Planar unit distances: the final grid of OpenAI's argument
Best proposal, not yet published : value ≤ 30 (Oct 10, 2026, PR #3)
- upper bound 100by OpenAIgithub.com
- upper bound 36by rohanarungithub.com
- upper bound 32by rohanarunPR #2
- Last checked
Cite this frontier
DiscoverIA. Planar unit distances: the final grid of OpenAI's argument [Frontier, checked on Oct 13, 2026]. DiscoverIA registry. https://discoveria.ai/en/frontieres/distances-unite/
Only the upper bound is tracked. The upper bound has moved 2 times since 2026, most recently on Oct 10, 2026.
Number of selected edges in the final algebraic witness (complete grid of N left points by j anchors, N·j edges) of OpenAI's proof of a power saving for planar unit distances, general (rank-two) case; a complete grid with at least this many edges suffices for the conditional conclusion. Smaller = better reduction.
A conditional reduction of one step of the proof: it changes neither OpenAI's 4/3−δ exponent nor supplies an explicit power saving. The source's generic-grid, negligible-height and private-valuation lemmas remain hypotheses. Line cases: 6×4 = 24 edges (one locus is a line) and 4×3 = 12 (same line) are treated separately by the repository; one 8×4 grid covers every case. Points from unmerged PRs are proposals.
Conditional result : yes
Linked announcement in the registry
- Not evaluated describes the scope of the Lean formalization
Every step, as a table : Upper bound (4 points)
Smallest number of edges of the final witness established (OpenAI preprint, initial commit, then PRs of the tracked repository).
| Date | Exact value | Nature | Authors | Source |
|---|---|---|---|---|
| Oct 10, 2026 | 30 | proposed | rohanarun | PR #3 |
| Oct 10, 2026 | 32 Current record | merged | rohanarun | PR #2 |
| Oct 9, 2026 | 36 Current record | merged | rohanarun | github.com |
| Sep 23, 2026 | 100 Current record | published | OpenAI | github.com |
Other tracked problems (8)
Report an error or suggest a source
Thank you, your message was sent.
Sending failed. Try again, or write to [email protected].