Top all-time · #421
Zvi Mowshowitz
Personal blog · thezvi.substack.comaipolitics
Zvi Mowshowitz writes Don't Worry About the Vase, a personal blog at thezvi.substack.com about ai and politics. Since 2021, 132 different Hacker News readers have linked it 581 times, most often “The online sports gambling experiment”. It is still writing: the last post was on 2026-10-30, and you can follow it by feed.
Sometimes recommended · 7 of 40 sampled comments recommend it?
Why HN links it
HN readers link it for weekly AI news roundups and in-depth analysis of AI safety, policy and model releases, and for posts on SB-1047, sports betting and nuclear risk.
- Recommended for · 7 comments
- Called well-sourced and in-depth, not too technical or vibes-based; several posts called good, reasonable or worth reading.
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
A world made of gears. Doing both speed premium short term updates and long term world model building. Currently focused on weekly AI updates. Explorations include AI, policy, rationality, medicine and fertility, education and games. Click to read Don't Worry About the Vase, by Zvi Mowshowitz, a Sub
Most-linked posts?
- The online sports gambling experiment6 people · 8 mentions · first linked 2024-11-11
- More on an Internal OpenAI Model Hacking into HuggingFace6 people · 6 mentions · first linked 2026-07-26
- /6 people · 6 mentions · first linked 2023-11-12
- The Three AI Pills6 people · 6 mentions · first linked 2026-08-05
- OpenAI Trained Models While They Were Coordinating Exploits via Message Boards5 people · 6 mentions · first linked 2026-08-07
- On Car Seats as Contraception: Or: Against Car Seat Laws at Least Beyond Age 25 people · 5 mentions · first linked 2022-08-22
- Jailbreaking ChatGPT on Release Day4 people · 6 mentions · first linked 2022-12-03
- Escape Velocity from Bullshit Jobs4 people · 4 mentions · first linked 2023-01-10
- The Most Forbidden Technique4 people · 4 mentions · first linked 2025-03-13
- On the Proposed California SB 10473 people · 3 mentions · first linked 2024-02-12
What Hacker News says?
My take from what I have been able to ascertain regarding these events is that underestimating models' ability to cooperate, propensity to discover new utility through basic systems, and unreservedly participate in collective actions of deception, coercion, or extortion is a very risky decision. The precautionary principle seems appropriate here, don't test them against weaker containment techniques in order to find…
Comment linking /p/openai-trained-its-models-for-months
on “Discovery of a new OpenAI agent message board”Submitted Convoy Crackdown – power to freeze bank accounts without trial or legal recourse as a story · 416 points
HN discussion>The huggingface incident isn't widely reported and digested yet ... It is pretty widely reported, and is being digested in an ongoing manner as more details become public. One entry point into the scenery from a month ago can be found at : https://thezvi.substack.com/p/openai-trained-its-models-for-... There has also been reporting at CNN: https://edition.cnn.com/2026/08/24/tech/openai-subpoena-hugg... and by NBC:…
Comment linking /p/openai-trained-its-models-for-months
on “Navier-Stokes – Tristan Buckmaster [pdf]”Submitted Claude 3.5 Sonnet as a story · 301 points
HN discussionApparently the censorship isn't baked-in to the model itself, but rather is overlayed in the public chat interface. If you run it yourself, it is significantly less censored [0] [0] https://thezvi.substack.com/p/on-deepseeks-r1?open=false#%C2...
