Hacker News: yusufozkan

ARC-AGI-3 Leaderboard

yusufozkan — Wed, 25 Mar 2026 17:30:57 +0000

Article URL: https://arcprize.org/leaderboard

Comments URL: https://news.ycombinator.com/item?id=47520530

Points: 5

# Comments: 0

Harness design for long-running application development

yusufozkan — Tue, 24 Mar 2026 16:40:30 +0000

Article URL: https://www.anthropic.com/engineering/harness-design-long-running-apps

Comments URL: https://news.ycombinator.com/item?id=47505430

Points: 5

# Comments: 1

A Ramsey-Style Problem on Hypergraphs

yusufozkan — Mon, 23 Mar 2026 17:24:13 +0000

Article URL: https://epoch.ai/frontiermath/open-problems/ramsey-hypergraphs

Comments URL: https://news.ycombinator.com/item?id=47492454

Points: 3

# Comments: 0

New comment by yusufozkan in "Our Agreement with the Department of War"

yusufozkan — Sat, 28 Feb 2026 20:50:28 +0000

This is the same company that started as a nonprofit dedicated to open AI safety research, then became a capped-profit entity, then effectively closed-source, then dropped the cap, and is now pursuing full for-profit conversion. Every single guardrail they've set for themselves has been quietly revised or removed once it became inconvenient. Anyone want to bet on how long those exclusions last?

Firstproof_oai [pdf]

yusufozkan — Sat, 14 Feb 2026 10:37:17 +0000

Article URL: https://cdn.openai.com/pdf/a430f16e-08c6-49c7-9ed0-ce5368b71d3c/1stproof_oai.pdf

Comments URL: https://news.ycombinator.com/item?id=47013427

Points: 2

# Comments: 0

1X World Model – From Video to Action: A New Way Robots Learn

yusufozkan — Mon, 12 Jan 2026 16:42:24 +0000

Article URL: https://www.1x.tech/discover/world-model-self-learning

Comments URL: https://news.ycombinator.com/item?id=46590896

Points: 3

# Comments: 0

New comment by yusufozkan in "The Llama 4 herd"

yusufozkan — Sat, 05 Apr 2025 19:14:25 +0000

> while pre-training our Llama 4 Behemoth model using FP8 and 32K GPUs

I thought they used a lot more GPUs to train frontier models (e.g. xAi training on 100k). Can someone explain why they are using so few?

New comment by yusufozkan in "4o Image Generation"

yusufozkan — Tue, 25 Mar 2025 20:19:16 +0000

This is another cool example from their blog

https://imgur.com/a/Svfuuf5

New comment by yusufozkan in "4o Image Generation"

yusufozkan — Tue, 25 Mar 2025 20:17:46 +0000

Generating an image of a completely full glass of wine has been one of the popular limitations of image generators, the reason being neural networks struggling to generalise outside of their training data (there are almost no pictures on the internet of a glass "full" of wine). It seems they implemented some reasoning over images to overcome that.

New comment by yusufozkan in "4o Image Generation"

yusufozkan — Tue, 25 Mar 2025 20:10:38 +0000

Are you sure you are using the new 4o image generation?

https://imgur.com/a/wGkBa0v