Blog

Runs, results and what didn't work.

Dated Kaggle competition runs with every rejected experiment, and practical guides to evidence-first ML work from Claude Code, Cursor or the command line.

GuideNov 15, 2026 · 9 min read

Evidence from ARC-AGI-3, Enveda CASMI 2026, Squares in Squares and Gemma 4

Autoresearch for Kaggle: What 136 Overnight Rounds on Four Live Competitions Taught Us

Karpathy’s autoresearch loop on live Kaggle competitions: what had to change, a medal-range result, 136 recorded rounds and the lessons, failures included.

Live competitions 4Rounds on record 136Best standing Bronze rangeLaunched unattended 94
Read the full story
Progress reportNov 14, 2026 · 8 min read

ARC-AGI-3 · Enveda CASMI 2026 · Squares in Squares · Gemma 4 Developer Agent

Our Kaggle Standings, November 2026: One Bronze-Medal Range, Three Live Runs

Our standings on four Kaggle competitions, 14 Nov 2026: ARC-AGI-3 in the bronze-medal range, Enveda 0.008 from bronze, and what each run taught us.

Competitions 4In medal range ARC-AGI-3Closest live Enveda · 0.008Rounds on record 136
Read the full story
Competition runAug 11, 2026 · 9 min read

ARC Prize 2026 — ARC-AGI-3 · Kaggle · $850,000 prize pool

ARC Prize 2026 (ARC-AGI-3): A Bronze-Medal-Range Run and Every Rejected Idea

ARC Prize 2026 ARC-AGI-3 on Kaggle: 9-hour no-internet runs, closed November 2, 2026, results Dec 4. Our rank: 313 of 4,926, bronze range. Every idea shown.

Public rank (final) 313 / 4,926Medal standing Bronze rangeBest public score 32.51Experiments 14 · 10 rejected
Read the full story
Competition runNov 14, 2026 · 8 min read

Enveda CASMI 2026 · Kaggle · $50,000 · closes 14 Dec 2026

Enveda CASMI 2026 on Kaggle: From Rank 3,856 to 1,421, 0.008 From Bronze

Our Enveda CASMI 2026 run: 52 rounds, public score 0.136 to 0.479, rank 1,421 of 5,305 and 0.008 below the bronze-medal line. What worked, what failed.

Public rank (15 Nov) 1,421 / 5,305Medal standing 0.008 from bronzeBest public score 0.479Rounds 52 · 10 rejected
Read the full story
Competition runNov 14, 2026 · 7 min read

Squares in Squares · Kaggle Playground · $50,000 · closes 18 Jul 2027

Squares in Squares on Kaggle: 46 Rounds, 22.2502 to 22.1756, and a Moving Field

Our Squares in Squares run: 46 rounds took the public score from 22.2502 to 22.1756, rank 326 of 2,052. Which merges held and which of our ideas did not.

Public rank (14 Nov) 326 / 2,052Best public (lower wins) 22.1756Best local 22.1580Rounds 46 · 17 improved
Read the full story
Competition runNov 14, 2026 · 7 min read

Gemma 4 Developer Agent · Kaggle · $75,000 · closes 2 Dec 2026

Gemma 4 Developer Agent on Kaggle: Noise, No-Op Adapters and Failed Runs

Our Gemma 4 Developer Agent run: 0.17 public, rank 2,249 of 5,691, bronze line 0.20. An identical resubmission moved 0.05; 20 of 24 rounds failed on compute.

Public rank (14 Nov) 2,249 / 5,691Bronze line 0.20Best public score 0.17Rounds 24 · 20 failed
Read the full story
GuideNov 5, 2026 · 7 min read

Evidence from Enveda CASMI 2026 and ARC-AGI-3

Kaggle CV vs Leaderboard: What to Trust When They Disagree

Your Kaggle CV and public leaderboard disagree: which do you trust? A recorded case where the public score rose 0.136 to 0.166 while local MRR@25 fell.

Submissions traced 16October CV ↓, public ↑November Both ↑ · 0.213
Read the full story
GuideNov 5, 2026 · 6 min read

Evidence from Gemma 4 Developer Agent and ARC-AGI-3

Kaggle Leaderboard Noise: The Same Submission Scored 0.12, Then 0.17

We resubmitted a byte-identical Kaggle submission and the public score moved from 0.12 to 0.17. How to measure leaderboard noise before you trust a gain.

Identical file 0.12 → 0.17Offline repeats 2–4 levelsRule Measure the floor
Read the full story
ResearchAug 18, 2026 · 9 min read

Claude Code vs Cursor for AI Research Workflows: Where Each One Wins

A practical comparison of Claude Code and Cursor for research synthesis, coding execution, and evidence-based iteration.

Workflows 2 comparedBest for Research + editingOutcome Clear split
Read the full story

More posts

Video walkthroughs

Each run, explained on the Forger Labs HQ YouTube channel.

▶ YouTube