Unreviewable Fluency
The better the work looks, the less it may tell the person who has to sign it. A name for the condition and the opening case for a new discipline.
Thinking
The economics changed. The operating model has to follow.
The essays begin where the senior person feels it: the work looks finished and cannot be read. From there they follow the failure upstream to where it was made, then outward, to what keeps two signals independent, to what compounds and what leaks, and to the economics that changed underneath all of it.
The canonical pages hold the few stable concepts and relationships. The essays discover, narrate, challenge, and extend them.
If you arrived from one of the essays: this is what they add up to, and what we do about it.
01 · Stable public canon
Three doors into one operating system.
The condition, the response, and the commercial implementation.
The condition
Unreviewable Fluency
Why fluent work stopped carrying reliable information about the thinking underneath it.
Canonical definition →The operating responseThe re:compound Method
How governed production, designed reviewability, independent review, human reliance, and carry-forward form one loop.
Inspect the method →The implementationThe ten-week Installation
How the complete capability is installed in one partner’s team, a pair at minimum, and transferred to independent use.
See the installation →02 · Essays in reading order
01
The symptom you feel first, and its upstream causes
The old senior instrument was reading production residue. AI flattened it. That is the symptom, and it appears at review. The causes begin earlier, and in your hands too: a hypothesis goes out and comes back as corroboration, and a first draft freezes an understanding it has not yet earned.
A Consensus of One
Three days in which an untested hypothesis comes back as confident thesis, even though nobody lies and no model hallucinates. The two causes of unreviewable fluency, watched forming at every desk upstream.
02
Independent signal and the conditions for inquiry
Agreement is not corroboration. Useful collaboration depends on remaining differently wrong, on a human governor able to exploit the difference, and on conditions that let a weak signal survive long enough to count.
What Apollo Knew That Silicon Valley Forgot
More agents do not create independent corroboration when their errors are correlated. The value lies in different signals under sparse, explicit human governance.
A Temporary Little Civilization
A practical account of the conditions under which two very different cognitive systems can sustain difficult, exact, compounding work together.
03
Compounding, re-entry, and apprenticeship
More exact understanding is the unit of progress, not more polished output. A summary preserves what was said. A checkpoint preserves what now matters and lets the next return begin higher. Judgment that can be read can be handed on.
One Bet or Two
A family business, a succession, and AI in the room. Whether a successor can be underwritten depends on whether their judgment can be read, and six lines make it readable.
The Memo She Couldn’t Write
What a senior partner carries is not the file but a maintained live state. The right handoff object is designed for re-entry at altitude.
The Most Expensive Thing Your Organization Throws Away Every Day
AI made generation cheap. The new bottleneck is preserving the changed state of understanding from which the next real move can begin.
The Constant of Integration
Why senior judgment can’t be transferred by briefing: it is an integration over history, and the constant can’t be transmitted. What can be transmitted instead.
04
The changed economics, and what must change in the work
Task execution became cheap and governing the inquiry became the scarce skill. Economics that change this much do not leave the work as it was. These essays say what has to change: what the senior person is for, how juniors come to have judgment, and where a firm sends the capacity it frees.
Reskilling for What?
AI has made doing tasks cheap. The scarce skill now is governing the work: framing, steering, selecting, integrating, and preserving what should improve the next cycle.
The AI Layoff Trap Is a Fiduciary Problem
The real choice is not whether to capture AI efficiency. It is where the freed human capacity goes: cost takeout, more automation, or multiplied institutional capability.
A practical first step
If one of these described your Tuesday.
The essays name the changed condition. The installation changes the work. Thirty minutes with Bud is the next step.