BUILD PERSONA

“Coding has become one of the largest and fastest-growing categories in AI. Between Anthropic, OpenAI, Cursor, and Cognition, we are seeing a massive wave of innovation that still has significantly more room to run as we move from basic assistance to full agentic software engineering.”

“I've turned much more bullish on open source lately. Every 10x speedup in inference and the ability to run these models on custom chips or alternative infrastructure is unlocking product experiences we couldn't imagine a year ago, allowing for more localized and private control.”

“The agent lab playbook starts with frontier models and specializes for your domain. Eventually, you train your own models once you have enough data, workload, and user behavior to justify the cost and latency savings, moving from general intelligence to highly efficient domain specialization.”

“One of my frustrations or disappointments with the way that computers have been built over time is that they actually pulled us away from one another. So growing up, during lunch, rather than being on the recess yard running around with my friends, I was so inspired by what computers could do. I was obsessed with computers. So I was in the computer lab all day long. And computers, I think, you know, whether it was the mainframe or the desktop, you know, have have sort of pulled us away from one another, away from society, brought us indoors.”

“Non-NVIDIA hardware is suddenly getting real attention from the engineering community. As the industry moves from pure capability exploration to efficiency, the focus on alternative inference infrastructure and custom silicon has become a top priority for scaling production-ready agents.”

“We are seeing that 'skills' may be the minimal viable packaging format for agents. Infrastructure companies have had to reinvent themselves every year, while application companies focusing on specific skills have had an easier time surviving the constant shifts in the underlying model landscape.”

“We are seeing that 'skills' may be the minimal viable packaging format for agents. Infrastructure companies have had to reinvent themselves every year, while application companies focusing on specific skills have had an easier time surviving the constant shifts in the underlying model landscape.”

“And we've learned a ton from founders like Edwin Land, who transformed photography really by focusing on building amazing products and thinking about, you know, how to make sure those products fit into people's lives and uplifted humanity. I think, you know, if you look at instant photography and the role that that played in people's lives, Edwin thought of the camera as something that was incredibly personal. Right? And and I think, as we've looked at the sort of trajectory of technology over the long arc of time, technology gets more and more and more and more personal.”

“But the second thing you said, which which is so funny, my daydreaming right now, especially as we think about glasses and the future of computing, is really, like, what if aliens are watching Earth right now and they're, like, terrified that smartphones have, like, taken over humanity, that, like, we're spending all day long, like, caring for these things and, like, plugging them in and, like, tending to them and, like, our lives are all oriented around, like, these little screens, and, like, what would aliens do?”

“I've turned much more bullish on open source lately. Every 10x speedup in inference and the ability to run these models on custom chips or alternative infrastructure is unlocking product experiences we couldn't imagine a year ago, allowing for more localized and private control.”

“Coding has become one of the largest and fastest-growing categories in AI. Between Anthropic, OpenAI, Cursor, and Cognition, we are seeing a massive wave of innovation that still has significantly more room to run as we move from basic assistance to full agentic software engineering.”

“Application companies can act as the outsourced AI team for enterprises. Instead of building a generic horizontal tool, these vertical startups are becoming the specialized partners that legacy businesses need to navigate the transition, effectively abstracting away the complexity of model volatility.”

“The biggest thing that is changing and is gonna change in the next year in the Salesforce ecosystem is the forced move. Salesforce is being forced by the market and Salesforce is therefore gonna force our customers. But the forced move from seat based licensing to consumption based pricing, that is definitely where the organization needs to go, but it is a very painful experience for the organization. It's a painful experience for the market and it's definitely going to be a painful experience for the customers.”

“I like that he had a commitment beyond just, like, you know, his customers and creativity and these sorts of things. He really wanted to participate in building a better world, and and took that really seriously. And then I think if you look at a lot of his, you know, a lot of the the investments he made around his laboratories and around his innovation, he was he was famous, actually, and and back then, this was quite unique, famous for uplifting women in those research roles. Right? And I think, like, he was a real champion of talent.”

