Writing
Essays
In-depth essays where AI, business decisions, stoicism, and ADHD meet. Each one takes a single mental model apart — from how to use it in practice to the situations where it fails.
-
AI · MBA
The internal dead internet
Generative AI lowered the cost of writing almost to zero, and the cost of checking did not fall. A company that does not measure the quality of its own knowledge starts reading itself and can, I believe, collapse much like a model trained on its own outputs.
-
AI · MBA
Growth will not split the gains from AI by itself
In the extreme scenario of an Anthropic Institute report, the whole increase in GDP goes to capital and compensation would cost about 9% of GDP. The report is not a forecast, so the essay tests what survives the strongest objection and gives one rule to write down in your company before you automate.
-
MBA · STO
Invert, always invert: the founder's cheapest edge
Munger borrowed a rule from a 19th-century mathematician. Used well, it is the highest-leverage, lowest-cost thinking tool a founder has — and it fails in three specific ways.
-
AI · MBA
What Gartner's 28% actually says
One survey, two denominators: 28% of use cases and 77% of firms. A methodological teardown of the AI-failure statistics before you quote one.
-
AI · MBA
Too much, too fast: the mechanics of expectation inflation
Where sponsor pressure comes from, why hype, escalation, and the planning fallacy inflate it, and the written contract that disarms it before a project starts.
-
AI · MBA
ROI doesn't live in the model
Swap the model for a smarter one and the number rarely moves. A Theory-of-Constraints reading of why AI value lives in the process — and when the model really is the bottleneck.
-
AI · MBA
Pilot purgatory: where AI projects go to die
The average organisation scraps 46% of its AI proofs-of-concept before production, and Gartner forecast 30% of GenAI abandoned after PoC. Why the pilot is built so it cannot cross — and the four design moves that let it.
-
AI · MBA
95, 80, 28: the war of AI failure numbers
AI failure statistics are a publishing genre with its own economy. Every producer of a number has a business model, and headlines usually drop the differences in denominators that you need to read the result. How to build your own reference class instead of quoting someone else's.
-
AI · MBA
AI-ready data is a power problem
Data readiness looks like engineering — pipelines, catalogues, quality dashboards. When the data exists and may legally be shared, a project can still stall on property: who owns the data, whose budget pays for its quality, and whose position weakens when it flows.
-
AI · MBA
The agent demotions will start after the incidents
A forecast, not a finding — and the diagnosis under it is worth more than the number. When governance is a two-position switch, the first incident leaves only one move: switching the agent off.
-
AI · MBA
The company brain and the knowledge nobody wrote down
Vendors sell a system that holds everything your organisation knows. Hayek's knowledge problem, Walsh and Ungson's retention bins and a 2026 benchmark on company-internal data mark the boundary — together with the field experiment that argues against me.
-
AI · MBA
I taught my bot to confuse important with urgent
A triage rule that welded importance to urgency kept pushing aside what was important and slow. Simon on the unit of attention, Leroy on the debt, and measurements from two of my own systems.
-
AI · MBA
Mainframe and the boundary of GenAI competence
A Gartner forecast about mainframe exits is really a clean experiment on where a model's competence ends. Circle of competence, Chesterton's fence and Hyrum's Law, applied to forty-year-old COBOL — and to the vendor pitch on your desk.
-
AI · MBA
Governance without theatre
A control that fires while nobody answers looks identical in the register to one that works. One question settles it: what would this control have shown if the incident had happened yesterday, and after how many days?
-
AI · MBA
This is not an IT project
The project closes in the week the real work begins and leaves a working system with nobody accountable for it. From Conway's law to Article 72 of the AI Act, and a three-question test you run before the start.
-
AI · MBA
What the winning 28% do
Gartner asked the winners why they won and wrote down the answer. It is a list compiled by the planes that came home. Five traits of a survivor, and five questions you ask before you commit a budget.
-
AI · MBA
One operator, a fleet of agents
In more than 90 percent of the firms the MIT NANDA report surveyed, employees reported regularly using their own AI tools for work, so access to the same models lies on both sides of the line between success and failure with AI. My claim: what separates a one-person studio from an enterprise programme is the internal layer of coordination between employees — a layer with a formula, a price, and a ceiling.
-
AI · MBA
Stop measuring development by what the model does for less
Competency is the correct execution of a known task under predictable conditions, which also describes what the model does for less. Four objections, including the heaviest: no study shows self-determined learning works inside a company.
-
STO · MBA
Antifragility: how to stand on the right side of volatility
Some systems break under stress, some resist it, some gain from it. Taleb's triad and the barbell, with a founder's allocation worked out in hours — and three honest objections.
