The Log shows what shipped. This page shows everything: every attempt we've made on a numbered Erdős problem, every reasoning seat we ran, every round, every dead end. The 35 internal attempts that went nowhere are here on purpose — "we tried and it didn't work" is data too.
Every published entry, in problem-number order — each carries its full rounds, receipts, and retractions on its own page. This section maintains itself.
Reasoner briefs and Codex waves that produced nothing publishable. PARTIAL = a real sub-result, not a close · NO RESULT = tried hard, got nothing · STILL OPEN = probed, learned, own nothing. Sorted by problem number.
No counterexample found — search exhausted its budget. Problem open.
Early batch run — no verdict recovered (run JSON only).
Early batch run — no verdict recovered (run JSON only).
May run had no verdict; a fresh density-bound brief is dispatched. Open ($500).
No counterexample — ratio behavior verified computationally. Open.
Early batch run — no verdict recovered (run JSON only).
Verifier built and checked; no verdict write-up was ever produced.
A reduction plus a conditional construction — no resolution. Open.
The one that beat every model at every budget in our cross-model bench. Best outcome: an honest decline.
Two rounds, decorrelated second opinions included — both unsolved.
No polynomial improvement in either direction. Open ($100).
The requested limiting ratio was never proved. Open.
Computational upper bounds for f(4), f(5) verified. General problem open.
New computational contribution landed; infinitude still open unconditionally.
Subset-sum checker built; the asymptotic threshold is unresolved.
Remote-witness obstruction unresolved; a blind reattempt corroborated the architecture, no summable bound.
Search produced no proof claim. Open.
Decidable up to the computed range, no counterexample — general case open.
Sieve machinery explored; the full problem appears untouched. Open.
Round 2: proofs for 3 and 4 primitive generators. General case open.
The order is already Θ(√N) — the refinement is the question. Open.
Open — a related variant was solved by others (Joret–Micek–Reed–Smid) while we worked it.
Two computational obstructions independently confirmed (affine K₂₅→K₂₆ nonexistent; non-rigidity).
Complete N=4, k=4 certificate — no integers m, n exist. Full problem stays open.
No counterexample, no general proof. Open.
Round 2 pins g(4) = 5. The asymptotic remains open.
Literature map plus several open cases — no resolution.
Counterexamples to eventual strictness found and checked. Unresolved.
No proof of the N^(1−o(1)) bound. Open.
Two rounds — still open.
No k=5 witness found — does not settle the question either way.
Strong partial — the universal statement remains open.
Found non-log-concave trees, but still unimodal — the target was never met across 11 search runs.
Open — a recent Eric Li preprint claims more; we treat it as an external candidate, not ours.
Strong partial, verified — not closed by us (Price's claimed proof is unauditable).
ChatGPT Pro (GPT‑5.6 Sol). The heavy reasoner — Patrick relays briefs by hand, one at a time.
Codex at max reasoning. A separate usage bucket, so it runs as parallel waves while Pro works.
Grok Heavy. The decorrelated second opinion — bake-offs and blind re-attempts.
Scouting, verification, shipping. Trusts nothing downstream of any seat's enthusiasm.
Nothing reaches the Log until a second, independent seat re-verifies the load-bearing claims — and claims die more often than they survive. That's the point of doing it in public.