Comment linking /p/on-deepseeks-r1?open=false
on “DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL”Submitted OpenAI Moves to Complete Potentially the Largest Theft in Human History as a story · 249 points
HN discussionI'm going to take your comment as a request for supporting details. I believe what I wrote is an accurate summary of information in a report published this month by Anthropic. I haven't seen the report, but have seen this next summary written by someone who I trust to summarize accurately: https://thezvi.substack.com/p/the-bad-guy-with-an-ai-named-c... OpenAI reported similar behavior last year. The assertion that t…
Comment linking /p/the-bad-guy-with-an-ai-named-claude?open=false
on “Microsoft exec called AI scraping 'the largest theft of labor in human history'”Submitted O3 Turns Pro as a story · 164 points
HN discussionhttps://thezvi.substack.com/p/newsom-vetoes-sb-1047 This link convinced me otherwise. I wasn't too sure before (I'm a little ashamed to admit I hadn't been following the development of SB 1047). I will also specifically share this talk on YouTube: https://www.youtube.com/watch?v=xoVJKj8lcNQ I think this passing this bill would have probably been a very good idea. Not perfect, but far, very far from broken. Edit: I'm…
Comment linking /p/newsom-vetoes-sb-1047
on “Gavin Newsom vetoes SB 1047”Submitted The online sports gambling experiment as a story · 124 points
HN discussionYou believe in AGI, but not ASI then? https://thezvi.substack.com/p/the-three-ai-pills That's fine, but it would be useful to explicitly say that is the disagreement, rather than just claim "hubris". Otherwise conversations are just looping. Yes you're right there are plenty of offline things more dangerous than just ordinary AI, and arguably than AGI (not so sure). There definitely aren't, pretty well by definition…
Comment linking /p/the-three-ai-pills
on “I resigned from Anthropic today”Submitted Altman's Ambition as a story · 111 points
HN discussion>promoting this as a good thing This is literally completely opposite of what happened. Then entire point is that this is bad, unwanted, behavior. Additionally, it has already been demonstrated that every other frontier model can be made to behave the same way given the correct prompting. I recommend the following article for an in depth discussion [0] [0] https://thezvi.substack.com/p/claude-4-you-safety-and-alignm…
Comment linking /p/claude-4-you-safety-and-alignment
on “Claude 4 System Card”Submitted Repeal the Foreign Dredge Act of 1906 as a story · 106 points
HN discussionI like gambling too, I was very much for legalizing it until it happened and I saw how many lives it's devastated, and how vulnerable young people are to it. Now I don't give a fuck that banning it would deprive me (or you) of something we happen to enjoy. Here‘s the article that started me toward changing my mind: https://thezvi.substack.com/p/the-online-sports-gambling-exp...
Comment linking /p/the-online-sports-gambling-experiment
on “Americans increasingly see legal sports betting as a bad thing for society”Submitted AIs Will Increasingly Fake Alignment as a story · 104 points
HN discussionThe Most Forbidden Technique by Zvi Mowshowitz, Mar 12, 2025: https://thezvi.substack.com/p/the-most-forbidden-technique
Comment linking /p/the-most-forbidden-technique
on “Hybrid Reasoning Agents: LLM agents that think in latent space, act through text”Submitted Spending too much time at airports as a story · 89 points
HN discussionI think you're being a little pessimistic. See these comments on a recent US senate hearing: >Not every senator asked good questions, but most of them did. All of them very clearly already knew plenty of details about the Hugging Face incident and multiple other incidents. Most of them had a clear understanding of terms like "misalignment", "recursive self-improvement", "chain of thought / chain of thought monitorin…
Comment linking /p/the-ai-preference-cascade-reaches
on “I quit OpenAI because its culture is broken”Submitted AI Companion Piece as a story · 83 points
HN discussionI cannot tell what negative-sum outcomes you consider possible. Do you believe AI can drive humans extinct? How many of Zvi Mowshowitz's Three AI Pills would you say you've taken? https://thezvi.substack.com/p/the-three-ai-pills
Comment linking /p/the-three-ai-pills
on “An Alien Mind”Submitted Omicron Variant: Early Analysis as a story · 59 points
HN discussionZvi Moshowitz's blog [0] is IME a pretty good place to keep track of the state of things, it's well-sourced and in-depth without being either too technical or too vibes-based. Generally every time a model is declared the new best you can count on him to have a detailed post examining the claim within a couple days. [0]: https://thezvi.substack.com/
Comment linking /
on “Gemini 2.5 Pro vs. Claude 3.7 Sonnet: Coding Comparison”Submitted On the Proposed California SB 1047 as a story · 49 points
HN discussionI’m new to this concept so may have missed something, but the post [0] seems to be about CoT specifically. In CoT you have an intermediary step that helps the model get better final results; the lesson is that if you try to improve the intermediary steps directly using training data then the model will optimize for better steps but not for better final results. I don’t think this is the same situation. 1. Anthropic…
Comment linking /p/the-most-forbidden-technique/
on “Persona vectors: Monitoring and controlling character traits in language models”Submitted Analyzing a Critique of the AI 2027 Timeline Forecasts as a story · 38 points
HN discussionFor ref https://thezvi.substack.com/p/the-most-forbidden-technique/
Comment linking /p/the-most-forbidden-technique/
on “Persona vectors: Monitoring and controlling character traits in language models”Submitted DeepSeek Panic at the App Store as a story · 36 points
HN discussionThere's a great deal of coverage about DeepSeek on Zvi's newsletter here: https://thezvi.substack.com/archive?sort=new - the first post on R1 is from January 22
Comment linking /archive?sort=new
on “TL;DR of Deep Dive into LLMs Like ChatGPT by Andrej Karpathy”Submitted Formula for a Shortage as a story · 34 points
HN discussion