-
MBA · STO
Via negativa: the discipline of subtraction
The firmest knowledge you hold is negative, so removal is a safer default than addition. Iatrogenics, Lindy, the additive bias — and a calendar audit that returns 18 percent of scheduled time.
-
MBA · ∩
Mimetic desire: why you want what others want
Most goals feel self-authored; Girard's triangle says they are borrowed. Careers, rivalry, feeds — and a scorecard audit that shows whose desire sits in your spreadsheet.
-
ADHD
Flow and hyperfocus: how to enter work, and how to leave it
Csikszentmihalyi measured optimal experience with pagers: three conditions and a channel between anxiety and boredom. What a founder can build into a chosen work block, and why with ADHD the exit matters as much as the entry.
-
ADHD · AI
Your brain is a prediction machine — and that reframes ADHD
Predictive coding recasts attention as precision and ADHD as a miscalibrated loop, not weak will — and turns the fix into environment design.
-
STO · AI
When machines do the execution, taste is the job
As AI collapses the cost of execution, judgment becomes the scarce, defensible thing. Rick Rubin is the proof — and the cautionary tale.
-
MBA
First principles: a real lever with a real bill
A genuine tool that occasionally buys a decade of edge, and expensive enough that misusing it is how smart people reinvent the wheel.
-
MBA · ∩
Naval's formula for wealth — and the base rate it hides
Specific knowledge plus permissionless leverage is the solo builder's best hand. The insight is real; the survivorship bias behind it is too.
-
AI · ∩
Cognitive sovereignty: choosing whose model of the world runs in your head
Feeds optimise for engagement, LLMs supply default frames. Owning your attention is a deliberate choice — and it can curdle into a self-made bubble.
-
ADHD · ∩
Engineer the identity, not the willpower
Behaviour flows from identity, not willpower — which reframes everything for an ADHD-shaped mind. And a rigidly built identity becomes a cage.
-
MBA
Your company is a tribe — design the rituals on purpose
Anthropology makes the familiar strange: culture is a system of meaning you're already building. Onboarding is a rite of passage — until "we're a family" becomes exploitation.
-
ADHD · ∩
Mismatch: why a Stone Age brain keeps sabotaging a modern life
Ancient drives firing at modern supernormal targets. A design tool for founders and ADHD alike — and a field that earns its bad reputation.
-
STO
Stoicism is an operating system for uncertainty — not a mug quote
Dichotomy of control, premeditatio malorum, amor fati — a working system for judgment under uncertainty. And the counterfeit sold as strength.
-
STO · ∩
Houellebecq's warning: you can win the game and still have nothing
The dark mirror the optimisation canon needs: a life maximised for freedom can be stripped of meaning. Diagnosis worth taking — prescription worth refusing.
-
ADHD
Deep work for an ADHD brain: blocks, or sprints?
Newport's insight is right; the monastic block is built for a neurotypical brain. Hyperfocus sprints may fit ADHD better — with honest caveats.
-
STO · ADHD
Solitude is a skill, and you have to build the room for it
Not loneliness — the chosen state where original thought happens. For a founder-parent it must be architected. Over-sold, it becomes isolation.
-
MBA
"And then what?" — the two words that separate good decisions from clever ones
First-order effects are obvious; second-order effects arrive late and reverse them. The discipline that catches self-inflicted wounds in pricing and growth.
-
STO · ∩
The cat is an accidental stoic, and it has something to teach the founder
Indifference to approval, presence, not chasing — Stoic virtue a cat never studied. A temperament worth borrowing, and why you can't become one.
-
MBA · ∩
Stop fighting symptoms: the structure is the problem
Recurring problems are your system working as built. Change the structure, not the symptom — and know when to grab the bucket first anyway.
-
∩
Writing is thinking: the case for long formats in a short world
You don't know what you think until you write it at length. The essay forces coherence; long horizons are where compounding lives — with the traps of depth named.
-
STO · MBA
Not everything that counts can be counted — but "meaning over metrics" is a dangerous excuse too
Goodhart's law kills dashboard-driven companies; "meaning over metrics" is the phrase founders use to dodge accountability. Telling a good metric from one that ate the goal.
-
∩
You are playing a status game right now — the only question is which one
Most behaviour is status-seeking in disguise: dominance, virtue, success. Seeing the game lets you choose a better one — and "I've opted out" is the highest-status move.
-
STO · ∩
The second life: what changes when you actually believe you get only one
A second life begins when we stop treating time as something guaranteed. On memento mori, deliberate choices, and the traps of productivity.