Ben Gawiser

Well, right now I'm thinking about what I'm doing next. Just finished up my involvement with my last startup.

So, you know, computer science degree, software development, then went and got an MBA, did some consulting, then did a lot of startups, then got tired of startups, ran a couple of software engineering and data science teams at Amazon, and then got dragged into one more startup. Now I'm thinking about what I'm doing after this.

That is a fantastic question. I have always loved technology. If you look deep into the archives of my hometown newspaper, you can find that they interviewed my third grade class and asked everyone what they wanted to be when they grew up, and I said computer programmer. I've always loved technology, and over the last couple of decades I've loved the intersection where technology meets business — those people who can actually sit in both places, comfortable whiteboarding and comfortable in the boardroom. And then really over the last decade I've focused a lot on ML and AI and analytics.

Absolutely. What you're seeing is an increase in productivity from individual developers with AI tools — your Claude AI, your ChatGPT, all those other things are really improving the individual productivity. So what that means is that each individual developer can actually get more done, and that they do, as you noted, need to be closer to the business. It's not just that they're sitting there, okay, write a function call that does this, because now you can basically do that with a prompt to an LLM. So they've got to understand the larger context about what they're doing. If you look at some of the orgs that have traditionally been better, their engineers always understood the business concepts. I think now it's sort of flowing downhill a little bit.

That is a fantastic question. When we think about sort of the coolest things that are coming down, everyone has gotten used to going to chat and saying, hey, how do I do this? And it kind of gives you a, you know, go to this site, click here, do that, do that. And it's pretty good about most things. What's really coming next is the ability to basically say, do that for me. And some of that's already working — you've got the operator functionality, you've got some other things. But as more and more companies create standalone agents that people can connect to and interact with, there's going to be a lot more functionality available to end users, and the ability to automate a lot of things. I was talking to someone just yesterday who was saying they had a claim for their glasses that they wanted to send to their vision insurance company, and they managed to basically completely automate it using ChatGPT and didn't have to fill out the forms themselves. Taking a task that's tedious like that — used to take you 15, 20 minutes and was really annoying because you were writing down the same thing three times — now can be done in a few seconds.

I just saw an article this morning which was talking about some people who are suing Google because they say that Gemini encouraged their family member to, quote, do a mass casualty event. And the only thing that saved a lot of people from dying was the fact that he didn't see the trigger that he wanted and ended up committing suicide himself. So it's really around data privacy, misinformation, and guardrails on LLMs and other agents.

I mean, look at what's happening with the Anthropic thing with the Department of Defense. That's a huge, huge deal. We need to have a conversation, not just as a country, but really as a species, as to what are the appropriate guardrails for this technology.

Yeah, fantastic.

Well, to a certain extent, I don't think it's changed at all. You go back 30 years, and everyone was moving from functional programming, C, to C++ or Java. They were making that change to object-oriented programming. Did it change how they hired people? Not really. They just looked for people who had object-oriented programming skills. A decade ago, you started looking for people who had done stuff in cloud native. So what it is, it's an adjustment. Someone who says, you know, I have never used a code assistant is going to have trouble finding a job. I think the industry is figuring out what is the productivity boost that each individual developer gets — how many people do I really need now? So at the end of the day, I don't think it fundamentally changes hiring that much, because the rule of hire a good team and get out of their way is still in place.

So I think it's a sculptor going from clay to using a 3D printer. It's a hell of a jump. Fundamentally, they're still making things. It's just that instead of taking hours and hours with the wheel to try to get it exactly right, they're typing in a bunch of specifications, a prompt, hitting enter, letting the machine do its thing, and coming back in a few hours and it's done.

Sure. It's going to be a change. All the people who were database administrators, where the company hosted its own Oracle database and there was the one person who was there who managed all the company's Oracle DBs — now, all of a sudden they're in the cloud, and you don't need a DBA like that anymore. Now you've got a site reliability engineer or a DevOps person. People's roles are going to change, and some people are going to get disrupted and moved around, and some people aren't going to make the changes that they need to make and they're going to find themselves on the losing end of it. But I think that's been true with any disruptive technology that we've seen.

So think for a second about late 70s, early 80s. What did a corporate accounting department look like? You had a lot of people, the visors on, sitting there with their green books and things like that. Maybe you got into the late 70s, they had a green screen terminal on their machine connecting to a mainframe. So maybe it's the 60s versus the 80s. But look at how corporate accounting has changed. You needed thousands of accountants in the 60s and 70s. By the time it got to the late 80s and everyone had Lotus 1-2-3 on their IBM PC XT on their desktop, there was a huge productivity change. So what you're seeing is, I think, equivalent to that.

Well, they said we'll never need more than 640 kilobytes of memory was sort of famously said by an IBM executive. And we've obviously shown that not to be true. We didn't get the flying cars we were all promised. Very few of us have gone to the moon. All those things that we all thought we were going to get 40 years ago — a lot of them we haven't. But I do remember someone saying in the 80s, eventually you're going to be able to watch any show you want at any time you want. And I thought that was completely ridiculous. But here we are, and we have Netflix and Hulu and all the other things now.

We have got to do something about misinformation. I think about the early days of the Internet. What was the promise of the early days of the Internet? Information would be free. Everyone would have access to information, and people wouldn't be able to lie anymore because they could be easily disproven by the Internet. That was, those of us who were online in the late 90s, that was all of our hope. And unfortunately, that dream, I think, has been completely crushed when you look at some of the troll farms and other things that have gone on to really drastically impact public opinion, not just in the United States, but all over the world.

AI is going to make it so much worse. 15 years ago, if you wanted to spread information, you had to get on Facebook yourself, write it yourself. And you were limited. That's why they hired tons of people in some of these places to do that. Now a single person could write an agent in a matter of minutes that could make hundreds of thousands, if not millions, of posts on social media.

No, we haven't solved for trust. And you know, you go back and read Satoshi's original paper on Bitcoin. It's absolutely brilliant. And it should have been someone's PhD dissertation. And that should have been the end of it.

Well, when it comes to things like that, Europe tends to lead the way. Look at the regulations they have in place for privacy, like GDPR and the digital market acts over there. California has done some stuff in the US with CCPA and some of the other things. So I think what we will see is Europe lead the way, then some of the more liberal states in the United States, then we'll start to see it out of Japan and other places. And then hopefully we'll manage to figure something out. But there's a lot of hope in that sentence.