Agent Wars VI: Ping Pong — A Panel of Expertise with Two AIs
A second AI can send the ball back. Three rounds of cross-model critique made the economics argument narrower, more honest, and ready for a real-world test.
Read articleTODD KELSEY, PhD · INDEPENDENT RESEARCHER
I study durable memory, human agency, education, and the economic transition created by increasingly capable AI systems.
Explore the work ↓CURRENT THESIS
“GPU context is a scratchpad,
not a filing cabinet.”
Our knowledge should outlive any model, vendor, or interface. The next generation of AI systems needs memory that is durable, portable, inspectable, and owned by the people who create it.
WRITING ARCHIVE
38 pieces on AI, education, work, and human agency, available here for durable access.
A second AI can send the ball back. Three rounds of cross-model critique made the economics argument narrower, more honest, and ready for a real-world test.
Read articleWhat does an accepted agent outcome earn after model use, review, rejected cases, overhead, and acquisition? A deliberately illustrative invoice and a testable pilot.
Read articleSPF, DKIM, and DMARC help important email reach its destination—and help recipients trust who sent it. Here is what each layer does, why setup is tricky, and how ChatGPT Work helped turn a one-time repair into a recurring health check.
Read articleSELECTED RESEARCH
Ongoing programs, working hypotheses,
and prototypes.
RESEARCH PROGRAM
A model-independent memory substrate for people and agents: durable sources, hybrid retrieval, clear provenance, and context assembled for the task at hand.
Store memory once. Retrieve it intelligently.
Active research
ECONOMIC RESEARCH
Tracking the gap between rapid AI capability gains and the slower institutional signals—employment, tax receipts, organizational design—that reveal structural change.
The shock can arrive before the statistics admit it.
Working series
EDUCATION DESIGN
Curricula and learning systems that move beyond tool fluency toward judgment, orchestration, verification, and durable intellectual independence.
Teach people to direct systems, not merely prompt them.
Field notes
PROTOTYPE
An evidentiary layer for decisions: sources, assumptions, tool and model versions, uncertainty, revisions, and the path from evidence to action.
Reasoning should leave a trail.
In development
ABOUT
Todd Kelsey is a researcher and educator working at the intersection of artificial intelligence, technical communication, institutional change, and human-centered systems.
After more than a decade in higher education—including building a four-course Certificate in Applied AI—his current work focuses on the infrastructure and public ideas needed for people to retain continuity, authorship, and judgment in an AI-mediated world.
Connect on LinkedIn ↗