Essay · 2026
Start Here: What Survives When The Surface Changes?
A short front door to the publication: the lens, the reader paths, and the free instruments for making better judgement usable.
Start here if AI makes you feel behind.
I do not write tool roundups.
I do not try to recap every model release.
I publish essays, Notes, and free instruments for one recurring question:
What survives when the surface changes?
That question is the front door to the whole publication.
Most AI commentary moves at the speed of the surface: model releases, benchmark jumps, tool launches, prompt tricks, funding rounds, leaderboards, screenshots, and arguments about who is suddenly ahead.
Some of that matters.
But if you only track the surface, you keep inheriting the feed’s anxiety. Every release feels like a reset. Every demo feels like a threat. Every new tool asks you to rebuild your map of the world.
The work here starts one layer lower.
The surface changes fast.
The substrate decides what compounds.
The Lens
Every system has a canopy and a substrate.
The canopy is the visible part: the demo, interface, prompt, benchmark score, dashboard, feature, public story, or model people are talking about this week.
The substrate is the part that still matters when the canopy gets repriced: the data-quality pipeline, evaluation contract, workflow integration, distribution channel, trust mechanism, relationship, failure memory, recovery path, or decision rule.
The mistake is not caring about the canopy. The mistake is confusing visibility with durability.
In AI, the canopy is often the model, prompt, framework, or benchmark score. The substrate is the surrounding machinery that decides whether output becomes reliable action: evals, contracts, traceability, recovery paths, user workflow, proprietary signal, and the human decision the system exists to improve.
In investing, the canopy is often the story everyone can repeat. The substrate is the channel, coupling, asymmetry, optionality, or structural durability that still exists if the story turns out to be partly wrong.
In knowledge work, the canopy is the note count, dashboard, folder, or AI summary. The substrate is whether any of it changes what you believe, decide, search for, or review at the moment of use.
That is why the central question is not:
Is this impressive?
The question is:
What still has a job after the environment changes?
The Test
Durability becomes useful when you stop treating change as a vibe and name the disturbance.
Not “AI gets better.”
Name the actual disturbance:
An open-source model matches your benchmark and runs on commodity hardware.
Not “markets get worse.”
Name the actual disturbance:
Your sector takes thirty percent multiple compression.
Not “distribution changes.”
Name the actual disturbance:
Your main channel stops favouring your format.
Once the change is named, the test becomes simple:
How much of what you built still has a job?
That is displacement rate. It is the honest measure of durability.
A thing can look weak today and be hard to displace. A thing can look dominant today and be a canopy bet waiting for the next shift to expose it.
This publication builds instruments around that question.
Not just takes.
Instruments.
The point is not vocabulary. The point is better judgement under moving conditions.
Start With The Problem You Have
You do not need to read everything.
Start from the problem you actually have.
I want to understand the core lens.
Read The Forest Floor Is The Product.
That is the cleanest statement of substrate versus canopy, and the best starting point if you want the philosophy behind the publication.
I build products, teams, systems, or roadmaps.
Run The Substrate Map.
It separates canopy work from substrate work across your last 90 days, then gives you a substrate-vs-canopy ratio you can actually use.
Use it when your roadmap looks productive, but you cannot tell what is compounding.
I need to stress-test a product, position, architecture, or bet.
Run The Displacement Rate Audit.
It gives you a 0-5 score for how much still works after a named large change.
Use it before you trust the story, ship the roadmap, defend the thesis, or assume the moat is real.
I invest or allocate capital.
Run The Investor’s Substrate Test.
It scores the structure underneath a single position across five axes, so you are not only reacting to the latest narrative.
I work on AI systems, evals, or agents.
Read Most Verification Is Just Bigger Classification, then You Are Not Comparing Models, You Are Comparing Contracts.
The recurring question is whether the system produces evidence you can replay, trust, and act on, or whether it only produces scores, access, and confident outputs.
The vocabulary is new.
Read The Lens Lexicon.
It defines the ten load-bearing terms behind the publication, with tests, examples, and common confusions.
What you should expect
You will not get generic AI news here.
You will not get tool roundups.
You will not get the posture that every new release requires a new identity.
You will get essays, Notes, worked examples, and free tools built around one recurring promise:
one structural lens at a time, written so you can use it.
Sometimes that means an essay.
Sometimes it means a short Note.
Sometimes it means a worksheet, audit, or map you can run on your own work.
The best outcome is not that you agree with every claim here.
The best outcome is that the next time something looks impressive, you ask a better question:
What is the substrate, and what happens to it when the ground moves?
If You Only Remember One Thing
Do not ask what looks impressive today.
Ask what still has a job after the environment changes.
If a single argument here changed what you were about to trust, the highest-leverage move is to subscribe on Substack. One piece a week, no filler.