I’ve been forming a relationship with AI tools.
At times, they’ve brought me real joy—opening doors to things I couldn’t do before. At other times, they’ve been frustrating, producing results that feel shallow, messy, or incomplete. What I’ve come to realize is that both reactions are valid, and they stem from a deeper tension in what AI actually is.
My Experience
I’ve used AI for research, writing, and general language tasks. I’ve also seen it used by a professional engineer to turn a simple wireframe I created into an effective working prototype with very little effort.
My most meaningful experience, though, was collaborating with another engineer to build a fairly complex drawing application. AI allowed me to participate in creating software that would have been far beyond my abilities on my own.
At the same time, I’ve watched my partner, who works in communications, go from light use of AI to relying on it throughout most of her day.
These different contexts revealed something important.
Two Ways of Using AI
My partner often complains about the outputs she gets from AI. Even though she understands how the systems work, the results feel inconsistent and unsatisfying. The text isn’t quite right. There are too many possible variations. Nothing feels definitive.
That makes sense, because for her, the artifact is the end product.
In my case, I used AI to generate code that I didn’t fully understand. The architecture, the math, even the structure of the code, were often beyond me. But that didn’t matter as much (to me, at the time), because what I cared about was the outcome.
When I said, “when the drawing canvas is clicked, do this,” the result was usually clear and testable. Either it worked, or it didn’t.
This difference turned out to be central.
Artifacts vs Outcomes
I ended up having what I can only describe as a dual quality experience.
On one hand, the artifacts AI produced (code, text) were often not good. The code in particular was messy, hard to maintain, and grew quickly into something unwieldy. It solved problems in brute-force ways, with no apparent concern for elegance or long-term sustainability.
On the other hand, the outcomes were often excellent. The application worked. Features behaved as intended. Bugs could be identified and fixed quickly. In some cases, the results were not just good, but surprisingly good.
This contrast depends on what you consider the “product.”
- If the artifact is the product (like writing), AI can feel lacking.
- If the artifact is a means to an outcome (like software behavior), AI can feel powerful.
Working with the Models
To get good results, I had to do a lot of preparatory thinking. I needed to be very clear about what I wanted and how to express it. The process became a sequence of carefully constructed prompts.
One (recommended) technique that stood out was asking the model to “think” first, to produce a plan before generating code. This helped me verify that it understood the goal and was likely to deliver the outcomes I wanted, before asking the model to generate the code.
When the desired outcome was clear and concrete, the model was surprisingly reliable. It was also very effective at debugging, systematically (though here too, not elegantly) identifying and resolving issues.
But the underlying pattern remained: the system was optimizing for getting to the result, not for producing something well-crafted along the way.
Static and Dynamic Quality
Reflecting on this made me think about Robert Pirsig’s notion of Static & Dynamic Quality. Pirsig likens Dynamic Quality to the leading edge of a train and Static Quality to all the cars that follow behind:
… Dynamic Quality is the pre-intellectual cutting edge of reality, the source of all things, completely simple and always new … It’s only perceived good is freedom and it’s only perceived evil is static quality itself … any pattern of one-sided fixed value that tries to contain and kill the ongoing free force of life.
Static Quality emerges in the wake of Dynamic Quality. It is old and complex. It always contains a component of memory … Good is conformity to an established pattern of fixed values and value objects. Static morality is full of heroes and villains, loves and hatreds, carrots and sticks. It’s values don’t change by themselves. Unless they are altered by Dynamic Quality …
AI systems, as they exist today, seem extraordinarily good at Static Quality. They are built upon vast amounts of existing material. They recombine, extend, and apply what has already been done. In many cases, they can do this better and faster than humans. But they appear limited when it comes to Dynamic Quality.
They don’t initiate intent. They don’t seek out the unknown. Even when applied to new situations, they tend to map the past onto the present rather than consider the potential of uncovering something fundamentally new.
Where this leaves Me?
There has always been a kind of dance between Static and Dynamic Quality.
Dynamic Quality pushes into the unknown, it’s where new ideas, directions, and possibilities emerge. Static Quality follows behind, capturing those discoveries and transforming them into patterns, systems, and knowledge that can be reused and built upon.
Until now, that balance has largely been governed by human limitations. Our ability to explore Dynamic Quality is constrained by our ability to harness, apply, and (ultimately) move past Static Quality. We forget things, overlook patterns, and struggle to manage complexity at scale.
AI changes that.
For the first time, we have access to a system that can hold and operationalize vast amounts of Static Quality, far beyond what any individual can manage. On its own, that doesn’t make it creative. But in collaboration with humans, something more interesting begins to emerge.
Humans regularly encounter (whether we like it or not) Dynamic Quality. We regularly encounter new challenges and sense into new directions, deciding what is worth pursuing. But now, we can pair that with an ability to bring to bear on Dynamic Quality unprecedented depth of accumulated knowledge (Static Quality).
This creates the possibility of a new kind of partnership: one where human sensitivity and capacity for the unknown are amplified by machine-level access to everything that is already known.
And if that balance truly shifts, then the relationship between Static and Dynamic Quality may fundamentally change in, as yet, unimaginable ways (so much potential and so much risk).
Leave a Reply