The tricks behind the speed

The same trick scales. For 27 + 45 + 33, don't add them in order — scan for pairs whose ones digits add to 10. Here 7 + 3 = 10, so pair 27 and 33 first: 27 + 33 = 60. Then 60 + 45 = 105. This "friendly pairs" habit is why some people seem to add long columns instantly — they're not doing it left-to-right, they're finding shortcuts inside the column.

You were drilled to align columns and start from the ones digit because that's how paper subtraction handles borrowing. But in your head there's no paper, and starting from the biggest place value means you often already know the answer's order of magnitude before you've finished. That's also what makes mental math fast to verify — if 47 + 38 gave you 195, you'd catch it instantly by having thought "40 + 30 = 70" first.

For subtracting from a round number like 100, 1000, or any power of 10, use complements to 9 (and 10 at the very end). To compute 1000 − 476: for each digit except the last, subtract from 9; for the last digit, subtract from 10. So 476 becomes (9−4)(9−7)(10−6) = 524. This is the same trick cashiers use to make change quickly.

Borrowing forces you to modify a digit you already "finished", which creates two moving parts in working memory. Counting up moves in one direction only — you're just answering "how much further?" three times and summing the jumps. Same answer, half the cognitive load.

For a two-digit number "AB", multiplying by 11 is really (A×100 + A×10) + (B×10 + B) = A×100 + (A+B)×10 + B. That's the same as writing "A _ _ B" with the sum A+B in the middle. When A+B ≥ 10 you carry: 78 × 11 → 7 _ (15) _ 8 → 8 _ 5 _ 8 = 858 (the 1 from 15 carries into the hundreds).

You can double and halve indefinitely because you're changing the shape of the product, not its size. This unlocks a family of shortcuts when one factor is 25 or 75. For 25 × (any multiple of 4): halve the multiple twice, then × 100 — e.g. 25 × 12 → halve to 6 → halve to 3 → × 100 = 300. For 75 × (any multiple of 4): same halving, then × 300 — e.g. 75 × 8 → 4 → 2 → × 300 = 600. Any time you spot a factor of 25, 50, or 75, look for a matching factor of 4 or 2 to combine with — the answer usually collapses to something × 100.

(10A + B)(10C + D) = 100·AC + 10·(AD + BC) + BD. That expansion has three groups — one for the hundreds place (AC), one for the tens (AD + BC), one for the ones (BD). The "vertically and crosswise" picture is just those three products drawn as a diagram: multiply the digits above each other (vertically) for the outer places, and multiply crosswise for the middle. It's basic algebra, dressed up as a visual pattern.

For ABC × DEF (both three-digit), you get five columns instead of three. Rightmost is C×F. Next is B×F + C×E. Middle is A×F + B×E + C×D. Then A×E + B×D. Leftmost is A×D. Same idea, just more crosswise pairs. In practice most people stop at two digits — beyond that, either estimate or grab a calculator.

In base 10, every power of 10 leaves remainder 1 when divided by 9 (10 = 9+1, 100 = 99+1, 1000 = 999+1, and so on). So a number like ABC = A·100 + B·10 + C has the same remainder when divided by 9 as A + B + C does — each place contributes just its digit. The same logic works for 3 (since 10, 100, 1000… all leave remainder 1 when divided by 3 too). This is also why "casting out nines" works as a checksum — old-time accountants used it to catch arithmetic mistakes.

There's no clean digit-based test for 7, but there's a tail-doubling method: take the last digit off, double it, and subtract from the rest. Repeat until the number is small enough to check. Example — is 203 divisible by 7? Take off the 3: 20 − 2·3 = 14, which is 7×2, so yes. It works but rarely saves time — for 7, most people just divide. Useful mostly as a party trick.

Divisibility rules multiply cleanly when the factors are coprime. That's why the rule for 6 is "divisible by 2 and 3" — because 2 and 3 share no common factor. The same idea gives you: 12 = 3 and 4, 15 = 3 and 5, 18 = 2 and 9, 45 = 5 and 9. But "divisible by 4" isn't just "divisible by 2 twice" — 6 passes the 2-test but isn't a multiple of 4.

If a shirt is $80 after a 20% discount, the original price isn't $80 + 20% = $96. That would be adding 20% of $80, but the discount was 20% of the original. The right approach: the sale price is 80% of the original, so original = $80 ÷ 0.8 = $100. General rule: to reverse an X% change, divide by (1 ± X/100). Getting this wrong is the most common percent mistake in everyday life.

If a poll goes from 40% to 44%, that's a 4 percentage-point increase — but a 10% relative increase (because 4 is 10% of 40). News articles conflate these constantly. When someone says "interest rates rose by 2%", ask whether they mean 2 percentage points (huge) or 2% of the current rate (tiny). It changes the answer by orders of magnitude.

A number ending in 5 has the form 10n + 5. Squaring: (10n+5)² = 100n² + 100n + 25 = 100·n(n+1) + 25. That last line is literally "n(n+1) with 25 tacked on". The shortcut is the algebra, in disguise.

For any two-digit number close to 50, use: n² = (50 + d)² = 2500 + 100d + d². So 53² = 2500 + 300 + 9 = 2809. Or 47² = 2500 − 300 + 9 = 2209. Same shape as the "near a base" trick, specialized for 50. Great for the 45–55 range where neither the "ends-in-5" trick nor the "near-100" trick is a clean fit.

