Truth seeking, a new paradigm
In 1994, the team behind General Magic imagined a world where software could leave your computer and act on your behalf. They envisioned agents that could find information, make reservations, and even initiate transactions on behalf of users:
Mary receives and pays her bills electronically. Her agent receives each bill as it arrives, verifies that Mary has authorized its payment, checks that it is in the expected range, and issues instructions to the bank. Mary's agent prepares for her a consolidated monthly report, and at tax time sends her accountant a report of her deductible expenses.
Thirty years later, agents are rapidly catching up to that vision and promise to change the future of how we work. The infrastructure they use, however, is still woefully antiquated, and nowhere is that more true than with payments. At Natural, we’re trying to fix this by creating the standard for how agents programmatically transact.
I recently joined Natural as a software engineer after graduating from the University of Chicago, where I studied mathematics and computer science. Coming into college, I had this vague dream of becoming a professor. I spent the first two years happily disappearing down different rabbit holes in mathematics and theoretical machine learning. I loved the feeling of spending days on a problem before finally feeling the “click.” Over time, though, I realized I wanted a tighter feedback loop and to directly impact the real world. By the time I reached my senior year, I knew that meant joining a startup. Ideally, it would be a killer team, working an important problem, and would offer the chance to do meaningful and interesting work. But finding a startup that fit the bill proved harder than expected.
Then, in January, I got a LinkedIn message from Kahlil.
Natural looked very different then. The team was working out of the Human Capital office on Pine street, and they had just raised their seed round. I was initially a little skeptical. I had a latent interest in financial markets, but, aside from the occasional Patio11 article, I knew very little about payments, let alone what “agentic payments” were supposed to look like.
To my surprise, I found myself flying to San Francisco for an onsite a few conversations later. I went in with an open mind, and that experience did more than anything else to convince me to join. Rather than giving me a neatly scoped interview problem, the team asked me to think through a north-star design for Natural’s risk system. I spent much of that week working with Walt. I’d spend the day thinking about some aspect of our approach to risk, he’d poke holes in my overly-abstract ideas, and I’d go back to the drawing board. By the end, we had converged on something much more aligned and grounded. I had gotten my first taste of startup life, and I loved the feeling.
Beyond the fun problems, I was most impressed by the team. Despite being a small group with an unreasonable amount of work, everyone was very generous with their time. I could turn to someone with a half-baked question about payments and get an hour-long tutorial in response. There was very little ego about whose idea was right, and people were remarkably candid about what they knew, what they didn’t, and where there was still some room for growth. The focus on truth-seeking was something I admired deeply.
Fast-forward to today, and I couldn’t be more excited about our direction. Fittingly enough, I’m now working on our risk and fraud systems.
Agent payments introduce a strange inversion to many assumptions of fraud detection. Historically, automation itself was a useful signal of abuse. A user making ten payments in a few seconds, moving through a checkout flow with machine-like regularity, or behaving identically across many sessions might reasonably attract scrutiny. But obviously, these might now be typical of the average Natural user. At the same time, fraudsters have increasing capabilities, with agents allowing attacks to become more sophisticated, better coordinated, and faster than ever before. We thus have the opportunity to define a new paradigm for thinking about risk, one better suited for a world where agents will drive the majority of payment volume.
If building the next generation of payments excites you, we’re hiring /careers!
