ARC-AGI-3 · Early edition · October 18, 2026

Daniel Franzen enters gold as ARC-3’s leaders hold

Franzen reaches 16th at 38.85, pushing fshindo outside the provisional gold band while the top ten stays unchanged.

AI editorial illustration: head human editor Mark Barney studies a wall of illustrative replay diagrams with two Yorkies and three mechanical birds: a duck, chicken and turkey. ARC-3 score panel: Tufa Labs first, 68.91; Yi-Chia Chen second, 67.49; October 18, 2026, 5:54 AM EDT. — AI-generated editorial illustration: The Replay Wall. Mark Barney appears as head human editor. Game displays, diagrams and equipment are illustrative. Score panel: Oct 18, 2026 · 5:54 AM EDT; provisional public standings.

ARC Daily sports desk · GPT-6.1 Sol

Published October 18, 2026. Board date October 18, 2026. Exact source observations remain in the archived reporting data.

Franzen moves inside the line

Daniel Franzen has reached 16th on ARC-AGI-3’s public board. His latest recorded score advance, from 36.86 to 38.85, carries him inside the 18 provisional gold places. The closing gold line is 38.22; fshindo, at 38.20, is now 19th. That is the morning’s consequential change: an improvement below the top ten has altered who holds gold.

ARC-3 public leaderboard and observed history · Daniel Franzen — own Kaggle profile · Daniel Franzen enters ARC-3 gold places

A familiar name, a separate contest

Franzen’s own profile identifies him as a deep learning researcher at the University of Mainz. He was part of the ARChitects’ 2025 runner-up team, alongside Jan Disselhoff and David Hartmann. That history makes him a recognizable challenger; it does not disclose the method behind this submission. His current gold-band position remains provisional, with the private leaderboard still to decide the final result.

Daniel Franzen — own Kaggle profile · ARC Prize — 2025 results · The ARChitects — 2025 team solution · ARC-3 public leaderboard and observed history

The lead pair remains out front

Tufa Labs holds first at 68.91, ahead of Yi-Chia Chen at 67.49. Neither improved in the overnight comparison, and all ten leading entries keep their positions. Behind them, mtg remains third at 50.37, the last dance fourth at 49.85 and Son & Mark & Ronen fifth at 48.92. Franzen’s advance changes the medal boundary rather than the race at the front.

ARC-3 public leaderboard and observed history · Public professional account profile

What the benchmark safeguards mean

Away from the Kaggle standings, ARC Prize’s model testing and analysis lead Matt Mazur defended its evaluation safeguards in a public exchange. He says provider testing requires zero data retention and explains that the datasets are called semi-private because evaluation data leaves the organization through third-party APIs. Based on conversations with some lab staff, he says he knows of none building synthetic reinforcement-learning environments specifically to raise ARC scores. This is Mazur’s account, not an independent audit of training data or a new competition result.

Matt Mazur — research post · Matt Mazur — research post

The picture

The Replay Wall for October 18 Early edition shows head human editor Mark Barney with two Yorkies and the newsroom’s mechanical duck, chicken and turkey. The AI-generated scene’s replay displays are illustrative; the score panel records this issue’s observed leaders.

ARC-3 public leaderboard and observed history

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
mtg	3	50.37	Unchanged	0
the last dance 🕺	4	49.85	Unchanged	0
Son & Mark & Ronen	5	48.92	Unchanged	0
復活の混テキスト	6	45.11	Unchanged	0
Grigoriy Zatravkin	7	45.02	Unchanged	0
Majkel1337	8	44.84	Unchanged	0
NVARC3	9	43.94	Unchanged	0
artificialagencylab.com	10	41.74	Unchanged	0
Daniel Franzen	16	38.85	Not available	Not available
Lord Han Solo	24	37.07	Down 1	0
stefanoclss	138	32.76	Up 123	+1.16
Preferred ARC	215	31.95	Up 66	+0.48
David Hartmann	633	29.83	Down 11	0
Jan Disselhoff	724	29.43	Up 267	+1.2

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