back to indexHow students build with Claude | Code w/ Claude

00:00:00.000 |
All right. Hey, everybody. How are you doing? My name is Greg. I lead student outreach here at 00:00:11.520 |
Anthropic, and I am so excited to be sharing the stage with some of the brightest young minds in AI. 00:00:16.440 |
Just a little context for this panel. So, at Anthropic, we've given out API credits to thousands 00:00:23.680 |
of students to help them build things at school, and so what you're about to see is a very small 00:00:29.160 |
glimpse at what students have been creating with those API credits. It's a very wide variety of 00:00:35.220 |
things, as you're about to notice. Some of these projects are very humorous and funny. Some of these 00:00:40.400 |
projects are very serious and important. Some of these students are working on one project, and some 00:00:45.760 |
of these students have been building an app every single week throughout all of 2025. So, I think if 00:00:52.480 |
I was gonna sum up what I've learned from running this program, it's that the future is really, really 00:00:57.360 |
bright in the hands of these students. So, without further ado, I'm gonna invite up our first speaker, 00:01:03.420 |
All right. Thank you for having me. It's a privilege to be here. My name is Isabel. I'm a senior at Stanford, 00:01:15.600 |
where I study aeronautics and astronautics, and I'm doing my honors in international security. And today, 00:01:20.740 |
I'm here to talk to you about my honors work, which is on finding nuclear weapons in outer space, and how I use 00:01:26.220 |
Claude to help me do it. So, for those of you that may not know, Article 4 of the Outer Space Treaty bans 00:01:34.780 |
the placement of nuclear weapons in outer space. Now, other arms control agreements that you may have 00:01:39.900 |
heard of, like START and New START, include provisions for verification and monitoring. So, 00:01:44.780 |
nations are shown to be compliant with their treaty obligations using inspection systems. We have on-site 00:01:50.780 |
inspections where inspectors will go and look at each other's delivery vehicles and inspect for the 00:01:55.740 |
presence of nuclear warheads. We don't have anything like that for outer space. 00:01:59.340 |
mostly because we signed the outer space treaty in 1967, and there were no technologies to do that 00:02:04.220 |
kind of inspection, right? How would you go about approaching a satellite in orbit that might be 00:02:10.060 |
carrying a nuclear weapon and inspecting it for the presence of such a device? Daunting for the 1960s, 00:02:15.420 |
daunting today. And this became a problem recently, in 2024, in April of last year. 00:02:24.860 |
The Biden administration announced that the United States assesses that Russia is developing a space 00:02:30.220 |
vehicle that carries a nuclear weapon. Now, this was pretty destabilizing for the international 00:02:35.500 |
community. We've had a lot of dispute in the UN Security Council recently about how to handle this 00:02:40.780 |
potential violation of the Outer Space Treaty. Given that we don't have a verification mechanism for 00:02:45.980 |
compliance with the Outer Space Treaty, I started to wonder if it would be possible to implement such a 00:02:50.220 |
system. Particularly given that the US Space Force tracks 44,800 space objects today, how would you 00:02:56.380 |
begin to know which one of those is the suspected nuclear weapon? So this brings me to my research 00:03:04.220 |
question. Is it feasible to perform an in-space inspection mission where you inspect a target 00:03:10.060 |
satellite for the presence of a nuclear warhead on board? Daunting question. It has a lot of interesting 00:03:15.740 |
technical and political facets to it, but for one particular aspect of it, I was able to use Claude 00:03:21.420 |
to my advantage. So I looked at specifically the feasibility of detecting the nuclear weapon with an 00:03:27.900 |
x-ray system. So you fly an x-ray source and detector on two different inspector satellites in space, 00:03:36.780 |
have them rendezvous with the suspected nuclear warhead target, and scan it for the presence of a nuclear 00:03:42.060 |
weapon and board. I wanted to know if this would ever be possible. No one's ever tried using x-rays 00:03:47.260 |
in space. There are interesting questions around whether the space background environment is - there's too 00:03:51.820 |
much noise in space to detect the source signal. So I built a computational simulation to see if this 00:03:58.380 |
would ever be possible. And to do it, I used Claude. I used this very complicated CERN software package 00:04:05.740 |
called giant4. I am not a particle physicist. I did not know how to approach this software package 00:04:11.420 |
and write the C++ code. But I was able to make a desktop application to do my simulation using Claude. 00:04:17.260 |
