Ron Rivest locked MIT's time capsule with a puzzle that takes decades of sequential math

In early April 1999 a time capsule was delivered to architect Frank Gehry with instructions to build it into the design of the building that would host MIT's Computer Science and Artificial Intelligence Lab. It held 50 items contributed by the likes of Bill Gates and Tim Berners-Lee, and was not to be opened for 35 years — unless someone cracked the cryptographic puzzle woven into its design. The puzzle's designer was Ron Rivest, the R in RSA, and he says it was not meant to be hard so much as slow: built to take almost exactly 35 years to compute.

The lock is a chain of nearly 80 trillion squaring operations. Because every step builds on the previous one, parallel computing is useless — throwing more machines at it cannot shorten the timeline, which is why Moore's law, not hardware budgets, sets the pace. Rivest estimated the answer should take about 35 years from 1999.

Bernard Fabrot, a self-taught Belgian programmer, stumbled on the puzzle in 2015. He noticed Rivest's Java reference code could be beaten by using the free GNU Multiple Precision Arithmetic Library, dedicated one CPU core of his home desktop, and ran it around the clock through vacations and power cuts. He told almost no one: if word got out, someone with a bigger CPU could overtake him. Three and a half years later he had completed the squarings and derived the solution, announcing it to MIT on April 15, 2019 — almost 20 years to the day after Rivest published the puzzle.

He did not know he was in a race. A group called Cryptophage, led by former Intel engineer Simon Peffers and researching verifiable delay functions for blockchains, had run an FPGA implementation of Erdinc Ozturk's algorithm since mid-March — about 10 times faster than a high-end CPU — and calculated it would finish on the evening of May 10. MIT told them Fabrot had beaten them. Rivest admitted he had overestimated his puzzle's difficulty: he never anticipated FPGA chips. CSAIL director Daniela Rus, it turned out, had not even known the puzzle existed.

Chain the squarings so each step needs the last: more computers can't buy speed, only elapsed time can.

Sequential dependency is the whole design: each squaring feeds the next, so buying more machines does not shorten the timeline.

The lock calibrates itself to Moore's law — Rivest sized the workload in 1999 so only decades of steady chip improvement could crack it.

It needs no guarding or secrecy: the instructions are public, and the answer is simply unreachable before its time.

Free software decided the race — a faster arithmetic library on one desktop core beat specialized FPGA hardware built for the purpose.

Some problems can be priced in time, not compute: sequential dependency is a resource money cannot parallelize away. A public answer that stays unreachable for decades needs no guards and no secrecy.

Fabrot handed MIT his solution on April 15, 2019, beating the Cryptophage FPGA team by weeks. Rivest conceded he had overestimated the difficulty because he did not foresee how capable FPGAs would become. The congratulatory phrase inside is to be revealed at the capsule's opening on May 15, 2019. The original instructions said to send the answer to the director of the Laboratory for Computer Science — a lab that no longer exists, having merged into CSAIL in 2003, whose director was unaware of the puzzle until Fabrot showed up.