I thought I was pulling separate threads—until the ground gave way beneath them.
A post about why we shouldn’t judge AI by whether it mimics humans. Another about synthesisers changing music. A piece on desktop publishing. Each felt like following a separate thread of curiosity—until I realised I’d been circling the same buried structure from different angles.
The moment of recognition came while drafting “Are We Trapped in an AI Cul-de-Sac?” Suddenly I could see the pattern beneath all the previous pieces. I hadn’t been writing random explorations. I’d been conducting an archaeological dig into how transformative tools actually work.
The Excavation
The synthesiser thread seemed to be about music history, but I’d actually found evidence that tools don’t replace human expression; they reshape it. Gary Numan didn’t become less human when he plugged in; he found new ways to be human through the machine.
The desktop publishing thread felt like workplace analysis, but revealed the same pattern. Those sign writers weren’t made obsolete; their craft was metabolised by new tools, absorbed and transformed rather than destroyed.
Each essay was revealing the same underlying structure: technologies succeed not by replacement but through situated integration. My oblique approach—following curiosity through adjacent domains rather than marching toward predetermined conclusions—turned out to be perfect archaeological technique.
The Whole Tapestry
When I stepped back and saw how all the threads wove together, what emerged was an accidental series: AI as Tool, Not Master. Six essays that challenge the replacement narrative dominating AI discourse:
Where Does the Thinking Happen When AI Writes? (August 2025)
Intelligence isn’t in the model; it leaks out in use.Can AI Be Creative? (July 2025)
Novelty has edges; let’s feel them.Let Generative AI Be Itself, Not an Imitation Human (February 2025)
Stop grading fish on tree-climbing.Cold Text, Gary Numan, and the Synthesiser Effect (August 2025)
Gary Numan found new veins of human through the machine.Why LLMs Are More Like Desktop Publishing Than the Industrial Revolution (June 2025)
Democratise, don’t pulverise.Are We Trapped in an AI Cul-de-Sac? (upcoming)
Scoreboard thinking is a dead end; process thinking is the through-road.
Connected excavations
The Power Loom Principle (May 2025): Productivity surges come from work redesign, not raw technology
Why We Keep Misreading Disruption (May 2025): The same critique of simplistic replacement frames
What the Pattern Reveals
The buried structure turns out to be about metabolisation rather than replacement. We’re not heading toward a future where AI replaces human intelligence—we’re seeing new hybrid forms of intelligence-in-context emerge.
The real transformation isn’t happening in boardrooms debating existential risk or labs chasing benchmark scores. It’s happening in the margins: paralegals jailbreaking ChatGPT, teachers generating personalised problems, teenagers storyboarding comics. While we argue about the map, they’re learning the territory through situated practice.
This suggests we’re in for more hybrid intelligence than artificial general intelligence. More distributed coordination than grand unified theories. The Romanian nurse running a language model on a dusty Dell doesn’t need consensus about whether the AI “really understands” anything—she just needs it to work within her existing workflow.
Where This Goes Next
The excavation continues, and new sites are emerging. The pattern I’ve uncovered suggests we’re in for more metabolisation than replacement. More hybrid intelligence than artificial general intelligence. More post-consensus coordination than grand unified theories.
But the archaeological method keeps revealing new directions to explore.
Embodiment: Synthesisers reshaped sound; LLMs reshape language. What happens when AI colonises gesture—robotic kitchens, exoskeletons, VR haptics? The metabolisation pattern may be moving from cognitive tools to physical ones.
Maintenance: Every creative revolution creates invisible guilds—guitar techs, colour separators, sys-admins. What maintenance class is forming around AI right now? Who calibrates the models, curates the corpora, patches the bias drift while the rest of us argue about consciousness? As David Edgerton showed in The Shock of the Old, most of technology is old and most effort is maintenance—yet AI discourse remains obsessed with the newest capabilities while ignoring the growing ecosystem of people keeping these systems actually working.
Tempo: Synthesisers let musicians speed up arrangement cycles; desktop publishing let designers tighten feedback loops from weeks to hours. AI compresses iteration even further—perhaps until the loop collapses into something closer to improvisation. What happens to a craft when iteration becomes indistinguishable from performance?
I’m also curious about your own archaeological sites. Where do you see AI metabolising work in your field—shifting practice without replacing it? What gaps have you noticed between how these tools are marketed versus how they actually get used?
Every deployment context reveals new artefacts. Every situated use uncovers another piece of the emerging pattern. The dig continues.
This series explores how AI changes the way we express ideas and do work, drawing on historical precedents from synthesizers to desktop publishing. Each piece approaches the question obliquely, treating theory as scaffold rather than architecture—letting patterns emerge through exploration rather than forcing them through predetermined frameworks.