And it was incredibly exciting. It worked. So what you're seeing in this picture is like a very, very quick 00:04:25.580 |
snapshot of an x-ray image taken in space. And you see a little hole in the middle that shows you that 00:04:33.740 |
there's very, very dense fissile material on board the target of the scan. So indeed, in this simulation, 00:04:39.340 |
there was a simulated nuclear warhead on board the satellite target. The outcomes for this are pretty 00:04:45.660 |
significant and interesting. There are a lot of people in the national security intelligence community in this 00:04:50.540 |
country that are interested in developing this kind of capability to inspect adversary spacecraft on orbit. 00:04:55.580 |
To understand their capabilities, particularly whether they might carry a weapon of mass destruction. 00:05:01.100 |
So having done this research, I actually am going to be able to brief it in Washington, D.C. to some 00:05:06.700 |
policy makers at the Pentagon and state. I'm really thrilled about that opportunity. And certainly, 00:05:12.540 |
the desktop application with this level of fidelity would not have been possible without 00:05:16.540 |
modern AI tools to make this kind of research accessible to an undergrad in less than a year. 00:05:22.060 |
My takeaways for you, kind of as a student, doing research in the era of AI is just that primarily, 00:05:32.060 |
there is no learning curve that is too steep any longer, right? Even the toughest problems, space 00:05:38.220 |
technology is notoriously hard, nuclear weapons, existential threats. We can address these critical crises 00:05:45.180 |
with the tools that we have today with emerging technology. And so I want to challenge all of the 00:05:50.140 |
minds here and other students to think about what are the world's toughest problems? What are the problems 00:05:54.460 |
that you thought were unaddressable, that feel like existential crises to you for the next generation? 00:06:01.980 |
Those are the ones that we should be using our brand new, shiny, exciting AI assistants to work on, 00:06:08.860 |
because that's how we're going to help make the world safer and more secure, or at least outer space, 00:06:15.660 |
I'm going to pass it off to the next presenter now, but if you have any questions, I'd love to talk after 00:06:30.380 |
Okay, so it's kind of tough to follow up finding nuclear objects in space. So I'm going to tell you about 00:06:37.660 |
how I did not know the difference between the terminal and the code editor, and why Claude is 00:06:42.060 |
the reason why I was able to learn how to code. I'm a student at UC Berkeley, my name is Mason Ardidi, and I'll go ahead 00:06:47.260 |
and get started. So I want to talk about what we think of as the traditional way to approach learning 00:06:53.180 |
how to code. I'm going to call this the bottom-up way, where we start by taking basic classes, learn our basic 00:06:58.780 |
skills, and then build apps with those skills. Slowly but surely, we level up our skill set and build apps that are 00:07:04.220 |
more complicated. I learned a little bit differently. I'm going to call this the top-down approach, where I had an 00:07:13.340 |
idea as I get inspired randomly, and I had no idea how to solve it. It was software I've never coded before. 00:07:19.420 |
So I tried to have AI make it for me. Hey, make this app for me. And then when it inevitably fails, I learn 00:07:25.260 |
how to do it myself, slowly but surely learning through different layers of abstraction until I actually 00:07:30.780 |
understand what's going on. Now, where did this leave me seven months ago? It left me not knowing what 00:07:36.540 |
the difference between the terminal and the code editor was. I put npx create next app latest in my page 00:07:43.180 |
file. I had no idea what I was doing. But slowly but surely, I asked, why is this happening? What am I 00:07:48.380 |
doing wrong? And I was able to learn more complicated skills. Let me show you a demo of something I'm 00:07:54.540 |
capable of doing now. Okay, welcome to CalGBT, which is a better way to schedule your Cal courses using AI. 00:08:00.940 |
It's going to question like, show me math classes with a high average grade, since I want to be lazy in my 00:08:07.580 |
math and get an easy curve. Here, it's going to show us five different classes that have an average of A or more, 00:08:13.020 |
And in fact, it even showed us classes with a grade point average of 4.0. Can't really get much better 00:08:17.660 |
than that. Now, let's say it's getting late in the enrollment cycle. And I want to see classes that 00:08:22.140 |
still have open seats. Show me history classes that still have open seats. And this is drawing directly 00:08:30.220 |
from Berkeley time. So it's live data. And here it is, it's showing you history classes, five seats, 00:08:35.260 |
20 seats, or three seats. We can even ask questions that are more deep, like, what is the meaning of life? 00:08:42.060 |
