Intelligence is collaborative.

Most people use AI on their own. Directed AI changes that. Talk to Shannon in Slack, connect related chats, and bring someone into an AI session with the context and tools they need to help.

A scenario

Imagine Bob is working with AI and gets stuck. Alice has solved a similar problem before. She joins his conversation, shares her earlier work, and corrects an assumption that was leading him in the wrong direction. They get to a better answer, and her contribution stays connected to her. As that work helps others, it builds a visible record of her expertise. We’re building toward a network where she can also get paid when her knowledge is used.

People supply goals, context, judgment, corrections, new knowledge, and expertise that improve what AI can do. Directed AI is built to bring those contributions into the work when they’re needed—and keep them connected to the people who made them.

Our mission

To direct the world’s intelligence to where it’s needed, and keep it owned by the people who create it.

Founding stories

Justin Palmer

My experience spans machine learning, fintech, and recruiting. Across those fields, I've built countless models, created datasets, designed evaluations, and put human expertise to work in products used by thousands of companies.

My career started working on language technology and evaluation at MITRE on novel joint speech transcription and machine translation models in the 2000s. In grad school, I explored how language could guide computer vision. Then I shifted to fintech and applied sequence models and large datasets to credit risk to help make data part of everyday operations at LendingHome. After that, I founded TopFunnel, where we focused on building people data, matching, and recruiting workflows that combined ML and automation with human judgment.

The ideas behind Directed have been developing this whole time. In 2017, I coauthored “Real-World, Man-Machine Algorithms”, describing a version of continuous learning that unified machine learning, task definition, data labeling, human contributions. Those ideas were at the heart of how we worked at TopFunnel.

In the post ChatGPT world, there is an opportunity to go bigger, based on these same fundamentals: define the problem carefully, measure what matters, and give capable people (and agents) the tools, context, and incentives to contribute their best work. We’ve spent years building the technical systems and human teams together. Directed brings that experience to a broader market for intelligence.

Marko Krkeljas

I’m an engineer and artist based in New York. At Directed AI, I focus on how information, decisions, and actions flow through networks of human and machine collaborators. Previously, at Barnard College, I led technical projects at the Computational Science Center and software development for the Academic Technology team, building educational tools and teaching programming and machine learning. Before that, I built data systems for renewable energy at Apex Clean Energy.

My art and engineering are a single practice, driven by the desire to bring something original into the world.

Tahmid Khan

I built Forms.md, an open-source tool for creating multi-step forms and surveys, and Halfmoon, a drop-in replacement for Bootstrap 5. My favorite part of open source is that anyone can pick up what someone else built, learn from it, and take it further. Nobody starts from scratch, and the people who helped get credit for it. Most work with AI doesn’t feel like that yet. It happens in private chats, and what you figure out disappears when you close the tab. Directed changes that: conversations can be shared and built on, the right people can jump in, and their contributions stay connected to them.

I like tools that connect people and AI in a meaningful way, rather than just running a billion agents. I'm also interested in novel UIs for AI and LLM applications.