About
Why We Started 73ai
AI agents are getting really capable. They can write code, manage infrastructure, automate workflows. But capability without control is a problem. The more powerful these tools get, the more important it is to figure out how to keep them safe, predictable, and transparent.
Most AI safety work is theoretical. We think the hard questions get answered by building real things and seeing what breaks. How do you give an agent enough freedom to be useful without letting it do something dangerous? How do you build harnesses that keep AI controlled without making it painful to use? These are applied problems, and the only way to solve them is through experiments.
That's why 73ai exists. We run experiments, build tools, and publish everything we learn. No paywalls, no gatekeeping. If we figure something out, it goes public. We believe the people building with AI deserve honest, practical research about what actually works.
Founder
pjay is a software engineer and AI researcher with about 10 years in the industry. He co-founded a fintech company (Spotlight), spent a long time building distributed systems and payment infrastructure at scale, and has published research on matching algorithms and code retrieval techniques for coding agents.
Now he's focused on AI safety for general-purpose agents through 73ai. Most of his current work is about how to build agent harnesses that keep AI tools controlled without making them painful to use.
X · LinkedIn · Personal site