Each dot is an essay; a bigger dot is a longer read. Pick an island to jump to its essays.
Tap an island to open it
Each dot is an essay; a bigger dot is a longer read. Pick an island to open its essays.
The newsvendor problem began, by later accounts, with a bank deciding how much cash to hold in 1888; it shows why the average forecast is the wrong order quantity.01
Jevons saw cheaper efficiency raise total use in 1865; whether cheaper AI analysis does the same for a planning team depends on one number.02
Capablanca worried in 1926 that memory had made top chess too safe; Fischer, then Kasparov, show where a team's real skill sits once the book runs out.03
Garrett Hardin said a shared pasture must be ruined; Elinor Ostrom went looking and found rules that kept commons alive, and the same rules fit a shared production line.04
In 1884 a Signal Corps sergeant claimed near-perfect tornado forecasts, and a simple rule would have beaten him by saying nothing; scoring your own probabilities exposes the same trick.05The newsvendor problem began, by later accounts, with a bank deciding how much cash to hold in 1888; it shows why the average forecast is the wrong order quantity.
A careful retailer held a data week, learned how bad its records were, and still kept the launch date. The fields no planner reads decided what happened next.
Aristotle tells of a philosopher who reserved every olive press before a big harvest; in 2020 some chip buyers reportedly gave up their place in line before a rebound.
In Gogol's Dead Souls a swindler profits because the census still lists the dead as alive; a replenishment engine trusts its on-hand record the same way.
A genuine disagreement inside the field, with an Indian foundry's real before/after numbers.
The Civil War gave English a word for goods whose low price hid their failure; the Dreamliner shows the same hole in a sourcing paper, with delay in place of rot.
Hewlett-Packard raised its factory costs on purpose and still cut total cost by a quarter, by moving one step to the other side of the ocean.
Nokia, Ericsson, a factory fire, and the difference between a supply chain that bends and one that breaks.
Suetonius blames drought for Rome's grain shortage, yet every remedy he credits to Claudius aims at the reliability of the sea route.
Ford freed a fortune without selling one more car, and Dell's 2002 filing shows the same trade in numbers a planner can recompute.
The Rosetta Stone shows what it takes to read one message in two languages, and a made-up product family shows the same step in an S&OP meeting.
Seventy years of research says an equation often matches expert judgment, and a study of supply-chain forecasts shows where your manual overrides sit on that record.
A number you can compute from your own order data this afternoon, and the four reasons it's probably bigger than 1.
A calculable number to replace the platitude about diversifying suppliers, plus a 2011 case that shows exactly what happens when nobody ran it.
A 1983 HBR framework that still beats most AI-scored supplier-risk dashboards, because it forces a strategy decision instead of a number.
The one technical question that separates a real digital twin from a rebranded what-if model. Ask it before you sign the next vendor contract.
Jevons saw cheaper efficiency raise total use in 1865; whether cheaper AI analysis does the same for a planning team depends on one number.
Forecasting contests have crowned opposite winners for decades; the new pretrained models join the field, and your own data has to pick.
Kasparov explained a 2005 chess upset as a result of process, and planning teams that allow overrides can test the same idea on their own log.
Michael Hammer's 1990 account of Ford's payables shows why an AI agent should arrive after the delete pass, and what the delete pass must protect.
Keynes said job loss from saved labor would be a temporary phase; the bank teller shows the arithmetic that decides how long the phase lasts.
Stanford payroll data finds early-career workers losing ground in AI-exposed jobs, and for a planning team the sentinel is also the supply of future seniors.
Maskelyne's 1767 Nautical Almanac was built so that agreement between independent hands counted as evidence. A bank's risk spreadsheet shows how far one unchecked divisor can travel.
AI coding tools make you feel faster. METR measured it, and the measurement disagrees with the feeling by 39 points.
Air France 447 handed control back to pilots who had never trained for that moment at altitude. An auto-approved queue hands back too, and nobody has planted an error to see who is watching.
A 1948 clock test and a short 1983 paper explain why a planning tool that handles the routine leaves its people worse prepared for the day it fails.
Hanoi once paid for rat tails and got tails without dead rats; a chatbot dashboard can pay for handled chats and get the same gap.
Capablanca worried in 1926 that memory had made top chess too safe; Fischer, then Kasparov, show where a team's real skill sits once the book runs out.
Hume described money that flows in before prices notice, and a video game's economist charts that same gap every month; planners can chart it for stock.
Thomas Schelling asked strangers where they would meet in New York with no way to talk; an AI card game shows what happens when the answer is missing.
Laplace bet millions to one on tomorrow's sunrise, and a Dota bot's public test shows how little a short clean record proves about a plan.
Christiaan Huygens priced a chance as a fixed sum in 1657, and the same pricing shows why a faster step that sometimes fails can lose to the slow one.
Defend, rush or click up: what an Age of Empires II build order says about capacity and investment timing.
Overreacting to one bad fight, or one bad week, makes the next decision worse. The MIT Beer Game shows how far that travels.
