Compounding vs Additive Gains

Take any improvement you’re paying for — a tutor, a tool, extra hours — and imagine it gone tomorrow. If you would still be more capable than before you had it, the gain changed your state, and the next gain builds on the new baseline. If you would snap back to where you started, the gain was additive: real while it lasted, gone when the input stops. Compounding works the way compound interest works: $100 earns $1, the balance becomes $101, and the next 1% is calculated on the higher number, so the gap widens every cycle. Adding 1% a day in a straight line yields 365% in a year; compounding the same 1% yields roughly 3,600% — about 37 times the start. The entire difference is whether each gain updates the balance. In learning, you are the balance.

The removal test

• Imagine the input gone. If you remain more capable without it, the gain changed your state and can compound. If you land back at baseline, the gain was additive — you rented a result.
• Same hour, different purchase. Both kinds of gain consume the same time and effort, so every additive hour is an hour that could have bought a compounding gain. On any single day the difference is invisible — $1.01 versus $1.00 — and over a year it is multiples.
• Expect the slow start. The compounding curve sits below the additive line early, then overtakes it by several fold. The additive line rises linearly, plateaus, and in some cases turns down.

Borrowed capability fails the test

• Tutoring lends gear. Extra resources, guidance, and explanation raise performance above real ability — a level-one character with level-six gear clearing level-five content. Remove the gear and the character is still level one; knowledge and results drop with it.
• Reliance scales the wrong way. Content gets harder and targets rise, so the borrowed support has to keep growing — until the binding constraint becomes how much your own brain can hold in the time available, which never grew. The fade-out and teacher-dependency research behind this lives in WPW.
• Hours behave the same. One hour of study buys an 80% result and two hours buy 90%, but drop back to one hour and the result drops with it. Past a threshold, added hours run the curve backward: stress climbs, efficiency falls, and sleep, friends, and hobbies pay the bill.
• Some additive spend is unavoidable. A purely compounding mix doesn’t exist; the discipline is knowing which kind of gain you’re buying before you spend.

Skill gains pass — and transfer everywhere at once

• Skills change the state. Recall of details, conceptual understanding, problem-solving from routine to curveball questions, execution at speed under time pressure: improving any of these upgrades the person who shows up to every future hour, so the next gain builds on a higher floor.
• Transfer multiplies the rate. Get 1% better at recalling details with no extra time, and the gain applies across all five subjects at once — effectively 5%. Subject-specific help has no such reach: science tutoring does nothing for maths.
• Same brushwork, different paint. Subjects differ the way blue paint differs from yellow; the skill of learning them is the same technique with the brush, which is why a gain in understanding one domain surfaces in the others.
• Output rises at constant effort. Changing how the hour is spent lifts the result of that same hour — 80 to 90 to 100 — with no added time and no rented support.

The grade is a speedometer

• Wanting the outcome changes nothing. Punching through a wall barehanded stays slow and bloody however much you want the hole; the odds move when the method moves.
• The result reads out the process. A grade indicates studying skill the way a speedometer indicates speed — tapping the dial doesn’t move the car. Improvement effort belongs on the processes upstream of the number.
• The outcome arrives as a consequence. Someone with strong technique, recall, understanding, problem-solving at speed, plus calm and time management is hard to imagine missing the top result. Aim the gains at becoming that person.

Running the test on any spend

• Courses and tools. A tool that performs the work delivers output; a tool that trains a capability you keep delivers state. The removal test separates them instantly.
• AI assistance. Drafts, code, and analysis produced for you are gear. The spend compounds only when the interaction leaves you better at the work once the assistant is closed.
• Knowing counts for nothing until used. An unpracticed technique might as well be unknown — familiarity breeds false security, and the nuances only surface in use. Until a technique runs regularly in real work, the removal test hasn’t even started; the full diagnostic lives in Are You Learning, or Just Using Techniques.

Links into the system

The state-change test picks which gains Marginal Gains should stack, and Marginal Gains in Practice runs the stacking; Loss Aversion explains why additive comfort feels safe enough to keep buying; Locus of Control owns the wider rule of spending attention on process instead of outcome; the dependency research behind borrowed capability lives in WPW.