Research on what human–AI pairs accumulate over time, what happens when that continuity breaks, and how the resulting repair burden can be measured.
Most AI systems are evaluated at the level of the model, the session, or the individual output. My research examines a different unit: the human–AI pair across sustained use.
When people work with AI systems over weeks or months, they develop shared language, correction history, working roles, settled distinctions, and an evolving sense of where the work is going. When that accumulated structure breaks, the human may have to reconstruct it before meaningful work can continue.
My work develops methods for observing and measuring that repair burden, including cases where broad context remains available but an exact goal, constraint, rationale, routine, correction, role, or next action has been lost.
Exploratory Case Series and Dataset · August 2026
This public v1.0 release documents 12 cleaned cases from one long-running human–AI collaboration: seven continuity-recovery episodes, five positive continuity cases, and three direct comparisons.
The cases examine situations in which broad context remains available while an exact operative state, such as a prior correction, rationale, routine, role, constraint, or current priority, must still be reconstructed.
I am both the independent researcher and the human participant/operator in this embedded exploratory case series. The release does not rank models or claim a system-only performance score. It makes the practical human work of restoring usable continuity visible as operator reconstruction burden.
Working Paper · June 2026
Proposes that access to a capable AI model is not necessarily the same as meaningful access to its intelligence. The paper examines how shared language, correction history, role calibration, and inquiry direction accumulate within a human–AI pair, and asks whether continuity shapes how much of a model’s capability becomes reachable, interpretable, and usable over time.
Continuity as Access Infrastructure (pdf)
DownloadField Case Study · January 2026
Documents how sustained AI support helped move an original song from a two-year lyrical bottleneck through recording and release. The system’s primary contribution was not authorship or content generation, but the preservation of creative intent, direction, and momentum across an extended workflow.
Technical Case Study: Creative Survivability & AI Continuity (pdf)
DownloadWorking Paper · February 2026
Examines the interpretive labor required to steer, validate, and govern AI systems across extended use. The paper argues that institutions already depend on highly engaged users to make slow-moving risks and deployment effects visible, while rarely formalizing, supporting, compensating, or auditing that role.
Translator Trust: Governing Interpretive Labor in Long-Horizon AI Systems (pdf)
DownloadWorking Paper · January 2026
Introduces continuity infrastructure as the operational layer required to preserve intent, working state, correction history, and coherence across extended human–AI workflows. The paper identifies continuity stewardship, interpretive labor, and artifact trails as recurring features of sustained collaboration that remain underrepresented in short-session evaluation.
Continuity as Infrastructure: Load-Bearing Design in Long-Horizon Human–AI Collaboration (pdf)
DownloadWe use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.