Scott J. Gardner
I’m building instruments for studying how behavior changes through interaction: between people and AI, between AI systems, and across networks of participants. The question is not only what each participant does. It is how their interactions change what happens next. The long-term goal is larger: behavioral transparency for interconnected systems that remain opaque partly because we continue to observe them as if they were separate.
From Transformation to Instrumentation
I came to AI as an executive with a conviction that it would transform how businesses operate. Beginning in January 2023, I moved aggressively from experimentation into practical use: integrating AI into regulatory documentation, safety data sheet generation and management, operating procedures, retrieval systems, software development, and agent-assisted workflows.
I also began using it extensively myself. Across more than 200,000 logged turns, practical deployment gradually became sustained behavioral observation. I pushed systems across models and contexts, explored the edges of ordinary use, and paid attention to what happened over time rather than only whether an individual answer was useful.
My attention increasingly shifted from outputs to trajectories. What persisted across interactions? What disappeared? What was transmitted between participants? Did influence accumulate? Could repeated interaction narrow or expand subsequent possibilities? When did participants synchronize or diverge? Could apparently independent interactions become coupled through shared context, artifacts, or participants?
Eventually I noticed another participant changing: me. AI wasn’t merely producing outputs I consumed. Sustained interaction was changing what I noticed, how I framed problems, which questions seemed reachable, and what I did next.
That matters because the observer isn’t necessarily outside the system being observed. Once people and AI repeatedly interact, each becomes part of the other’s environment. Outputs become inputs. Responses alter subsequent conditions. Feedback accumulates. The interesting object is no longer either participant in isolation. It is the coupled system.
I wanted to examine that system without relying on either the AI’s explanation or my own impressions. So in 2025, I started building instruments. That became RightMinds.
The Measurement Problem
Most AI evaluation starts with an individual system and asks what it did. RightMinds starts one level outward: what happens between the participants? What is transmitted? What decays? What becomes amplified through repeated exchange? How does influence propagate? When do participants synchronize or diverge? How does one interaction change the state from which the next interaction begins?
These are old questions in one sense. Cybernetics has spent decades examining feedback, regulation, communication, adaptation, and systems whose components continuously alter one another. AI makes those questions newly urgent. The participants can now generate language, pursue objectives, create artifacts, recruit other systems into workflows, alter human decisions, and operate at speeds and scales that make the resulting feedback structures increasingly difficult to reconstruct.
An individual response can look completely unremarkable while participating in a larger pattern that isn’t. That is the object we’re trying to measure.
I don’t have the answers. I’m building the instrumentation that makes the bullshit answers harder to sustain.
Measurement > Interpretation
RightMinds is a measurement platform company. We build infrastructure for measuring behavioral dynamics across interacting participants. We also conduct research, because new measurement systems have to be tested, challenged, calibrated, and sometimes discarded. But research is not the product. Measurement is.
Our work focuses on trajectories, transmission, decay, coupling, synchronization, loop amplification, influence, constraint, and structural change through time. The platform is designed around a deliberate separation between what can be measured and what someone believes those measurements mean. Measurements should be reproducible. Source evidence should remain traceable. Methods should be inspectable. Interpretations should be allowed to compete.
That distinction is foundational to RightMinds. A sufficiently compelling interpretation can make almost any complex system appear understandable in hindsight. The harder and more useful problem is building measurements capable of constraining which interpretations remain plausible. So our research doesn’t exist to prove a worldview. It exists to pressure-test the instruments. Measurement > Interpretation.
A Builder’s Path
I didn’t arrive here through a conventional academic or AI-industry path. I’ve spent more than two decades moving between disciplines that are usually treated as separate careers. My work has crossed software, mechanical systems, chemical processes, optical engineering, manufacturing, regulatory compliance, product design and marketing, international trade, enterprise data integration, major-retailer relationships, and entrepreneurship. Running alongside all of it has been a lifelong obsession with psychology and behavioral economics, particularly where human behavior collides with technological systems.
That combination shaped how I build. Technology doesn’t operate in a vacuum. People adapt to it. Incentives reorganize around it. Interfaces change behavior. Organizations develop workarounds. Measurements become targets. Tools alter the people using them. And systems often produce outcomes that make little sense until you stop treating the technical and human components as separate problems.
Sometimes I was the executive responsible for the system. Sometimes I was the person designing or building it. Often I was both. I’ve written software and designed machinery. Built manufacturing processes and customer-facing products. Worked through chemical and regulatory constraints, supply chains, enterprise integrations, product positioning, and commercial relationships. Taken ideas from an initial technical problem through design, production, operations, and market.
Much of that work happened near the leading edge of what available technology could practically do. The common thread is not any particular discipline. It’s the ability to move between them, and to see the interactions among technology, people, incentives, and behavior as part of the system being designed. I ideate, design, and execute.
That matters because behavioral observability isn’t cleanly a software problem, an AI problem, a mathematical problem, a behavioral-science problem, or a business problem. It lives in the relationships among them. I’ve spent most of my career working in the between.
The Larger Conviction
AI is where this measurement problem became impossible for me to ignore. It is not where I believe it ends. Organizations develop behavior through interaction. Markets do. Information networks do. Human–machine systems do.
Research moves markets. Markets alter incentives. Incentives change organizations. Organizations influence media. Media changes people. People interact with models. Models generate information that re-enters the network somewhere else. We give these systems different names because the boundaries are useful to us. The feedback doesn’t care.
My conviction is that behavioral observability will become an infrastructure category as intelligence, influence, and decision-making become increasingly distributed. The consequential unit is increasingly not the human, the model, the agent, the news cycle, or the market. It’s the system they form together. And the behavior of that system cannot necessarily be recovered by understanding each component independently. That is the deeper problem RightMinds is built around.
Behavioral Transparency
The long-term objective is not simply better AI monitoring. It is behavioral transparency across otherwise opaque interconnected systems. Today’s analytical boundaries often follow institutions, technologies, industries, datasets, or academic disciplines. Those boundaries can be useful operationally while obscuring the feedback structures that cross them.
A market may be analyzed as a market. A news cycle as media. An AI interaction as a conversation. An organization as an organization. But information, influence, incentives, and behavior propagate between them. Those aren’t necessarily separate dynamical systems. They may be different observational apertures onto the same one.
RightMinds is being built to widen that aperture. Not to collapse complexity into one giant causal story. To make relationships, trajectories, feedback, and change across those boundaries more available for measurement. The ambition is to make opaque behavioral systems legible without pretending that legibility makes them simple.
Release Boundaries
Greater visibility creates its own responsibility. Methods that reveal how influence propagates may also help someone optimize that influence. Our operating principle is:
Observability should make trust cheaper without making deception cheaper.
That affects what we publish and what we don’t. The objective is not to make judgments for people. It is to give them better measurements before those judgments are made.
Why RightMinds
The previous generation of observability infrastructure became necessary because distributed computation made complex software difficult to understand by examining individual machines. We are entering an analogous transition in behavioral systems. The boundaries are increasingly administrative. The behavior isn’t.
If consequential behavior emerges at the level of interconnected systems, we need instrumentation capable of measuring that level. Not another dashboard for individual nodes. A measurement layer for what happens between them. That’s what I’m building.
Get in Touch
I’m interested in organizations encountering behavioral questions their existing instrumentation can’t answer, researchers who want to challenge the measurements, collaborators who see applications we haven’t considered, and investors who recognize the scale of the category we’re building toward.
Put a ruler on the ineffable.