Jason Goodison, General Compute CTO on the Future of AI Inference

**Grace Gong** (0:03)
Hi, Jason. Welcome to Venture with Grace.

**Jason Goodison** (0:06)
Hey, thanks for having me.

**Grace Gong** (0:08)
I want to give a quick shout out to Shreen, who connected us over her dinner, and yeah, and then we're here.

**Jason Goodison** (0:15)
Shout out to her. Yeah, Shreen's fantastic.

**Grace Gong** (0:17)
Okay. I want to start from your entrepreneur journey right before General Compute.
You were a YC founder and you started two companies before for the General Compute. What were some core lessons that you've learned early on in your career shaped you into who you are today?

**Jason Goodison** (0:35)
Wow. Great question. Very loaded.
I did a lot of different stuff and I learned a lot of lessons. With my last company, SuperCall, what was interesting is that I as a solo founder, scaled up to 50K MRR at the height of the business, and I couldn't really find a path to making it really big company like a billion dollar company.
I think one thing I learned from that was that it's so important who you pick as a co-founder, it's so important who you pick as your team, and that I was so antsy to get something built and to get customers and to build a product and make a name for myself, that I skipped over the most important step, which was go out there and find the right problem and the right team. So I was building voice agents and I had a bunch of customers in Mexico that were using Spanish voice agents. Mind you, I don't speak Spanish, but I had all these customers and there were thousands of calls happening in Spanish every day, and I was listening to it and reading transcripts, and I think I got it wrong on both accounts. It's like I was solving a problem for a market that I didn't understand because I couldn't speak Spanish. Then second off, I just didn't have the right team, it was just me. So that was one of the biggest things I learned. With YC, I learned how important it is to have the right network too. And so I don't come from the traditional tech background. My dad worked in a factory, my mom cleaned houses.
I didn't really know that this world existed at all. And YC, when I got into it and moved to SF, it was the most eye-opening experience of my life. And it was really good to build a network very quickly in San Francisco, and I would recommend it to anyone. It's a great way to surround yourself with really smart, ambitious people. And so that was really, really important, I think.

**Grace Gong** (2:26)
For sure. I totally agree on, first of all, you have to find the right problem and the right team to solve the problem. And then you obviously, having a network is extremely important.
I want to start with, I feel like General Compute is super interesting in the inference space. Why don't we start with, why do you feel like inference is becoming the new, I want to say battleground, but I guess the new important area for AI infrastructure, especially for the agent workflow?

**Jason Goodison** (2:56)
Yeah. Great question. I think it was always inevitable. Let's say you train a model. Have you ever heard the phrase build once, sell twice?
That's the backbone of how to make money in a scalable way. It was always going to be the case that we were going to stop building once and eventually, we were going to have to sell twice. I think the last few years of the AI craze have been, okay, how much bigger can we make these models? How much smarter can we make them? At some point, someone had to be like, okay, let's start making money off of them. There wasn't really the productivity boost and the diffusion of AI into the economy that everybody wanted until recently. Anthropic, their revenue has exploded since the beginning of the year. They finally have that product market fit moment with AI, where you can use it in your daily life every single moment really for any task you're working on. I think that was the moment where everyone started being like, okay, AI is obviously going to have huge tremendous productivity gains and inference is the only way to do it. You can't train a model to be more productive if you can't run it.
There's just a huge wave of investment into this part of the economy.
I think that obviously I'm biased. I think we have one of the most unique angles of anyone working in this industry right now. But I'm excited to see how it plays out.