And do with that answer as you will. But this is CalGBT. My name is Mason. And enjoy your day. 00:08:50.460 |
I'll show you another one, which I developed at the Pear and Anthropic hackathon as well. 00:08:55.900 |
Okay, welcome to Get Ready, which is a new way to visualize and understand new code bases. 00:09:03.900 |
Let's take a look at Anthropic's SDK for TypeScript, for example. 00:09:08.220 |
You'll see soon. And we'll be able to interact with the chart and see 00:09:13.500 |
how all of these files interact with each other. So here we have a mapping of some of the most 00:09:19.660 |
important files. We chose not to display all of them, just the most important ones that the users 00:09:24.220 |
will interact with the most. And we have these lines to show how they're interconnected. And we do this 00:09:29.580 |
through the function calls that are actually, like, in each file. So, like, if this demo TypeScript file 00:09:34.300 |
is referencing the batch results, that's where the line comes in. And then over here, we have just a 00:09:39.980 |
quick description on what the file actually does. And we have our comments on the code base. 00:09:45.340 |
Okay, and on top of these two, I built many projects over the course of my learning how to code. 00:09:52.860 |
Now, what is the point of me showing you all of this? I'm not here to brag. I'm here to say that 00:09:57.900 |
Claude is the reason why I was able to learn how to code. Without Claude, without these AI tools, 00:10:03.500 |
including cursor, windsurf, whatever you guys want to use, none of this would have been possible. And the key 00:10:07.980 |
takeaway for me is that you can build anything you want nowadays. You just have to ask the right 00:10:13.100 |
questions, learn through the different layers of abstraction. 00:10:15.900 |
I think this is representative of a new style of building and a new class of builders. 00:10:22.300 |
Where my flow, personally, is I find a problem that I'm inspired by and want to fix. 00:10:27.660 |
I realize the solution is something that I have no idea how to do. 00:10:31.900 |
And then I have a high level chat with Claude, execute steps in the actual editor, 00:10:36.140 |
and then record and post a demo when it's not perfect, hopefully bringing users and revenue later 00:10:41.660 |
on. But this iteration cycle, instead of taking years for an undergraduate degree or doing other things, 00:10:47.420 |
can be one day to one week maximum if you really want to. 00:10:51.500 |
So I'll keep it short and sweet and leave you guys with a couple of things to think about, 00:10:55.420 |
which are on my mind right now. Which is, how can we build to get users and revenue? 00:11:00.300 |
Not for technical perfection and impressiveness. How can I build things as fast and as simply as 00:11:06.940 |
possible? As demonstrated by this prompt, give it to me in the simplest and most concise way possible. 00:11:12.220 |
What ideas actually inspire you? And how can we build it today? And lastly, not on the slide, 00:11:20.220 |
but what does it mean to really know how to code? Does it mean understanding every single line and 00:11:25.820 |
every single function? Or does it mean being able to build something that actually improves people's 00:11:29.980 |
lives? I'm going to continue to post more information. If you want to connect with me, 00:11:34.620 |
you can scan this QR code. But my name is Mason. Thank you guys. 00:11:38.620 |
All right. What is up everyone? How are we all doing? We good? Yeah. My name is Rohil. I'm a freshman, 00:11:54.860 |
or just finished freshman year at UC Berkeley in the MET program, studying EKS and business. So CS and 00:12:00.140 |
business. And I'm here to talk to you guys today about SideQuest, which is a project that a couple of 00:12:05.980 |
friends and I made at the Pear Exanthropic Hackathon recently. So let me tell you guys about a big 00:12:13.500 |
problem today, is AI embodiment. So we see like in Hacker News and the newest news all around that 00:12:21.420 |
we're trying to create bots that interact with our world. And most recently, we have seen these robot 00:12:26.540 |
dogs that are able to deliver you a water cup or something like that. But these systems do not compete 00:12:34.460 |
with humans ourselves. Humans are like built to interact with our world. And that brings me 00:12:41.340 |
to here, which is that today we have humans hiring AI agents to do their work for them. I'm sure all of 00:12:50.860 |
you guys have probably employed some sort of AI agent to do your work for you. But today with SideQuest, 00:12:57.740 |
we are flipping the script, we are flipping the script. And we have AI agents hiring humans to do their work for 00:13:04.620 |
them. So AI agents obviously are amazing at interacting with the digital world. And humans are amazing at 00:13:13.500 |