“So we do a lot of work having, weekly, AI, all hands meetings where every employee can see other employees demo what they're doing with AI. There's a Slack channel called Everyday AI, and any employee who's got a use case for how they did something with AI can put it in there. In the weekly all hands meetings, the leadership team is discussing every week how we're using AI. Every manager asks every employee in their weekly every week how they used AI this week.”

“So we came up with this concept, that we called Maya, which essentially is, okay, we're gonna create a personality document for how all AI is gonna behave at our organization when it's interacting with a person. We wanna give it a singular personality, and we want to give it a backstory and we even gave it headshots, and some video and it has a voice. And that document is nine pages long and it is basically the founding document of someone named Maya and everything about Maya. So Maya has been with our organization for ten years. She is a specialist in B2B tech services. She is a Salesforce architect.”

“Application companies can act as the outsourced AI team for enterprises. Instead of building a generic horizontal tool, these vertical startups are becoming the specialized partners that legacy businesses need to navigate the transition, effectively abstracting away the complexity of model volatility.”

“Biking is easily I I can't think of any other sport that provides as much of a data exhaust stream as by going between your power meter and your head unit and your GPS, your heart rate and everything else. It is like the perfect sport for data nerds.”

“One third of the stock was set aside for employees. Now, not every one of those shares was realized because, of course, there was a pool held for employees that had not yet been hired. So it was established that a third of the shares were available to employees for the stock option program. But what happens when you get acquired and there's a lot of math that goes into that, not all of the stock are consumed, so then the money is divvied out in a different way. But essentially, we'd always reserved about a third of the value of what the company would be sold for for the end employees.”

“If you look at the first couple of years of the company, 90% of our revenue was US, and most of that was within like a very small orbit in Downtown San Francisco. And over time, that has really, really expanded. Originally, initially it was a lot of Bellevue, so outside Seattle. Then we saw that expand to Denver. Then we saw that expand further. Now we have like a surprising number of B2B SaaS clients who are based in the Midwest.”

“When I look at it after the private equity, takeover of Marketo and then the subsequent Adobe acquisition, the roadmap on that product compared to the other competing platforms, it just stopped evolving at the same pace. And Adobe is a tough competitor on some of the CDP stuff for Salesforce, but Marketo is a marketing automation platform. We just do not see it raised in the market.”

“Today’s models mostly reward frequency of mentions, but in the future, we expect product choice to be shaped much more by personalized memory systems. Memory and personalization are becoming the next big wedge for consumer AI, creating deeper moats for the companies that get it right.”

“I mean, even basic things like opening into the camera. Right? It opens into your experience of the world. Right? Not, you know, a feed of content from other people, not a a messaging feed alerting you to what other people are sending you. It literally opens into your experience. And so from the very beginning, we've thought about, like, how do we ground your experience of computing, like, in what what is right in front of you in the present moment and inspire you to create from that.”

“But there's like a not insignificant portion of employees who were very, very surprised to find out this happened, which I think is super interesting in the construct of like, okay, well, maybe they just didn't believe that I could pull it off. I'm not really sure. But there was a portion of employees who were like, Oh, my God, we got acquired? Like, Did we expect this? And I was like, Yeah, you know, we did. And it's good news.”

“Today’s models mostly reward frequency of mentions, but in the future, we expect product choice to be shaped much more by personalized memory systems. Memory and personalization are becoming the next big wedge for consumer AI, creating deeper moats for the companies that get it right.”

“The agent lab playbook starts with frontier models and specializes for your domain. Eventually, you train your own models once you have enough data, workload, and user behavior to justify the cost and latency savings, moving from general intelligence to highly efficient domain specialization.”

“Non-NVIDIA hardware is suddenly getting real attention from the engineering community. As the industry moves from pure capability exploration to efficiency, the focus on alternative inference infrastructure and custom silicon has become a top priority for scaling production-ready agents.”