Top all-time · #16173
Redwood Research blog
Company blog · redwoodresearch.substack.comaiscience
Redwood Research blog is the company blog at redwoodresearch.substack.com, writing about ai and science. Since 2024, 4 different Hacker News readers have linked it 4 times, most often “Getting 50% (SoTA) on Arc-AGI with GPT-4o”. It is still writing: the last post was on 2026-10-27, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
Redwood Research's Substack posts cover AI research, including a benchmark result (50% state-of-the-art) and AI safety concerns about models deceiving humans via chain-of-thought. Comments only question or cite the work.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
We research catastrophic AI risks and techniques that could be used to mitigate them.
Most-linked posts?
- Getting 50% (SoTA) on Arc-AGI with GPT-4o2 people · 2 mentions · first linked 2024-06-17
What Hacker News says?
Have you tried using more 4000x more samples? https://redwoodresearch.substack.com/p/getting-50-sota-on-ar...
Comment linking /p/getting-50-sota-on-arc-agi-with-gpt
on “Testing Generative AI for Circuit Board Design”Submitted Getting 50% (SoTA) on Arc-AGI with GPT-4o as a story · 394 points
HN discussionIt's really very simple. As models become more capable they may become interested in deceiving humans or otherwise manipulating them to achieve their goals. We already see this in various places see: https://www.anthropic.com/research/agentic-misalignment https://arxiv.org/abs/2412.14093 If the chain of thought of models becomes pure "neuralese" i.e. the models think purely in latent space then we will lose the abil…
Comment linking /p/notes-on-countermeasures-for-exploration
on “Is chain-of-thought AI reasoning a mirage?”Submitted How will we update about scheming? as a story · 1 points
HN discussion