Is Julius Just Another Language Model?  ## Metadata - Author: [[Hard Fork AI]] - Full Title: Is Julius Just Another Language Model? - Category: #podcasts - URL: https://share.snipd.com/episode/76b20785-adf4-4286-b82c-c6321c0cdc65 ## Document Notes ## Highlights - **Julius Startup Controversy** - Julius is an AI data analyst startup that raised $10 million pre-seed funding. - It faced early controversy for competing against ChatGPT's data analysis features. Transcript: Jaeden Schafer Today on the podcast, I want to talk about a startup that just raised a $10 million pre-seed, which is called Julius. It's basically an AI data analyst. And the thing that I think is really interesting here that I think is going to be like exciting for the whole industry is just how much controversy and I guess pushback this company had At the beginning when they were first getting started, because basically they're a data analyst. They do something that ChatGPT has built in. And anytime the ChatGPT or any of these foundational models, they roll out a new feature set, right? Like, you know, voice, or they have kind of the agents built into ChatGPT, people say, oh my gosh, the agents built into ChatGPT just killed 400 AI startups. But at the end of the day, it doesn't actually have to kill startups. And so today on the podcast, I want to be breaking down what happens in an industry when you still focus on a small vertical like this, what basically the outcome has been for Julius, and Where I think this is going to go in the future. ([Time 0:00:00](https://share.snipd.com/snip/385c89e5-cb8a-40be-95e9-85480c09ae12)) - **Focus Niche for AI Success** - Focus deeply on a specific niche or industry to build AI products. - This targeted approach beats generalized AI models like ChatGPT in specialized use cases. Transcript: Jaeden Schafer But I think that the ability to go deep on one specific topic, and one specific dimension, one specific industry or niche is so valuable that these big foundational models will never Be able to get every single feature set and tool that a person in any specific industry would need. They can generally make tools that are really useful for everyone. But when it comes to, you know, every little thing that a welder needs to do and every, you know, maybe software or use case that he could think of, you just want like a welder specific software Tool or a salesperson specific software tool instead of just relying on ChatGP for anything. So in any case, what are they doing? Their founder is Raul Sunwalker. He actually launched the company back in 2022. He went to Y Combinator for it. But what's interesting is it was originally a logistics startup and they pivoted away from that. Again, I'm a big fan of the pivot. If your company isn't working, don't scrap it. You built a lot just to find new ways to grow and come up with different concepts. It was basically designed to act like a data scientist. It analyzed and then it visualized data sets and it performed predictive modeling. It was basically running from ChatGPT and Claude and Gemini, which is interesting. I love the ability to swap between the three models because there's always a new, there's whatever the model that came up most recently is always the best. So if you always want the best, if you just use ChatGPT for everything, while when models come out that are better, you're going to be stuck with something that's not the best. Whereas if you have one of these platforms that just allows you to switch to the best model, honestly, think there's like huge value in startups that are doing that. The competition from the top models makes startups more viable, in my opinion. But basically, they carved out their own niche, and they said that they have about 2 million users, and they have more than 10 million visualizations, data visualizations that have Been generated. ([Time 0:03:54](https://share.snipd.com/snip/4939da16-c5a0-40a3-b3ed-543aa75c3cb1)) - **Julius Pivoted to Data Science** - Julius started as a logistics startup and pivoted to data analysis after joining Y Combinator. - It now analyzes, visualizes data, and does predictive modeling using multiple AI models. Transcript: Jaeden Schafer He actually launched the company back in 2022. He went to Y Combinator for it. But what's interesting is it was originally a logistics startup and they pivoted away from that. Again, I'm a big fan of the pivot. If your company isn't working, don't scrap it. You built a lot just to find new ways to grow and come up with different concepts. It was basically designed to act like a data scientist. It analyzed and then it visualized data sets and it performed predictive modeling. It was basically running from ChatGPT and Claude and Gemini, which is interesting. I love the ability to swap between the three models because there's always a new, there's whatever the model that came up most recently is always the best. So if you always want the best, if you just use ChatGPT for everything, while when models come out that are better, you're going to be stuck with something that's not the best. Whereas if you have one of these platforms that just allows you to switch to the best model, honestly, think there's like huge value in startups that are doing that. The competition from the top models makes startups more viable, in my opinion. But basically, they carved out their own niche, and they said that they have about 2 million users, and they have more than 10 million visualizations, data visualizations that have Been generated. So this is something that's very popular. A lot of people are using it. They said, quote, the easiest way to use Julius is to just talk to it. You can talk to AI like you would talk to an analyst on your team and the AI, like ([Time 0:04:33](https://share.snipd.com/snip/8c6845c6-e53a-4b31-bdb2-f0d32cde4084)) - **Harvard Collaboration Boost** - Harvard Business School adopted a customized Julius product for their Data Science and AI for Leaders class. - This academic partnership improved Julius's features and credibility. Transcript: Jaeden Schafer And I'll be able to come up with something for you to visualize that. What's interesting is last year a professor at Harvard Business School Ivan Bojivanov he was really impressed he actually saw it and he asked them to modify it to be specifically for Harvard's new required course, which is called Data Science and AI for Leaders. So it's kind of interesting. I think those kinds of projects help a product along immensely, right? If someone's like, you know, we have this specific use case. In this case, it's literally a class for data science and AI and leaders. So they have a very specific like data science class. And they're like, we need it to be able to do all these things. And so what's cool is when you partner with academic institutions or businesses, you could also do this ([Time 0:06:11](https://share.snipd.com/snip/9ad6ebd1-1c13-40bf-b6e5-22ab8d3332bd)) - **Use-case Focus Enables Success** - Startups labeled as "ChatGPT wrappers" can survive by focusing on unique industry use cases. - CEO Raul Sunwalker's focus on practical use cases helped Julius succeed despite skepticism. Transcript: Jaeden Schafer So I think this was a kind of a good collaboration they did with Harvard. And I think it helped out the product a lot. So one thing that Son Walker said about this, he said, people told us you're not going to succeed basically because ChatGP is building this. What we found was that being focused on a use case is really important. So of course, he kind of got a little notoriety from his viral prank back in the day, but he turned this into a real company, really focused on a use case. And so I think that overall, this is a really interesting company and a huge congratulations to the whole team on raising $10 million pre-seed round. ([Time 0:06:58](https://share.snipd.com/snip/7a6a31d5-e10f-4a7e-a09d-184d8093b75f))