What ten thousand hours of Dota taught me about KPIs: the scoreboard can lie, and planning dashboards can too.
Laplace imagined a mind that sees everything; Tetris champions see one piece ahead, and in 2018 taking it away broke them, which says something about how far ahead a plan must look.
Thomas Reid wrote in 1785 that a chain of reasoning is only as strong as its weakest link; Factorio shows the same law in belts, and one difference between links and stages is the whole lesson.
Renaissance mathematicians split a pot by the score and got absurd answers; planning post-mortems that judge a call by its result repeat the mistake.
A Prussian general who distrusted every report still sent out scouts, and the arithmetic he lacked gives planners a ceiling price for any demand signal.
In 1796 an Astronomer Royal dismissed an assistant for slow timings, and a 2013 StarCraft II study suggests planning teams make the same mistake when they count busyness and ignore lag.
Garrett Hardin said a shared pasture must be ruined; Elinor Ostrom went looking and found rules that kept commons alive, and the same rules fit a shared production line.
A young student in Vienna saw that Einstein's theory risked being wrong and Adler's could not be; most business cases are written like Adler's.
A Roman senator showed that the same votes can give different verdicts depending on how they are tallied, and Kenneth Arrow proved the trap has no clean exit; a planning meeting can still pick its rule first.
Homer's hero told his sailors to disobey him on purpose, and a frozen planning window works on the same trick, as long as the person who begs cannot reach the knot.
A perfect record feels like proof and, Hume argued, cannot be one; two simple rules say how much a clean run of weeks still allows.
A donkey between two haystacks starves because it cannot stop choosing; Herbert Simon's answer, good enough, has a price you can calculate before round one of an RFQ.
Safety stock works for a product with two years of history and quietly misleads for the one you launch next month; Frank Knight drew the line in 1921.
Chesterton told reformers to find out why a fence stands before clearing it; a break-even number says how sure you must be before you delete an approval step.
A target can save lives and bend the data around it in the same year; a simple count shows which of your own KPIs has already started to bend.
A perfect map has one flaw, and Lewis Carroll's farmers found it before Borges did; planning models walk toward the same flaw one attribute at a time.
In 1884 a Signal Corps sergeant claimed near-perfect tornado forecasts, and a simple rule would have beaten him by saying nothing; scoring your own probabilities exposes the same trick.
James Lind's tiny scurvy trial worked because the effect was enormous; most business pilots chase small effects, and the arithmetic of the control group says how many stores they need.
A physician's perfect record taught him the worst lesson, and the same trap sits in your demand history: some SKUs teach you, and some only flatter you.
Francis Galton's crowd weighed an ox within a hair, yet insurance underwriters pricing the same policy disagreed wildly; the difference is whether anyone saw the spread.
A surgical checklist cut complications in eight hospitals and moved almost nothing across Ontario, and the gap shows what to measure before you roll out a sign-off.
Smith's pin factory is taught as a story about skill, yet his own text says the time lost between jobs is larger than we would guess.
Montaigne left public life to write, was pulled back into office, and kept calling his book a trial; his reason is a working method for planners who only share finished slides.
Fred Brooks watched a late operating system get later as people were added, and his arithmetic about channels now applies to software agents.
Victorian commissioners admitted that Brunel's railway was faster and still ruled against it, and the same logic asks how far a famous 1956 shipping pilot's saving could reach.
A paperback at the price of a cigarette packet only works if someone else carries the printing risk, and publishers found that trick two centuries before Penguin.
Economists credit Pigou with the congestion toll, but his famous two roads say something else, and Venice's day-tripper fee shows why the difference matters.
An Air Force study found no average airman among thousands, and the same trap sits inside every average delay a planner reports.
A count's promise in 1255 and a king's writ in 1295 give the same rule for shared costs, and a Dutch water board shows how hard it is to keep.
Singapore built a harbor out of other people's cargo, and the arithmetic that made it work also makes it a bet.
A rejected timetable draft became the rhythm of a country, and its rule is to fix the handover first and fit each step to it.
The Kumbh Mela has a bathing day every planner can see coming years ahead, and in 1954 and again in 2025 the failures came on that day, with the counts still disputed.
Harry Beck deleted distance from the Tube map, and a study of London riders found that they trusted the map's distances over their own travel times.
Athens kept Theseus's ship by replacing its planks, and Ise renews its shrine on a fixed cycle; both raise the question a planner faces with an old model.
Japanese rail crews point at a signal and name it, and the famous 85 percent claim behind the habit rests on a study I could not find.
Mumbai's dabbawalas are famous for a Six Sigma rating that one magazine article never gave them, and the trail of that number shows how a spoken answer turns into a statistic.
I redrew myself in twenty visual styles, from a cave wall to a handheld screen, and every plate is code, not an image.
Four cities, four writers, and what's still standing when you go looking for the address.
The Stoics were describing your inbox.
I write when I’ve learned something. Every piece is dated, so you can see the rhythm for yourself.
I send these by hand, only when I’ve written something I’d want to read myself. No schedule, no sales.