Posts by Peter Evans-Greenwood
The Relationship is the Skill
Whether or not hallucinations are a bug depends on what you’re trying to do
thepuzzleanditspieces.substack.com/p/relationsh...
My latest Substack post, ‘The Relationship is the Skill’, the one where I show how the same AI output was a verification crisis for one person and a prospecting tool for another, and neither of them was wrong.
The Empty Chair
The rebalance is real. The reinvention isn’t here yet. And the industry is preparing for the wrong guest.
open.substack.com/pub/thepuzzl...
My latest Substack post ‘The Empty Chair’. The one where I argue the chair is ready, the preparation is sincere, and the guest won’t be in the room where the invitation was written.
The Physical AI playbook is looking a lot like the Metaverse one
www.noemamag.com/what-the-ai-...
Seth is answering 'what kind of thing would have to exist for consciousness to be present' and Friedland wants to answer 'what does the configuration produce.' Those aren't in tension—they're just different questions.
"Biological not silicon" is being read into Seth's essay by Friedland's own biases. Which makes Friedland's critique less a correction of Seth and more a… change of subject.
Autopoiesis, self-production, is a defining characteristic of living systems. Seth's point is that the self-sustaining metabolic work is constitutive, not incidental. And the thermodynamics argument is the hard one to hand-wave. Living systems must continuously resist the decay.
But Seth's argument is explicitly not a substrate binary. The move is more specific: even a single neuron is a spectacularly complicated biological machine, busy maintaining its own integrity and regenerating the conditions and material basis for its own continued existence.
The consciences debate continues in NOEMA, with Barton Friedland’s ‘What The AI Consciousness Question Conceals‘ attempting to take down Anil Seth's 'The Mythology of Conscious AI'.
The Price Was Never the Point
Pricing symbols are precipitates of institutional form. You cannot change one without changing the other.
thepuzzleanditspieces.substack.com/p/the-price-...
My latest Substack post, ‘The Price Was Never the Point’. The one where I show why BMW’s heated seat subscription failed and Texas Instruments’ identical move forty years earlier became a business school classic. Same logic. Opposite outcomes. The difference is structural.
First Words argued the attention economy ended the subsidy, not the reader. Hollywood ran the same playbook—institutional identity replaced earned commons—and got the same result. Hail Mary is the market correcting.
open.substack.com/pub/thepuzzl...
Writing is not thinking
Writing is the externalisation of thought into symbols and manipulation of those symbols. That's not thinking. It's just one cognitive loop among many.
buff.ly/bRz4Bzh
It’s a failure of quality. ‘Hachette/Orbit’ was attached to a market signal, not to judgment. The imprint was laundering the Goodreads score into a prestige credential and downstream quality control failed. No wonder the publishers are in trouble. AI is just a convenient villain.
First Words
The attention economy didn’t break readers. It ended the subsidy. Attention is the symptom, not the pathology.
thepuzzleanditspieces.substack.com/p/first-words
My latest Substack post, ‘First Words’. The one where I make the case that the first words are an audition—not an introduction. And then run the audition without telling you.
My latest Substack post, ‘The Engine That Changed Categories’. The one where I argue that the beam engine and the LLM are in the same chapter. Neither is the platform. Both are the coalfield.
open.substack.com/pub/thepuzzl...
The SBC reckoning this post described two years ago is now just called Tuesday. Worth reading if you want to understand why the AI-efficiency layoffs were always coming—and why the AI framing is cover for something that was structurally inevitable.
nextbigteng.substack.com/p/unraveling...
Investment says frontier. Outputs say late-period. Trust the inputs? Or the outputs?
Maybe AI is different. Maybe it’s a meta-instrument that generates new categories rather than being one. But that argument needs to do some work before it earns its place. Appealing to it without evidence is just special pleading.
Look at AI through the same lens. The investment curve says foundational moment—capital mobilisation you associate with electrification, or the electric guitar. Output metrics are different: marginal capability improvements, better autocomplete. Real, but late-period in texture.
By the mid-90s the tools were in place and affordable. Change didn’t stop—but it shifted register. DAWs enabled micro-rhythmic refinement, new textures, production detail. Real change, but at a finer grain. No new timbral primitives.
Music changed faster 1950-1995 than it has since. The reason isn’t mysterious: each decade had a new instrument category. Amplification, electric guitar, synths, samplers. Each opened a new possibility space, and musicians explored it. Music as practice & experience transformed.
The Engine That Changed Categories
How the steam engine stopped being a tool and became a platform—and what that tells us about AI
thepuzzleanditspieces.substack.com/p/the-engine...
The 1990s–2000s boom was the category management story, repeated across the economy. Just-in-time manufacturing, yield management in airlines…
Which raises the obvious question about AI.
This provides a mechanism for both the productivity paradox and adoption lag. The paradox exists because the PC delivered narrow tool productivity—which doesn't show up in TFP. The lag exists because building the digital network takes time.