Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Claim. For every nondecreasing with and eventually, the values
with are dense in for every . Taking
answers the second question as well. Plasma AI's
formalization
proves the two statements as Erdos420.erdos_420.parts.iii and
Erdos420.erdos_420.parts.ii.
Covers. The parts logarithmic_density and monotone_density of
Problem 420. The first question,
part divergence, asking whether for large
, is not covered.
Formalization. The version retained on
the source card
pins Lean and Mathlib v4.35.0-rc4. Its proof derives a uniform
power-saving prime-number theorem in arithmetic progressions from a vendored
Lean proof of the Dirichlet seven-eighths zero-free theorem, combines a
protected CRT covering with Maynard's sieve and a Selberg upper-sieve tail
bound, and concludes by sliding the factorial window. The zero-free theorem
is a proved dependency, not a hypothesis of the final statements.
formalization.yaml credits Jacob Parish as the formalization's author and
records assistance from Claude Code, model claude-opus-5-5.
Acceptance. Accepted on 2026-10-20 on the formalization. The independent
reproduction
rebuilt every Problem 420 module in a fresh source tree, using checked caches
for the vendored dependencies. Comparator accepted both targets. Separate
empty-environment kernel replays checked their complete proof closures, with
no unsafe or partial declarations and only propext, Classical.choice and
Quot.sound; both target types and all their statement dependencies matched
the challenge.
The independent statement-fidelity review found both statements faithful to the two density questions, and a distinct fresh-context grader gave PASS. The grader ruled the disclosed general-instruction exposure immaterial. This acceptance covers the two density questions and assigns no native claim tier.
Depends on. The formalized density theorem, whose dependency closure includes the Dirichlet seven-eighths zero-free theorem.