Once you have a rough estimate x for √n, you can improve it with x_new = (x + n/x) / 2. Example: to refine our √50 ≈ 7.07 estimate: 50 / 7.07 ≈ 7.072, average with 7.07 = 7.071. That's four correct decimals from one step. This works because if x is too small, n/x is too large — averaging them meets in the middle. Every iteration roughly doubles the number of correct digits.

Estimating square roots mentally is useful when you're computing standard deviations, RMS values, or diagonal distances — anywhere a Pythagorean formula appears. For example, the diagonal of a 5×12 room is √(25 + 144) = √169 = 13 exactly. But if the room were 5×11, you'd need √146 ≈ 12.1 — bracket between 12² = 144 and 13² = 169, position ≈ 2/25 ≈ 0.08.

Because 2¹⁰ = 1024, roughly 1000, powers of 2 and powers of 10 stay in step: 2²⁰ ≈ 10⁶ (a million), 2³⁰ ≈ 10⁹ (a billion), 2⁴⁰ ≈ 10¹² (a trillion). This is why kilobyte, megabyte, gigabyte, terabyte match up with 10³, 10⁶, 10⁹, 10¹² even though they're really powers of two. Handy when converting bit sizes to human sizes.

Anything that doubles regularly grows staggeringly fast. Fold a piece of paper 42 times and it reaches the moon (2⁴² ≈ 4.4 × 10¹² sheets thick, and paper is about 0.1 mm). Compound interest at 7% doubles roughly every 10 years, so $1 000 becomes $8 000 in 30 years. Every time you see "doubles" in a problem, reach for the powers-of-2 table — you'll often skip several steps of arithmetic.

Cubes come up whenever volume is involved. Doubling the side of a cube gives 8× the volume. Tripling gives 27×. That's why a small increase in package size (or a bigger pizza) often gives much more than you'd expect — you're paying for a linear increase but getting a cubic one. Same rule for pricing spheres, watermelons, and ice cubes.

The Fibonacci sequence — 1, 1, 2, 3, 5, 8, 13, 21, 34… — is defined by "each term is the sum of the previous two". What's surprising: the ratio of consecutive Fibonacci numbers converges to φ ≈ 1.618, the golden ratio. So Fibonacci behaves geometrically in the limit, with ratio φ. This appears in flower petals, spiral shells, and, unfortunately, more Instagram design captions than it deserves.

If second differences are constant, the sequence is quadratic (like squares). If third differences are constant, it's cubic. If Nth differences are constant, the underlying formula is a polynomial of degree N. This is Newton's method of finite differences, and it's how mathematicians historically fit polynomials to data before calculators existed.

Sometimes the pattern is an index rule: term n is n² + 1, or 2ⁿ − 1, or n(n+1)/2. Try substituting n = 1, 2, 3 into simple formulas and see what matches. If even that fails, you might be looking at primes, digits of π, or a puzzle that's supposed to have no rule. Those are rare in this app — every Sequence-mode problem here has a clean answer.

To find how long an investment takes to double at a given interest rate, divide 72 by the rate. At 6% annual growth, doubling takes 72 ÷ 6 = 12 years. At 9%, it takes 8 years. The rule works because ln(2) ≈ 0.693, and 72 is close to 69.3 while being much easier to divide (72 has factors 1, 2, 3, 4, 6, 8, 9, 12). Bankers have used it for centuries.

Physicist Enrico Fermi was famous for estimating impossible-sounding quantities — "how many piano tuners are in Chicago?" — by chaining rough guesses. The technique: break the question into factors you can roughly estimate, and your over- and under-guesses often cancel. Say 3 million people, roughly 1 piano per 20 people → 150,000 pianos. Tuned once per year → 150,000 tunings. One tuner does 4 per day × 250 working days → 1,000 tunings per year. So 150,000 ÷ 1,000 → about 150 tuners. The exact answer isn't the point — knowing it's "about 100", not "about 10,000", usually is.

Most real-world estimates have input error of at least 5–10% (you don't know the exact bill total, tax rate, etc). Adding more precision than the inputs deserve is fake accuracy. The right habit: round every input to one or two significant figures, do the arithmetic, and round the answer to match. If someone hands you an answer with more precision than the inputs justified, be suspicious.

The order isn't a law of mathematics — it's a shared convention so we don't need parentheses everywhere. It was standardized in the late 1800s alongside modern algebra notation. Before that, mathematicians used their own local conventions and disambiguated with grouping symbols. If humanity started over today, we might choose a different order, but PEMDAS is what we all learned, so it's what we all use.

Two related problems: (1) "implied multiplication" (writing 2(3) instead of 2·3) is treated by some conventions as higher priority than explicit ÷, and by others as equal. (2) The ÷ symbol itself is ambiguous — real mathematicians use / with clear grouping or write fractions. If an expression is genuinely ambiguous, the right answer is to rewrite it, not to argue about PEMDAS. Well-written math should never depend on knowing the order-of-operations rules for equal-priority cases.

Work from the innermost outward. For 3 + (4 × (5 − 2)²), start with 5 − 2 = 3, then 3² = 9, then 4 × 9 = 36, then 3 + 36 = 39. Some notation uses brackets [ ] and braces { } for nested groups to make the levels visually distinct — mostly a style choice, but it helps when things get deep.