The Atlas
Atlas helps operators judge where AI changes work, where it merely shifts cost, and what to inspect before buying, deploying, or trusting it. Start with 001-010 as the core path; use 100-127 as concept deep dives and reference notes.
Core path — 001-010
- 001By the Time Anyone Looked, the Option Had Closed.The Recognition ThesisCORE
For time-sensitive options and exposures, recognition can bind before capital.
- 002When Uncertainty Pays YouConvexityCORE
Capped loss and open upside describe practical asymmetry; convexity is the separate test that the payoff stays below its chords.
- 003The Most Valuable Thing an AI Can Say Is "I Don't Know"The Invisible FrontierCORE
A tested abstention policy can mark where reliable answering ends and judgment returns to the operator.
- 004The AI Wins You Can Measure Are Table StakesSelectionCORE
Measure the gross gain. Then ask what remains after rivals acquire the same capability.
- 005Cheap to Make, Expensive to CheckThe Verification EconomyCORE
For operators, the decisive price is not generation alone but verification: offload is defensible only when a relevant check is reliable, independent enough for the risk, and cheaper than remaking the work.
- 006The Reverse CentaurWho Really DecidesCORE
A reverse centaur is an org design where people still approve the work, but the premise they check was supplied upstream by the system.
- 007Who Drew the Menu?Agenda ControlCORE
Agenda control is the power to draw the option set before the chooser arrives; the visible pick can only run over what reached the menu.
- 008When a Company Wiki Becomes a Graveyard, Fix the ClockKnowledge EscrowCORE
Knowledge escrow captures operating evidence during the work, then packages it only when a real question arrives.
- 009The Cell Is Not a DecisionRisk, Measured WrongCORE
A risk matrix can triage and route sign-off; colour alone cannot measure the tail.
- 010The Quadrant Is a Price, Not a PlaceThe Map Redraws ItselfCORE
Operational events can remain rare and severe while moving from unwatched improvisation into watched, repeatable handling.
Concept deep dives — 100-127
Reference notes for terms, models, and source ideas used by the core path. They support the spine; they are not a separate theme.
- 100The Shadow Option: The Move Hidden in What You Already BuiltThe LexiconREFERENCE
A shadow option is a possible strategic move embedded in a firm's resource bundle but not yet recognised.
- 101The Recognition Engine: Continuous Cheap Attention on Discontinuous PayoffsThe LexiconREFERENCE
A recognition engine is continuous, cheap attention pointed at discontinuous payoffs: public-but-unread sources scanned, rare flags triaged, decisions handed back to the owner.
- 102The Frequency-Impact Matrix: A Triage Grid Dressed as a RulerThe LexiconREFERENCE
A frequency-impact matrix is a two-axis triage grid: probability on one axis, impact on the other, useful for escalation but too coarse to measure the tail.
- 103Verifiable Space: Where the Answer Can Be CheckedThe LexiconREFERENCE
Verifiable space is work where proposed answers can be checked cheaply against a proof, calculation, measurement, registry entry, or deadline.
- 104The Prevention Paradox: The Disaster You Stop Never Arrives to Thank YouThe LexiconREFERENCE
Rose's prevention paradox pairs large population benefit with little benefit to each participant; quiet-year accounting creates a related problem for firms.
- 105Latent Exposure: The Loss You've Already Agreed ToThe LexiconREFERENCE
A latent exposure is an unrecognised dependency, contract term, or knowledge concentration that can turn today's operating setup into a future loss.
- 106Knightian Uncertainty: When No Defensible Reference Class Supports the OddsThe LexiconREFERENCE
Knightian uncertainty names cases where risk cannot be priced from a reliable reference class.
- 107The Reverse Centaur: The Machine Holds the Reins, You Hold the PenThe LexiconREFERENCE
A reverse centaur is an arrangement where a machine fixes the premise, pace, or target and the human is left to execute, approve, or absorb blame.
- 108The Decision Premise: How Upstream Assumptions Shape the ChoiceThe LexiconREFERENCE
A decision premise is the fact or value assumption a choice treats as already settled.
- 109Range Compression: Five Buckets, a Thousandfold GapThe LexiconREFERENCE
Range compression is what happens when a risk matrix forces continuous probabilities and losses into a few ordinal buckets.
- 110The Real Option: A Right, Never an ObligationThe LexiconREFERENCE
A real option is the right, not the obligation, to make a larger commitment later after paying a smaller premium now.
- 111The Option Chain: One Move Opens the NextThe LexiconREFERENCE
An option chain begins when exercising one option puts you in a new position with possible moves awaiting recognition and investment.
- 112Convexity: When Volatility Starts Paying YouThe LexiconREFERENCE
A function is convex when every chord lies on or above its curve, equivalently when its slope does not decrease; capped loss with open upside is one example.
- 113The Jagged Frontier: The Edge You Can't See Until You're Past ItThe LexiconREFERENCE
A jagged frontier is the uneven edge of model capability: tasks that look equally hard to a person can sit on opposite sides.
- 114Out of Distribution: A Map of Where the Data LivedThe LexiconREFERENCE
Out of distribution is a relative label: a case falls outside a specified reference distribution.
- 115The Base Rate Trap: When the False Alarm Rate DecidesThe LexiconREFERENCE
The base-rate trap is rare-event arithmetic: even a low false-positive rate, multiplied across many non-events, can bury the true hits.
- 116Metacognitive Demand: The Load Shifts from Doing to JudgingThe LexiconREFERENCE
Metacognitive demand is the burden of knowing whether you can judge AI output after the machine has done the producing.
- 117Prediction vs Decision: The Machine Forecasts, You JudgeThe LexiconREFERENCE
Prediction is a claim about what may happen; decision is the accountable choice about what to do.
- 118Differentiated Value: The Win Everyone Can Buy Is Not an EdgeThe LexiconREFERENCE
Differentiated value is the context-bound value that remains after copyable AI savings become table stakes.
- 119Rare for You, Common for the CorpusThe LexiconREFERENCE
Rare-for-you means a case with no local routine but many documented precedents in the corpus.
- 120The Generation-Verification Gap: When the Right Answer LosesThe LexiconREFERENCE
A model can generate a correct candidate and still fail because its verifier cannot reliably select it.
- 121The 70% Problem: Fast to Prototype, Hard to FinishThe LexiconREFERENCE
The 70% problem is Addy Osmani's field observation that AI can make a prototype arrive quickly while production work resists the same shortcut.
- 122The Three Lines Model: Independence Is a Role, Not a HeadcountThe LexiconREFERENCE
The Three Lines Model separates management's action and challenge from internal audit's independent assurance to the governing body.
- 123Tacit Knowledge: We Can Know More Than We Can TellThe LexiconREFERENCE
Tacit knowing is the background integration of cues that lets a person recognize, judge, and act without fully specifying every contributing detail.
- 124The Illusion of Choice: A Menu Hides the Decision Better Than a VerdictThe LexiconREFERENCE
Here, the illusion of choice describes a menu that preserves the feeling of authorship while hiding who framed, ordered, and omitted the options.
- 125Agenda Control: The Draw Shapes the PickThe LexiconREFERENCE
Agenda control is the power to define which alternatives reach the table before anyone chooses among them.
- 126Deskilling: The Skill That Catches the Error Is the One Automation Quietly TakesThe LexiconREFERENCE
Deskilling is the slow loss of unaided skill after automation takes over the routine work a human used to practise.
- 127Affordable Loss: Size the Bet by What You Can LoseThe LexiconREFERENCE
Affordable loss is an effectuation rule for genuinely uncertain ventures: cap the stake by what you can survive losing.