Top all-time · #14308
Vlad Feinberg
Personal blog · vladfeinberg.com · on HN as vladfaiprogrammingdata
Vlad Feinberg's personal blog at vladfeinberg.com covers ai, programming and data. Since 2019, 5 different Hacker News readers have linked it 22 times, most often “Gemini Pretraining [slides] [pdf]”. It is still writing: the last post was on 2026-05-10, and you can follow it by feed.
Too few sampled comments (1) to say how often HN recommends it.?
Why HN links it
HN readers link the author's slides from a 2025 Princeton talk, shared in a thread on Gemini Flash pretraining. The comment only gives PDF and X links, with no remarks on the writing.
Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Blog
Most-linked posts?
- Gemini Pretraining [slides] [pdf]2 people · 2 mentions · first linked 2025-04-27
- How to Land a Frontier Lab Job2 people · 2 mentions · first linked 2026-05-18
What Hacker News says?
https://vladfeinberg.com/ I mostly post about stats or programming topics. I only really try to put something up if it's a particularly hot take on a useful topic, like * a reduction from causal to statistical inference https://vladfeinberg.com/2019/12/01/metaphysics-of-causality... * an exploration of what the best way to listen to experts is https://vladfeinberg.com/2020/01/05/stop-anytime-multiplicat...
Comment linking /2020/01/05/stop-anytime-multiplicative-weights.html
on “Ask HN: What is your blog and why should I read it?”Submitted Vectorizing Ragged Arrays as a story · 28 points
HN discussion"Ridiculous" is not a word I'd use... sometimes symbolic derivatives are intractable, in which case complex step derivatives provide high-precision numerical derivatives which are otherwise unattainable. https://vladfeinberg.com/2020/07/05/discrete-residues.html
Comment linking /2020/07/05/discrete-residues.html
on “Differentiable programming from scratch”Submitted Beating Spark with a Better Map Reduce as a story · 24 points
HN discussionI think the problem with observing interventions is it will result in impractically large sample complexity to derive the same causal conclusion as causal calculus armed with assumptions of DAG structure. I think metaphysically speaking the approaches (observing interventions vs causal calc) aren’t meaningfully different in terms of inferences you can make with infinite data, see my similar observations to yours: ht…
Comment linking /2019/12/01/metaphysics-of-causality.html
on “Why machine learning struggles with causality”Submitted Complex Hash Collisions as a story · 10 points
HN discussionPDFs: https://vladfeinberg.com/assets/2025-04-24-princeton-talk.pd... op: https://x.com/FeinbergVlad/status/1915847967313154400
Comment linking /assets/2025-04-24-princeton-talk.pdf
on “Gemini Flash Pretraining”Submitted Solving the General RCT2 Maze Problem with Sympy as a story · 6 points
HN discussionI was hoping for this take—-of course LLMs have an internal model, and we can externally verify how accurate it is statistically. And the more you train it the better it gets. And for discrete spaces you might even get a perfect model eventually with nonzero probability. Since you mentioned different layers, one question to ask is if your poetry example is “just a combination of existing rare patterns,” one that’d b…
Comment linking /2019/12/01/metaphysics-of-causality.html
on “Do Large Language Models learn world models or just surface statistics?”Submitted Gemini Flash Pretraining as a story · 3 points
HN discussionI ended up needing this so often for graph processing, and for values which might be inexact if using floating point, that I saved the formula in a blog post. https://vladfeinberg.com/2020/03/07/subset-isomorphism.html The formula can be "oblivious" to the final size of the matrix too, which is helpful if you're doing some sparse ML training on edges (e.g., GNNs).
Comment linking /2020/03/07/subset-isomorphism.html
on “What scientists must know about hardware to write fast code”Submitted Distillation with Linear Models as a story · 3 points
HN discussionSubmitted How to Land a Frontier Lab Job as a story · 2 points
HN discussionSubmitted Gemini Pretraining [slides] [pdf] as a story · 2 points
HN discussionSubmitted Vectorizing Raged Arrays as a story · 2 points
HN discussionSubmitted Weekly Financial Summaries with Postgres as a story · 2 points
HN discussionSubmitted How to land a Frontier lab job as a story · 1 points
HN discussionSubmitted Crinkle Crankle Optimization as a story · 1 points
HN discussionSubmitted Stop Anytime Multiplicative Weights as a story · 1 points
HN discussionSubmitted The Semaphore Barrier as a story · 1 points
HN discussion
