Reduction: what remains when the excess is removed.

A reduction is not a subtraction — it is a concentration. The signal does not change; only the water that was obscuring it is removed. What the reduction reveals was always there. Heat and time made it visible.

#:reduction_grammar

Concentration as Epistemology

What is removed

Water evaporates. Volatile compounds escape. What remains is what was not volatile — the sugars, acids, glutamates, minerals that give the sauce its character. The reduction is an act of disclosure, not creation. Learning science names this signal extraction: removing noise to make signal legible without altering the signal itself.

Irreversibility

A reduction cannot be undone. Add water back and you have something thinner, not the original. This is the ^ quality of committed gesture — once the Maillard reaction has occurred, once the caramelization has happened, the substance has become a new substance. The past is permanent. Work with what it made.

Patience as method

Reduction cannot be rushed past certain temperatures without burning. The process has a natural pace — a thermodynamic constraint that is also a pedagogical constraint. Spaced practice, distributed attention, low and slow: the depth of concentration scales with the time allowed for it.

The moment of finish

A reduction that goes too far is a burned fond. There is a point of rightness — a viscosity, a color, a smell — that is felt before it can be measured. This is expert intuition: the result of enough reductions that the pattern is internalized. The recipe cannot describe the finish; only practice can teach when to stop.

~"reduction_applications"

Where the Principle Transfers

Agent: extend this section. The principle transfers into editing, curation, abstraction, syllabus design, and any domain where removing the inessential is the primary skill.

%draft Editing a draft Remove sentences that are true but inessential. The meaning does not change — it concentrates. A reduced draft is not a shorter draft; it is a draft where the signal is stronger per word.

%syllabus Designing a curriculum A reduced curriculum contains only the generative constraints — the principles that, once learned, unlock everything else. Everything else is noise until those constraints are internalized.

%codebase Abstracting a codebase Finding the interface that hides the implementation detail. The abstraction is a reduction — removing the specifics until what remains is the contract that other parts can depend on.

%composition Visual composition Removing elements until the image holds its structure without them. The painter's reduction is cropping, simplification, value compression — the same epistemology as the sauce.

#:reduction_cs

Reduction in Computer Science

Reduction is abstraction. Remove implementation detail until only the contract remains. The interface IS the reduction — what survives when everything volatile is boiled away.

The abstraction funnel. Implementation detail (volatile) evaporates through layers. What remains at the narrow end is the interface — the contract other code depends on. The reduction is complete when nothing left can be removed.

%interface Interface as reduction An interface hides everything that can change behind everything that must stay stable. trait Read { fn read(&mut self, buf: &mut [u8]) -> Result<usize>; } — that's the sauce after the water is gone. Every implementation detail is volatile; only the signature survives.

%map_reduce MapReduce Literally named after the operation. Map distributes, reduce concentrates. The reduce step combines partial results until one answer remains — the computational equivalent of boiling a stock. The signal was distributed across nodes; the reduction brings it home.

?[compression] Lossless compression Huffman coding, LZ77, zstd — remove redundancy without losing information. The compressed file IS a reduction: smaller, same signal, reversible. Lossy compression (JPEG, MP3) is the culinary reduction: irreversible, and the judgment of what to discard is the expertise.

#>refactoring Refactoring as reduction Extract a function, inline a variable, replace conditionals with polymorphism. Each refactor removes structural noise. The behavior does not change — it concentrates. The code says the same thing with fewer concepts. That's the finish point: when removing anything further would lose meaning.