. I asked him where he learnt about overfitting, he said @brilliantorg Learning has now changed!

0:42
Sue avatar
Sue@suekhim·Nov 3

Yesterday: Read my 3 year old Oh, The Places You’ll Go! 3 yo, today: I don’t want to go to school. I want to go to cool places. I WANT TO GO TO COOL PLACES!!!

Post attachment
Sue avatar
Sue@suekhim·Sep 25

A great learning experience has a sense of flow. Having a tutor take 5-6s to respond kills that. Switching to @wafer_ai for inference made Brilliant’s tutoring sessions feel fluid and fun again, and every session metric improved along with that.

Post attachmentPost attachment
Gokul Rajaram avatar
⟳ Sue retweeted
Gokul Rajaram@gokulr·Sep 25

This is an awesome customer story (I’m a proud @brilliantorg investor and a huge supporter of their mission). We heard the same themes over and over again when speaking to @wafer_ai customers.

Post attachment
wafer avatar
⟳ Sue retweeted
wafer@wafer_ai·Sep 24

@brilliantorg 3x’d their tokens per second and cut inference costs by 50%, by making its AI tutor fast enough to answer in real time instead of speculatively prefetching answers. To keep Koji, their AI tutor, responsive during math and coding lessons, the Brilliant team was speculatively prefetching AI-generated responses so students wouldn’t have to wait. On Wafer's dedicated GLM-5.2 endpoint, Brilliant reached ~250 ms time to first token and 300+ output tokens per second. This was a 3x throughput increase over their previous provider. That speed let Koji answer learners in real time, so the Brilliant team could stop pre-generating responses and remove the prefetching layer. Read how Brilliant made the switch! 🧵 Link in Thread

Post attachment
Gagan Biyani 🏛 avatar
⟳ Sue retweeted
Gagan Biyani 🏛@gaganbiyani·Sep 22

Today, we’re launching an ambitious new school called The Horowitz Andreessen Academy. Based in San Francisco, The Academy serves the most promising young high school graduates. We think this can be an elite institution that attracts top tier talent. One that prepares students for the future rather than remaining stuck in the past. The #1 goal is to help students learn to build, which is the most important skill in the AI era. They'll learn primarily by pursuing their own projects, either individually or in groups. There are classes and guest lectures, too, from some truly amazing people who have built modern-day Silicon Valley. The Academy is designed as a network, since that’s the reason students go to school in the first place. Core to that network are our 10 Founding Partners: Anduril, Anthropic, Coinbase, Google, Meta, NVIDIA, OpenAI, Palantir, Replit, and Stripe. The network includes over 50 hiring partners and over 200 speakers and mentors. To join as a hiring partner or faculty member, you can apply on our website. We raised $42M in funding led by @a16z. I'll be CEO and @pmarca and @eriktorenberg will join me on the board. Applications are open for our Founding Class Fellowship, which will be one year and tuition-free. Eventually, pending regulatory approval, we plan to offer a two-year program that charges tuition, similar in cost to an elite private university. We're looking for the most unusually ambitious young builders on the planet. Come join us in San Francisco: theacademysf.com/

1:57
Sue avatar
Sue@suekhim·Sep 17

"Students on University of California campuses may struggle with basic math, but they are graduating at record rates."

Post attachment
Sue avatar
Sue@suekhim·Sep 16

Task 2: 3-dimensional thinking. Will get to writing this up in a few days! If there’s interest, I’ll also at some point fill in details from sessions #2-#10, which covered a big range: logic, code execution, goblin currency conversion, and more.

Sue avatar
Sue@suekhim·Sep 16

They still have lots to learn about organization. If the problem were changed to a 3-high tower with 4 colors, it would overwhelm their current method. But it was nice to see them break down the problem differently once they realized brute force wouldn’t be enough.

Sue avatar
Sue@suekhim·Sep 16

They rushed to find the 1 yellow and 3 oranges, which they did with about 20 seconds left. They found every tower with no duplicates, all on their own, with zero help! They weren’t certain they'd found everything. But it was cool to hear them hypothesize that maybe there were 0 more, or 3 more (but it couldn’t be 1 or 2 more).

Sue avatar
Sue@suekhim·Sep 16

They decided to organize all the towers into 3 groups based on the color at the top of the tower: green, orange, and yellow. When they did that, they observed that there are 9 greens, 8 yellows, and 6 oranges. This was amazing to see, because in prior sessions they had not demonstrated any ability to organize information. They then discussed amongst themselves that maybe there are more of everything, but we at least know that because of symmetry, there has to be at least 1 more yellow and at least 3 more oranges. They now had 5 minutes left.

Sue avatar
Sue@suekhim·Sep 16

In the first 15 minutes they came up with 23 towers. Then they got stuck and felt like giving up: “This is everything.” “We’ll never know if we got everything.” They wallowed in their own sorrow for about 5 minutes. Across sessions #2-#10 their combined brute force often got them to a solution, but they were weak at systematizing cases. So we were especially impressed with the next 10 minutes.

Post attachment
Sue avatar
Sue@suekhim·Sep 16

The first task was a harder sequel to the combinatorics task from session #1. Before, they made 3-block towers using 3 colors, with each color appearing once: 6 possibilities. This time, color repetition was allowed: 27 possibilities. Their job was to build a city with every possible tower. We gave them 30 minutes.

Post attachment
Sue avatar
Sue@suekhim·Sep 16

5 yo math circle #11. This time we made it very hard for them, with 2 tasks. It was an hour of watching them truly struggle – and eventually, discover new mathematical superpowers. The materials and instructions are simple, so you can try this at home. Here’s what we did 🧵

Post attachmentPost attachmentPost attachmentPost attachment
Sue avatar
Sue@suekhim·Sep 15

Give a kid a potato and they’ll eat for a day. Teach them to plant potatoes and you’ll have to teach them crop rotation next. Teach them to teach themselves and you’ll never have to talk to them again.