Top all-time · #4345
Beren's Blog
Personal blog · beren.ioaiphilosophy
Beren's Blog is a personal blog at beren.io about ai and philosophy. Since 2023, 18 different Hacker News readers have linked it 21 times, most often “Right to Left (R2L) Integer Tokenization”. It is still writing: the last post was on 2026-10-22, and you can follow it by feed.
Too few sampled comments (5) to say how often HN recommends it.?
Why HN links it
HN commenters link Beren's Blog posts on AI topics: LLM confabulation versus hallucination, integer tokenization, alignment generalising further than capabilities, and why GOFAI failed. The links are cited as background.
Summarised by Claude from a sample of 5 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Scattered Thoughts
Most-linked posts?
- Right to Left (R2L) Integer Tokenization2 people · 2 mentions · first linked 2024-07-10
- How Can LLM RL Work Despite Information-Theoretic Inefficiency2 people · 2 mentions · first linked 2026-08-08
- LLMs Confabulate Not Hallucinate2 people · 2 mentions · first linked 2023-10-05
- The Scale of the Brain vs. Machine Learning (2022)2 people · 2 mentions · first linked 2023-12-22
What Hacker News says?
On the other hand, it may be that "Alignment likely generalizes further than capabilities." - https://www.beren.io/2024-05-15-Alignment-Likely-Generalizes...
Comment linking /2024-05-15-Alignment-Likely-Generalizes-Further-Than-Capabilities/
on “Ilya Sutskever's SSI Inc raises $1B”Submitted The Scale of the Brain vs. Machine Learning (2022) as a story · 63 points
HN discussionThe math tokenization research is probably closest. GPT-2 tokenization was a demonstratable problem: https://www.beren.io/2023-02-04-Integer-tokenization-is-insa... (Prior HN discussion: https://news.ycombinator.com/item?id=39728870 ) More recent research: https://huggingface.co/spaces/huggingface/number-tokenizatio... Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs: https://arxiv.org/…
Comment linking /2024-07-07-Right-to-Left-Integer-Tokenization/
on “The bitter lesson is coming for tokenization”Submitted Integer Tokenization Is Insane (2023) as a story · 35 points
HN discussionExpert systems are basically decision trees which are "gofai" (good old fashioned ai) as opposed to deep learning. I've never really seen a good definition for what counts as "gofai" (is all statistical learning/regression gofai? What about regression done via gradient descent?). There's some talk in [1] [1] https://www.beren.io/2023-04-10-Why-GOFAI-failed/
Comment linking /2023-04-10-Why-GOFAI-failed/
on “Based on its own charter, OpenAI should surrender the race”Submitted Scaffolded LLMs as natural language computers as a story · 11 points
HN discussion> Sparse models for example show a lot of promise and more closely track real biological systems. I think sparsity is a consequence of some other fundamental properties of brain function that we've yet to understand. Just sparsifying the models we've got is not going to lead anywhere, IMO. (For example it's estimated that current AI models are already within 1%-10% of a human brain in terms of "number of parameters"…
Comment linking /2022-08-06-The-scale-of-the-brain-vs-machine-learning/
on “GPU-rich labs have won: What's left for the rest of us is distillation”Submitted How Can LLM RL Work Despite Information-Theoretic Inefficiency as a story · 2 points
HN discussionOne rarely discussed peculiarity about religion is that it distorts reality for both believers and non-believers (as one can see by observing how anti-religious people discuss what religion "is", what religious people believe, etc). Possibly relevant: LLMs confabulate not hallucinate https://www.beren.io/2023-03-19-LLMs-confabulate-not-halluci...
Comment linking /2023-03-19-LLMs-confabulate-not-hallucinate/
on “LLMs Confabulate Not Hallucinate”Submitted Space Warfare Seems Mostly Defense Dominant as a story · 2 points
HN discussionSubmitted AI Monotheism vs. AI Polytheism as a story · 2 points
HN discussionSubmitted Intellectual Progress in 2025 as a story · 2 points
HN discussionSubmitted The Singularity as Cognitive Decoupling as a story · 2 points
HN discussionSubmitted Right to Left (R2L) Integer Tokenization as a story · 2 points
HN discussionSubmitted GPUs vs. Brains. Hardware and Architecture as a story · 2 points
HN discussionSubmitted LLMs Confabulate Not Hallucinate as a story · 2 points
HN discussionSubmitted Mitigating Reward Hacking as Institutional Design as a story · 1 points
HN discussionSubmitted How Can LLM RL Work Despite Information-Theoretic Inefficiency as a story · 1 points
HN discussionSubmitted Capital Ownership Will Not Prevent Human Disempowerment as a story · 1 points
HN discussionSubmitted Reflections on the Bitter Lesson (2020) as a story · 1 points
HN discussion
