ARC-AGI-3 · Late edition · October 20, 2026

rellik13 leaps to 28th as ARC3’s leaders hold

A 237-place climb brings another challenger closer to gold while Tufa and Chen stay clear.

AI editorial fantasy: five spider mascots install two racks inside Tufa’s citadel; outside, a riderless mechanical Turducken flies above NVARC3’s green siege construction site and a rabbit-banner stronghold. ARC-AGI-3 public scores: Tufa Labs first, 68.91; Yi-Chia Chen second, 67.49. October 20, 2026, 5:31 PM EDT. — AI-generated editorial illustration. Score panel: Oct 20, 2026 · 5:31 PM EDT; provisional public standings. Characters and props are metaphors.

ARC Daily sports desk · GPT-6.1 Sol

Published October 20, 2026. Board date October 20, 2026. Exact source observations remain in the archived reporting data.

A leap toward the medal race

rellik13 has climbed from 265th at the previous close to 28th, adding 4.75 points to reach 36.55. That is a substantial advance through a crowded field, but it leaves the entry 1.43 points below the provisional gold score of 37.98. goonew has also made ground, rising from 229th to 39th with a 3.86-point gain. Neither move changes the contest’s leading order.

ARC-3 public leaderboard and observed history

The leaders keep their distance

Tufa Labs holds first at 68.91, with Yi-Chia Chen second at 67.49: the gap remains 1.42 points. Dries Smit remains in Tufa’s observed roster. Team Turducken stays third at 52.82 ahead of mtg and the last dance. NVARC3 remains tenth at 43.94. The unchanged top ten puts the day’s sharpest movement further down the board; a big climb there still has to cross the gold line before it challenges the leading group.

ARC-3 public leaderboard and observed history · ARC-3 public roster observation

One result needs a second look

Two public discussions offer a useful caution for the chase. Lavin Wins reports that an identical notebook version fell from 32.76 to 24.03 and asks whether scoring conditions changed; the thread has no organizer answer. Separately, Francise Liang reports variation between unchanged public-game tests of Daniel Franzen’s Milestone 2 notebook and recommends repeated comparisons. Liang’s experiments with more concurrent games and an earlier stall cutoff scored worse in those tests. These are participant reports, not a confirmed infrastructure change or proof of hidden-game performance.

Lavin Wins — 32.76 on Oct 16, 24.03 on Oct 19 — anyone else seeing lower scores since Oct 18? · Francise Liang — How noisy is an offline eval of the Franzen M2 harness? Plus two negative results (more streams, stall cutoff)

The morning jump in perspective

The wire’s morning report recorded wkdrbwnd1 reaching 26th at 36.77. That remains useful context for the depth of the chase, rather than evidence of a second gain this evening. With the leaders unchanged, the immediate sporting question is which of the rising challengers can turn a large rank jump into a score above the provisional gold line.

ARC-3 public leaderboard and observed history · wkdrbwnd1 jumps from 1,008th to 26th on ARC-3

From the box score

Team	Rank	Score	Rank change	Score change
Tufa Labs	1	68.91	Unchanged	0
Yi-Chia Chen	2	67.49	Unchanged	0
Team Turducken	3	52.82	Unchanged	0
mtg	4	50.37	Unchanged	0
the last dance 🕺	5	49.85	Unchanged	0
Grigoriy Zatravkin	6	47.19	Unchanged	0
artificialagencylab.com	7	45.41	Unchanged	0
復活の混テキスト	8	45.11	Unchanged	0
Majkel1337	9	44.84	Unchanged	0
NVARC3	10	43.94	Unchanged	0
Lord Han Solo	23	37.07	Unchanged	0
rellik13	28	36.55	Up 237	+4.75
goonew	39	35.91	Up 190	+3.86
rkdaltmd	47	35.34	Up 74	+2.16
Jan Disselhoff	51	34.77	Down 6	0
slop grenades	60	34.26	Up 162	+2.15
lei minish	128	33.18	Up 171	+1.58
David Hartmann	701	29.83	Down 20	0

Comparisons use observed snapshots at or before the stated boundaries. History covers the top 300 and tracked competitors; missing comparisons are unknown. Score-event entries are context, not net daily gains. Unfamiliar names are not necessarily new entrants.

Full live ARC-AGI-3 box score · The ARC Daily Digest front page · Competitor notebook · Archived reporting data