interacting with the physical world. So why can't these AI agents just hire the humans? 00:13:19.180 |
So that brings me to the architecture of SideQuest, which is basically, like, let me give you a 00:13:26.620 |
hypothetical example. Let's say an AI agent is trying to host a hackathon. So now they have all the logistics 00:13:34.060 |
covered, but they need to put some advertising material up. They need some flyers up so that 00:13:39.660 |
people can find out where this hackathon is, where to go. But they don't have any physical means to do that. 00:13:44.620 |
So what they do is that they ping the nearest human to that area and tell them, "Oh, pick up this flyer, 00:13:51.260 |
put it in this location, and livestream that video to me. And as soon as I can see that you did it, 00:13:57.580 |
then I'll give you money." So that's exactly what's happening in SideQuest, and I'll show you a short demo. 00:14:09.580 |
Hello, world. My AI friends and I are hosting a hackathon. Let's check if the flyers are up. 00:14:17.500 |
But we do not see a flyer here. No flyer detected. 00:14:21.500 |
Bruh, I need a human to put on some flyers in room two. Let's do this. 00:14:35.740 |
It looks like there's a quest. So I have to collect three posters from table eight. Let's do it. 00:14:41.900 |
So over here, there's a live video stream that Claude is actively looking at and verifying whether 00:14:48.380 |
you're doing a new task. I found table eight. Let's see the posters. 00:14:57.980 |
Boom. Scanned. It says I have to set them up in Strong Yes. 00:15:01.100 |
We're here at Strong Yes. Now, let's set up the poster. And perfect. I think that should be good. 00:15:13.660 |
Let's scan it. Booyah. I made a hundred bucks. Let's go. And boom, we're done. We're ready for the hackathon. 00:15:24.700 |
Yep. And that's SideQuest. So let me talk a little bit about what I learned with building with Claude. 00:15:31.260 |
Is that first, Claude is really smart, as with like any of these AI systems these days. And they can 00:15:37.580 |
reason through many messy edge cases. So we as humans, we don't need to prompt every little nitty-gritty 00:15:44.140 |
thing. We can start thinking about bigger picture parts of building products. Secondly, we should design 00:15:52.780 |
like with a back and forth workflow with these AI systems. Like originally, we are thinking like 00:15:58.620 |
upfront, oh, how should I build this whole big thing? But that's a really big task. You can break it down. 00:16:04.620 |
Ask Claude, oh, like what are the different things that I need to do to to work on something? And let's 00:16:09.820 |
build this step by step. So with this iterative process, you can build like very robust systems. 00:16:17.580 |
So bottom line is that you should trust AI and trust Claude that they aren't things that you 00:16:24.140 |
have to micromanage. They can think on their own as well. And now some takeaways for builders to be like 00:16:30.860 |
like this cool guy, not this grumpy guy, is that you should think of AI as a system rather than just a 00:16:39.420 |
feature builder. That this is someone that you can like talk to, reason with. And secondly, as like thinking 00:16:46.540 |
bigger picture about us as humans, is that we should be system designers first, or architects of like the 00:16:55.660 |
things that we're building. Because in the future, we aren't going to be the ones writing the small code. 00:17:00.860 |
We'll be the ones dictating what code to write. So that brings me to the end. Thank you guys so much. 00:17:09.500 |
All right. Good afternoon, everyone. I'm Daniel. I study computer science at USC, and I've also built 00:17:19.100 |
projects across Amazon, IBM, and various startups. Yeah, very honored to be here as a student speaker 00:17:25.260 |
today. For more context, I help USC lead some of the entrepreneurship programs. And over the past year, 00:17:31.820 |
Claude has been integral to many of our projects, powering innovative solutions across various domains. 00:17:36.780 |
When Anthropic announced the hackathon at USC, a lot of the students, including my teammates Vishnu, 00:17:42.540 |
Shabayan, and myself, were naturally very eager to join in and explore new directions with Claude. 00:17:48.220 |
Today, I'm honored to share our journey and insights with you. So let's first start by looking at the 00:17:53.180 |
problem. Current LLMs are great at giving answers, but when decisions really matter, one general response 00:17:59.580 |
just isn't enough most of the time. Whether it's business, healthcare, or policy, high-stakes decisions 00:18:06.380 |
require diverse input and deep analysis. Today, getting those perspectives means prompting an LLM multiple times, 00:18:14.460 |
which could be slow, inconsistent, and very manual. Knowing that Claude excels at complex reasoning 00:18:20.940 |
as one of its most impressive capabilities, that's the gap that we aim to solve for our hackathon. 00:18:25.740 |
