Invention without surrendering judgment
Most conversations about artificial intelligence begin with what machines can do. This book begins elsewhere: what should people and AI systems invent together, and who remains responsible when the work matters? A language model can draft a proposal, suggest a software design, or analyze a case in seconds. Speed is useful, but it can also create a dangerous illusion that a plausible answer is a validated solution.
Human–AI co-invention asks for a disciplined partnership. People define values, negotiate the problem, listen to affected communities, judge acceptable risks, and remain accountable. AI systems may contribute alternative perspectives, structured synthesis, rapid prototypes, and critical challenges. Neither efficiency nor novelty alone is sufficient. An invention becomes meaningful when people can use it, challenge it, and trust the process that produced it.
This revised book links rigorous inquiry to practical creation. It combines a twelve-week course with a studio workflow: investigate a real problem, build a prototype, evaluate performance and collaboration, record the provenance of contributions, and design responsible deployment. The end product is both a working artifact and a defensible account of how it came into being.
Learners are encouraged to experiment boldly but report cautiously. When a test fails, preserve the evidence. When a model produces an attractive but unsupported claim, verify it. When a stakeholder rejects a design assumption, revise the assumption. These habits transform prompting from a convenience into a serious method of invention.
Author’s invitation
Use this text to build something real. Start with a person rather than a platform. Keep the human ability to ask “why,” “for whom,” and “what could go wrong” at the heart of every experiment. The strongest co-inventor is not the person who delegates everything to AI, but the person who learns to direct, question, and verify it.
From Automation to Co-Invention
1.1 Three Different Relationships
When people say they “use AI,” they may mean three very different things.
Automation hands a well-defined task to a machine and removes the human from it. A spam filter is automation. Success means the human no longer has to think about the task.
Augmentation keeps the human in charge while a system speeds up or improves parts of the work. A spell checker or code autocomplete is augmentation. The human's goals and judgment stay fixed; the tool makes execution cheaper.
Co-invention is something else. Here the human and the system each contribute to defining the problem and shaping the solution. Ideas move in both directions. The human may be surprised by what the system proposes, and the system's output is reshaped by human taste, context, and values. Neither party could have reached the result alone, and the human remains accountable for it.
The distinction is not about how advanced the technology is. The same model can be an automation engine in one workflow and a creative partner in another. What differs is the structure of the collaboration.
Beyond the textbook: making a decision
Automation is most defensible when the goal, exceptions, and success measure are stable. Augmentation is preferable when a person still needs to interpret a recommendation. Co-invention becomes useful when the problem itself is under negotiation. A project can move among these modes during one afternoon; labeling the system does not tell us how agency is actually distributed.
End of sample. Excerpts taken from the 100-page expanded edition. The full ebook is not available for unrestricted download.