These essays are not surveys of the field. They are attempts to work through specific problems in human-AI reasoning — to think in public about what the practice actually requires.
Why most AI collaboration fails before it begins
The failure is almost never the AI's. It is the human's — specifically, the failure to establish a bearing before entering the exchange. Without orientation, even the most capable AI system produces output that is technically correct and practically useless.
The dialectical method as a protocol for human-AI exchange
Hegel did not anticipate large language models. But the structure he described — thesis, antithesis, synthesis — turns out to be a remarkably precise description of how genuine human-AI collaboration proceeds when it proceeds well.
What your exchanges with AI reveal about how you think
Every exchange you have with an AI system is a record of your reasoning in motion. Not a log of inputs and outputs — a record of how your mind approached a problem at a specific moment. That record, analyzed, is a mirror.
On the difference between using AI and thinking with it
There is a version of human-AI collaboration that is simply outsourcing with extra steps. The human provides a prompt; the AI provides an answer; the human accepts it. This is not collaboration. It is a more sophisticated form of search — and it produces correspondingly shallow results.
The navigational metaphor and what it actually means for reasoning
A navigator without a bearing is not lost — they simply have no way of knowing whether they are going anywhere useful. The same is true of a thinker without orientation. The metaphor is not decorative. It is precise.
Clark, Chalmers, and the case for AI as cognitive component
In 1998, Andy Clark and David Chalmers asked whether cognition could extend beyond the skull. The answer they gave was yes — under specific conditions. Those conditions are now, for the first time, met by a widely available technology. The implications are significant and largely unexplored.