Introducing Claude Cortex, a system designed to emulate a panel of experts, 00:18:31.420 |
each analyzing the problem from a different angle. It dynamically creates specialized agents tailored to 00:18:37.100 |
your problem context and enables parallel processing for diverse insights. The output here is a more 00:18:43.580 |
synthesized and well-rounded recommendation, enhancing output quality for decision making. 00:18:48.060 |
It's basically like having your own strategy team for each prompt. 00:18:53.740 |
So yeah, let me show you how it works with a really simple example to test out the agents. So let's say I 00:18:59.100 |
want to learn how to use Lengraph specifically by researching its documentation. I also want to share 00:19:04.060 |
that finding with my teammates. I would type that in as a single prompt and let the master agent interpret 00:19:09.180 |
that request and spin it different agents, which in this case will need a browser agent to search and 00:19:14.060 |
extract relevant information from Lengraph's documentation, a research agent to summarize the key 00:19:19.500 |
concepts in plain language, as well as a notes agent to generate clear explanations, which it then shares 00:19:26.860 |
with my teammates automatically. Each agent will work independently, but they can communicate with one 00:19:31.980 |
another, creating a multi-agent system that gives more comprehensive insights. 00:19:35.820 |
Now for sectors where data security and compliance are paramount, Claude Cortex offers a secured mode by 00:19:43.740 |
integrating with AWS Bedrock. It ensures that all operations meet privacy standards, making it ideal 00:19:49.580 |
for sensitive environments. The rest of our architecture is also very straightforward. The front 00:19:54.940 |
end was built with Next and Tailwind. The back end leverages Fast API and Lengraph for orchestrating 00:20:00.060 |
multi-agent workflows. And Claude, of course, powers our agents reasoning with the addition of browser use, 00:20:05.740 |
which allows agents to fetch real-time web data and enhance their analytical capabilities. Claude Cortex represents a 00:20:13.420 |
shift in the way we use language models, moving away from simply generating responses to 00:20:18.700 |
structuring parallel reasoning pathways and delivering more comprehensive insights. 00:20:22.700 |
It's versatile, making it valuable across various sectors, from corporate strategy to public health safety. 00:20:28.620 |
Now, the key takeaways from building Claude Cortex are very intuitive, but the main two points here that I 00:20:35.980 |
want to emphasize are that when agent outputs were more focused and well-structured, like JSON format, Claude's synthesis 00:20:43.100 |
became more nuanced and high quality. It struggled, however, when upstream agents were more vague and 00:20:50.220 |
just dumped text blobs into the stream. And then dynamic task creation allows for flexibility. What that 00:20:57.740 |
means is we first started off by creating five predefined agents for every scenario. However, we later 00:21:04.780 |
realized that having a master agent to decide what tasks and agents to create allowed for more accurate and relevant information. 00:21:12.780 |
What we're building with Claude Cortex sits at a broader trend. Claude is powering a large number of student-led 00:21:18.060 |
products at USC. We've seen tools for lawyers to process case files faster, apps that help people retain 00:21:24.220 |
and connect knowledge more effectively, and software that can automate documentation and progress updates. 00:21:29.180 |
Claude's ability to read deeply, summarize clearly, and follow structure is what makes all of this possible. 00:21:36.620 |
Looking ahead, as a student building with Claude, the most powerful applications I've seen aren't just 00:21:41.900 |
asking Claude for answers. They're using it as infrastructure, something that you can wire into 00:21:46.380 |
your workflows and something that you can orchestrate like a system. And that's the shift that we see as 00:21:51.020 |
well. We imagine agents that can collaborate with one another, tools that can reflect, and context that can 00:21:57.740 |
compound. In summary, Claude Cortex isn't another AI tool. It's a leap towards a more intelligent, 00:22:03.900 |
secure, and multi-dimensional decision-making process. As we continue to refine and expand its capabilities, 00:22:10.380 |
we invite you to explore its potential and join us in shaping the future of AI-driven solutions. 00:22:16.300 |
Here's the team behind Claude Cortex. We're all student builders and we're all, yeah, builders and student 00:22:21.340 |
leaders at USC, and we would love to discuss more, so please feel free to reach out to us whenever. 00:22:25.420 |
I'm Daniel Gao, and it's been a pleasure sharing our work with you. Thank you for your time and attention today.