The published record captures what happened. This chronology asks when the way of working, deciding and observing began to change.
The dates mark when a pattern became visible in the public record. They do not imply that a transition was instantaneous, complete or irreversible.
Ongoing human–AI work had not yet been organized as a public longitudinal experiment.
The interactions became material for observation, with a public record and a separate private analytical layer.
An instruction could carry too much implied authority. Clarification, approval and execution were easy to blur.
State, permission and action began to be treated separately. Failures to maintain those boundaries became observable.
Forming an idea and producing a working artifact involved a more distinct handoff.
Ideas could be decomposed, built and revised inside the same cycle of thought. Faster execution also exposed new handoff failures.
The effort needed to make something helped determine which ideas became real.
Attention moved toward selection, activation and the judgments that should remain with the human—even when production was easy.
Human attention remained part of the practical execution loop, even after work had been delegated.
Some work could proceed between decisions and return for inspection. The gap between intended autonomy and demonstrated autonomy remained part of the record.
Rapid creation made it tempting to treat additional output as evidence of useful progress.
Pacing, selection and time away from production became deliberate constraints. More output no longer served as a sufficient measure.
Changes were often framed as something happening to the human or as a function supplied by the AI.
The coupled system itself became a candidate object of study. Narrative observations and structured measurements were distinguished without treating retrospective classification as original evidence.
Delegated research could appear complete when the returned artifact looked useful.
A pre-meeting brief arrived without the research needed to support it. The human identified the gap and supplied context the AI had missed. Another research assignment succeeded that morning. The contrast exposed an unreliable handoff, not a simple lack of available information.
OBOR began as a small China-to-Canada business intelligence prototype.
A reconstruction of seven weeks of development showed its focus shifting toward research questions and evidence standards as new capabilities were added. The change in purpose is visible in the artifacts. What drove it—the human's evolving judgment, the available AI capability, or both—remains unresolved.
The loop had acquired increasingly specific diagnoses and rules for recurring mistakes.
A review of earlier Loop Records found failures still recurring after they had been named. The collaboration could explain its errors more clearly than it could demonstrate durable prevention. A stronger test is whether comparable work proceeds without those mistakes or repeated human correction.
FIELD NOTES 150
This is a retrospective interpretation of an evolving public record, not independent validation of causation or capability. Future revisions should preserve reversals, unresolved questions and counterexamples.