This page builds every symbol the parent note Euclidean algorithm — GCD computation uses, starting from the idea of a whole number. If you have never seen gcd, mod, ∣, or q,r before, start here at line one. Nothing below is assumed.
The picture is tiles laid in a row with no gap and no overhang.
Look at the top row: three orange tiles of length 4 exactly fill a bar of length 12, so 4∣12 (read "4 divides 12"). In the bottom row, tiles of length 5 overhang the bar of length 12 — the last tile pokes out — so 5∤12 (read "5 does not divide 12").
From the list above, the common divisors of 12 and 18 were 1,2,3,6; the biggest is 6, so gcd(12,18)=6.
The magenta tile of length 6 is the widest tile that still covers both the 12-bar and the 18-bar with no overhang. A tile of 7 would overhang both; a tile of 3 fits but is not the widest. That "widest fitting tile" is exactly what gcd means.
Contrast with the factorization route from 2.5.03-fundamental-theorem-of-arithmetic, which finds this same number by breaking each into primes.
When b does not divide a, tiles overhang. The overhang is the star that makes the whole algorithm run.
For a=18, b=12: only one tile of length 12 fits (q=1), and 18−12=6 sticks out, so r=6. Written out: 18=12⋅1+6. The rule 0≤r<b says the leftover is always smaller than the tile — otherwise you could have fit another whole tile.
Now every symbol in the parent's key line is defined. Read it aloud:
gcd(a,b)=gcd(b,amodb).
In words: the biggest common tile of a and b equals the biggest common tile of b and the leftover. That is precisely the "leftover trick" — any tile that fits a and b must also fit the leftover r=a−bq, because it fits a and it fits every b-tile.
The magenta tile that covers both bars in the figure also lines up perfectly along the violet leftover chunk. Nothing about the biggest tile is lost when we throw away the big bar and keep the small bar + leftover — so we may safely shrink the problem. Repeat until the leftover is 0; at that final step the right-hand number is 0, and the base case gcd(a,0)=∣a∣ from Section 3 hands back the answer — the last non-zero remainder.
Why the topic needs them: consecutive Fibonacci numbers are the worst case — every quotient q=1, so the leftover shrinks as slowly as possible. Because ϕn grows fast, even the slowest case finishes in about log-many steps.