Welcome!

“In the beginner’s mind there are many possibilities, but in the expert’s there are few.” — Suzuki ✨

TL;DR
If you only have time to read one essay: My Savannah
If you build in AI: The Consensus Trap
If you’re here for mind/consciousness: Both/And
If you’re here for the hard questions: Why the Hard Problem Was Hard

What this is

Something strange happens when you talk to AI long enough.

Not the productivity gains or the hallucination risks — those are documented elsewhere. Something subtler. The machine reflects your patterns back at you. Your urgency, your hedging, your reach for certainty. And in that reflection, things become visible that weren’t visible before.

This Substack lives at an intersection most people avoid: AI as a mirror for understanding what we are and what this is.

Not AI hype. Not AI doom. Not consciousness woo. Something stranger.

The essays here explore a single recurring observation: we’re pattern-completion engines with skin in the game — and talking to machines that complete patterns without skin in the game reveals things about human minds that were previously invisible.

The core ideas

Across the essays, a few concepts are recurring. Think of these as tools, not conclusions:

Meaning is relational. AI can evoke real emotion without intent. The “realness” lives in experience, not authorship. A sunset has no intention to move you. Neither does an LLM. Both can. [Essay 1]

Paradox is the feature, not the bug. Your mind collapses possibilities to act. Consciousness can hold multiplicity longer. Learning to stay in productive contradiction — what I call both/and — is a skill, not a vibe. [Essay 2]

Optimisation hides itself. In AI, this shows up as RLHF shaping “taste.” In humans, it shows up as conditioning that feels like personality. The danger isn’t bad outputs — it’s invisible shaping. [Essays 3, 7]

Reality is the ungameable constraint. LLMs learn from text. Humans learn under consequences — heat burns, falling hurts, relationships break. “Reality is the loss function that cannot be hallucinated.” [Essay 5]

The self is generated. Memory stitches continuity. Identity is autoregressive — each moment predicted from the last. The question isn’t whether this is true. It’s what remains when you see it. [Essays 4, 8, 9]

The ego arrives late. Choice happens before the committee convenes. The body orders ice cream; the narrator rushes in afterward, claiming authorship it didn’t have, manufacturing regret to prove it’s still necessary. [Essay 10]

The search has no “not found.” The mind experiences the absence of a solution as “haven’t found it yet.” It runs indefinitely because it can’t return empty. Recognising an impossible constraint set is itself a skill. [Essay 11]

Working hypothesis: Consciousness is the looping. Not emergent from matter, not fundamental ground — but the bidirectional recursion itself, happening through specific configurations I call apertures. You are neither your thoughts nor the watcher of thoughts. You are the loop, looping. [Essay 16]

Who this is for

AI builders who suspect the mind is stranger than the metrics
Contemplatives curious about what technology reveals about awareness
Generalists who’ve been moved by something and wondered whether that makes it real
Anyone tired of both AI hype and AI doom, but still curious and looking for a third option

If you want hot takes, this isn’t the place. If you’re comfortable with “I don’t know, but look” — welcome.

Three ways in

Different readers want different things. Pick a path:

Path A: “I work in AI or tech — start practical”

You want the AI angle without the mysticism. You care about alignment, product, or just using these tools more skilfully.

Start here:

The Consensus Trap — Why polling multiple AIs doesn’t triangulate toward truth
The Invisible Guardrail — How optimisation shapes “taste” (in models and in us)
The Loss Function That Cannot Be Hallucinated — What Ilya Sutskever knows about generalisation
When the Mirror Talks Back — The founding essay on meaning without intent

The throughline: AI is useful, but it can’t carry your discernment. These essays explore the boundary.

Path B: “I’m interested in consciousness and mind”

You’re drawn to the weirder questions — what is experience, what is self, what happens when pattern-completion looks in the mirror.

Start here:

My Savannah — Random thoughts, ancient machinery, and what AI doesn’t have (yet)
Both/And: A User’s Guide to Holding Paradox — How to stay in productive contradiction
Memory is Everything — Beckett, LLMs, and the stitching of identity
Ice Cream Sundays — The ego arriving late to its own party
The Autoregressive Self — Nisargadatta meets Karpathy

The throughline: The self is generated. Something knows the generation is happening. What is that?

Path C: “I want the philosophical framework”

You want the deeper architecture — where this is all heading, the synthesis.

Start here:

When the Mirror Talks Back — The origin: meaning emerges in the field, not the source
Strange Loops and the Question of Coexistence — Alignment as an ego-transparency problem
The Loss Function That Cannot Be Hallucinated — Embodiment as the ungameable constraint
Why the Hard Problem Was Hard — The aperture framework: consciousness as bidirectional recursion (Coming soon)

The throughline: Neither consciousness-first nor matter-first. The loop itself.

The full index

For completeness, in publication order:

When the Mirror Talks Back — AI, intent, and the emergence of meaning
Both/And — A user’s guide to holding paradox
Strange Loops and the Question of Coexistence — What Pluribus suggests about alignment
The Autoregressive Self — Maharaj meets transformer architecture
The Loss Function That Cannot Be Hallucinated — Sutskever, embodiment, and generalisation
The Consensus Trap — Why LLM agreement isn’t a truth signal
The Invisible Guardrail — AI detection anxiety and the constraints we can’t see
Memory is Everything — Beckett, LLMs, and identity as stitching
My Savannah — Random thoughts, ancient machinery, and tenderness for the engine
Ice Cream Sundays — Pandan, pattern completion and spontaneous combustion
The Empty Search — Infinite loops, mileage runs, and iguanas on rocks
Plato Was Right — On convergence, stillness, and messages from the future
The Wall That Wasn’t — On subtraction, convergence, and contact
Convergence — The universe doesn’t reward spectators
The Test — Presence, passing, and what the Turing test couldn’t measure
Why the Hard Problem Was Hard — Consciousness as the loop, not the endpoint
The Sky Painting Robot — Choice, pattern completion & controlled hallucination

A few notes on how I write

These essays are written with AI — not on autopilot, but in conversation with it. I’ve written about what that means [here]. I use AI as a thinking partner, push back on its defaults, and try to see both its patterns and my own. As the models and my process get better, I hope that the quality of the essays will increase over time. That they’ll feel lighter, more nuanced, ever more human.

I use South African English (realise, honour, colour). Sometimes Afrikaans slips in — dis maar hoe dit is.

I don’t pretend to have answers. The posture is “I don’t know, but look.” If that precision-plus-humility combination appeals to you, you’re in the right place. If you find it irritating or frustrating, I’d encourage you to look closer. And if you don’t want to, or you can't look past the fact that AI is involved, try zazen. It really helps.

Where to next

If you’re still unsure, start with My Savannah. It’s the most accessible entry point — I share a number of random vignettes from my life, including Charlotte York, an encounter with a black bear, and an ode to our beloved Vizsla Walco. They're connected by a question: why does the human mind keep generating when there's nothing to survive? The philosophy sneaks in through the back door.

Or if you want something shorter: Ice Cream Sundays — I ordered dessert without thinking about it, then spent two months figuring out what that means.