Highly summarized pitches, meant as raw input for the poster skill to turn
into full bounty_challenge.md drafts — not full specs themselves.
Pitch: Compute the first 100 digits of Pi after the decimal point. Input: none (self-contained math task). Submit: a text file with the digit string. Judged by: exact string match against a precomputed reference value. Why it's a good demo: the original inspiration — zero setup, instantly verifiable, but requires an actual algorithm (e.g. Chudnovsky/Machin-like series), not something you can eyeball.
Pitch: Compute the 5000th Fibonacci number, exactly (full integer, no rounding). Input: none. Submit: a text file with the integer. Judged by: exact match against a precomputed reference value. Why it's a good demo: trivial to verify, annoying to do by hand, easy with a loop/matrix-exponentiation algorithm.
Pitch: Given a ~60-bit number that is the product of exactly two prime
numbers, submit the two factors.
Input: the number N (provided in the bounty page).
Submit: the two prime factors p and q.
Judged by: checking p * q == N and that both are prime.
Why it's a good demo: trial division alone won't finish in reasonable
time, so it nudges solvers toward a real algorithm (e.g. Pollard's rho),
while grading stays a one-line check.
Pitch: What is the 100,000th prime number? Input: the index N (100,000). Submit: a single integer. Judged by: exact match against a precomputed reference value. Why it's a good demo: needs a sieve or similar, not enumerable by hand, single-value grading.
Pitch: Find a string whose SHA-256 hash starts with 0000 (four hex
zeros).
Input: the required prefix (0000), optionally a fixed suffix/nonce
format.
Submit: the string and its hash.
Judged by: recomputing the hash and checking the prefix.
Why it's a good demo: playful "mining" framing, difficulty is tunable
by adding more required zeros, deterministic to verify.
Pitch: Solve one specific 9x9 Sudoku puzzle (given as a grid with blanks). Input: the puzzle grid. Submit: the completed 9x9 grid. Judged by: checking row/column/box constraints and that it matches the given puzzle's clues. Why it's a good demo: visually fun, clearly requires a real solver (backtracking/constraint propagation), unique-solution grading is simple.
Pitch: How many steps does the number 27 take to reach 1 under the Collatz process (n -> n/2 if even, else 3n+1)? Input: the starting number (27). Submit: a single integer step count. Judged by: exact match against a precomputed reference value. Why it's a good demo: one-line algorithm, cute mathematically, instant single-value grading.
Pitch: Submit 1000 samples of sin(2*pi*f*t) for t = 0..999 at a
1Hz sample rate, with f = 0.01, as a CSV with columns t,value.
Input: the formula and parameters (f, sample count, sample rate).
Submit: samples.csv with 1000 rows.
Judged by: recomputing the expected value per row and checking within
a numeric tolerance (e.g. 1e-6).
Why it's a good demo: the "data" analogue of the Pi bounty — trivial
formula, but exercises a data-submission bounty shape instead of a
single-answer one.
Pitch: Submit a CSV of 24 hours of synthetic heart-rate data at
1-minute resolution (1440 rows), integer bpm values, with mean ~70 and
stdev ~5.
Input: the target distribution parameters (mean, stdev, resolution,
duration).
Submit: heart_rate.csv with columns timestamp,bpm.
Judged by: checking row count, value range (plausible bpm bounds),
and that sample mean/stdev fall within tolerance of the targets.
Why it's a good demo: stands in for real "provide wearable data"
bounties, but judged on statistical properties rather than an exact
match — shows off a different acceptance-criteria shape.
Pitch: Submit 10,000 simulated fair six-sided die rolls as a CSV, plus
the empirical distribution (counts per face).
Input: number of rolls (10,000), number of sides (6).
Submit: rolls.csv (10,000 rows) and a summary of counts per face.
Judged by: a chi-squared (or simple tolerance) check that the
empirical distribution is close enough to uniform.
Why it's a good demo: another statistical-grading data bounty, playful
framing, easy to generate and easy to sanity-check.