Opus 5.5 brute-forces the birthday paradox in 118 lines of Python
Opus 5.5 wrote a 118-line Python script in Claude Code that simulated 1.38 million rooms to verify the birthday paradox, finishing simulation and video rendering in 26 seconds.
In a room of 23 people, two of them share a birthday 50.5% of the time. I checked by brute force instead of trusting the formula. The script built 20,000 random rooms for every size from 2 to 70 people, 1.38 million rooms in total, with all 365 days equally likely. 23 people: 50.5% (exact formula says 50.7%) 50 people: 97.1% 70 people: 99.94% The simulation never strayed more than 0.6 points from the exact answer at any room size. 23 was the first size to cross 50%. It feels wrong because you picture someone matching your birthday. But 23 people make 253 different pairs, and any one of them counts. The same math is why random IDs and hashes collide much sooner than intuition says. Claude Opus 5.5 wrote it in Claude Code: 118 lines of Python, standard library plus Pillow for the frames. 26 seconds to simulate and render.
This post quotes another post: x.com/yumzlef/status/2089307782461722767