When I started posting about AI this year, I did not realize that I was beginning my own version of David Bowie’s Low album.

I use that comparison carefully. Low matters here not as a code book or a track-by-track template, but as an allusion to emergence, fracture, atmosphere, and a break in method that only becomes visible after the fact. It was not just another Bowie record. It marked a turn: a new working condition, a new tonal register, a new way of hearing what the medium could do.

Only later did I realize that my early 2026 AI posts were no longer behaving like commentary. They had become a sequence.

Reading Bowie in Berlin: A new career in a new town, by Thomas Jerome Seabrook, helped me see that. What first appeared as a run of separate essays now seems, in retrospect, to have been a cycle: not planned, but emerging with its own order, pressure, and tone. The posts were not simply accumulating as observations about a fast-moving technology. They were recording a shift in how I understood AI itself.

The cleanest rule I have found is this: a post belongs in my Low sequence only if it marks the turn from AI as output machine to AI as medium, something to investigate, govern, and work within under conditions of method, control, texture, and recurrence.

That rule gave the sequence its shape as a suite, not just a chronology.

I. The Break

The first movement breaks with the inherited frame.

The Inquest marked the decisive turn away from treating AI as something to admire and toward treating it as something to investigate. Human-in-the-Loop Is Systems Stewardship made clear that human presence alone was not enough; what mattered was stewardship of boundaries, invariants, and judgment. Moving Beyond Prompts to Protocol-Governed AI and Prompting or Negotiating? pushed the argument further. Prompting was proving too small a frame for serious work. The real issue was no longer wording, but control. The End of the Magic Wand closed the door on the fantasy that these systems could be treated as frictionless helpers so long as one got the prompt right.

That was the break. AI was no longer just a tool to query more cleverly. It had become a problem of operating conditions.

II. Inside the Instrument

The second movement is where the work moved inside the medium.

Building the Stochastic Sandpit for AI opened the way for thinking about AI as a space for bounded experimentation rather than a vending machine for polished outputs. Playing the Guardrails now seems to me like the track that revealed what the whole run was really about. Jimi Hendrix matters here because he knew distortion was not just damage. In the right hands, it became part of the instrument. Eno matters because he understood medium, system, and environment. He knew error could be held inside a structure and made productive. Edge belongs here too, for the same reason: sound is not just played; it is designed, staged, and governed. Bowie understood the larger lesson. Fracture, interruption, and atmosphere were not problems to be cleaned up. They were part of the composition.

That was the turn for me with AI. The flaws and unruly behaviors of these systems are not always just bugs on the way to perfection. In exploratory work, they can sometimes be studied, played, even used. But only under discipline. Distortion is only interesting when someone is still playing the instrument. Otherwise, it is just noise. That is where the control plane enters the picture. Without it, you are not working the medium. You are being worked by it.

The Long Session Trap deepened that realization by exposing the hidden cost structure of sustained AI work. The promise of leverage could quietly turn into administrative burden. The session itself could become the work.

III. The Control Plane

The third movement widens from craft to architecture.

Vibe Coding and the Control Plane put the matter in its clearest form: what is at stake is not convenience, but whether you have surrendered the control plane itself. The Real Legal AI Risk Is in the Handoffs shifted the focus from isolated outputs to workflow architecture and extended the thinking beyond law. The Protocol Layer pressed the case that rigor has to be designed into the work, not added later as a moral flourish.

This was the point at which my language changed. The inherited vocabulary of prompts, outputs, assistants, and better results had begun to fail. It was useful, up to a point. But it could not carry the weight of what I was actually seeing. AI was no longer presenting itself simply as a tool. It was showing itself as a medium with textures, pressures, distortions, and design contradictions of its own.

That changed the stakes. The question was no longer how to get better answers from the machine. The question was how to govern the conditions under which the work could remain trustworthy, usable, and alive.

IV. Standing Waves

The final movement is the ending suite.

The Threshold Moment marked the point where something in my hearing changed. AI as the Unreliable Witness and the Appearance of Completion sharpened the forensic problem: fluency can improve even while coherence degrades. Surface completion can become its own deception.

Standing Waves is the closing track because it stays inside the instrument. It does not try to widen outward into institutions, markets, or professions. It ends with a field note. In sustained AI work, a session can develop patterned persistence: recurrences, pressures, interferences, carry-forward effects. The real unit is no longer the isolated prompt and answer. It is the condition of the session itself.

That was the note I had been moving toward without fully naming it. I wanted to change my ear, not reach a conclusion, let alone mastery.

The Final Mix

The core album, as I hear it now, consists of twelve tracks:

The Inquest
Human-in-the-Loop Is Systems Stewardship
Moving Beyond Prompts to Protocol-Governed AI
Prompting or Negotiating?
The End of the Magic Wand
Building the Stochastic Sandpit for AI
Playing the Guardrails
The Long Session Trap
Vibe Coding and the Control Plane
The Threshold Moment
AI as the Unreliable Witness and the Appearance of Completion
Standing Waves

Companion Tracks

Not every AI post I wrote this year belongs in that final mix. Some of the other essays from the same period still belong to the larger body of work. They carry themes, pressures, and discoveries that helped define the suite. But not every strong track belongs on the final album. A few now feel more like companion essays, side paths, or adjacent experiments: part of the same season, part of the same investigation, but not part of the final mix I hear as my Low album.

That distinction matters. I did not set out to make a concept album. I recognized one after the fact.

What ties these posts together is not that they are all about AI. It is that they document a coherent shift in understanding. They record the point at which AI stopped being, for me, mainly a matter of prompting for better outputs and became a matter of medium-awareness, disciplined experimentation, and control.

This Is KIPL

This sequence is my Kennedy Idea Propulsion Laboratory in miniature. It is not interested in frictionless automation stories. It is interested in what kind of medium AI is becoming, what kind of governance it requires, and how to work inside it without surrendering judgment.

That is the practical side of these essays. The more personal side is simpler. They helped me see that the real work was not getting AI to perform on command. It was learning how to investigate it, govern it, and, when appropriate, play it.

Reading Seabrook’s Bowie in Berlin helped me recognize the pattern. Writing Playing the Guardrails helped me hear it. Standing Waves gave me the closing note. Looking back now, I can say that this run of essays was my Low: not a polished conclusion, but the record of a method shift.

What Low left behind was not just a record, but a vocabulary: fracture, restraint, atmosphere, and a path that later post-punk and experimental music would keep following. In my much smaller way, that is what I was listening for here too.

Coda

I think I may also have written the first track of what could become my Heroes album.

But next albums need incubation.

That one does not belong here yet. I expect to hear more of it later, if it is real.

[Originally posted on DennisKennedy.Blog (https://www.denniskennedy.com/blog/)]

Like this post? Buy me a coffee

DennisKennedy.com is the home of the Kennedy Idea Propulsion Laboratory

DennisKennedy.Blog is part of the LexBlog network.