Top all-time · #20057
John Wang
Personal blog · johnjwang.com/postaistartups
John Wang's personal blog at johnjwang.com covers ai and startups. Since 2026, 3 different Hacker News readers have linked it 12 times, most often “How Claude's watermarking (probably) works”. It is still writing: the last post was on 2026-09-25, 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 this post when discussing how Claude's text watermarking works. One commenter summarises it as breaking token generations where probabilities are nearly tied, which is a statistical method rather than visible markers.
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.
Personal website of John Wang - previously co-founder / CTO of Assembled and an early engineer at Stripe. Writing about AI, engineering, and startups.
Most-linked posts?
- How Claude's watermarking (probably) works3 people · 5 mentions · first linked 2026-08-12
What Hacker News says?
It's done with a lot more subtlety and embedded directly into the content, no Unicode shenanigans. The basic idea is just to break token generations where the probability is nearly tied in favor of the side that matches the secret key. With a long enough text block you can be statistically certain if the generation was using the key. From another comment: https://johnjwang.com/post/2026/08/12/how-claude-watermarkin.…
Comment linking /post/2026/08/12/how-claude-watermarking-probably-works/
on “Text AI watermarks will always be trivial to remove”Submitted The engineering behind the US Strategic Petroleum Reserve as a story · 281 points
HN discussionThere exist methods to detect what kind of watermarking tool that someone is using, and most of the big tools have specific signatures that you can look for. For Anthropic, it’s highly likely that the watermark is a SynthID type mark similar to the one that Sean is talking about (I actually ran the analysis here https://johnjwang.com/post/2026/08/12/how-claude-watermarkin... ). When we get confirmation of whether al…
Comment linking /post/2026/08/12/how-claude-watermarking-probably-works/
on “Text AI watermarks will always be trivial to remove”Submitted Why are executives enamored with AI, but ICs aren't? as a story · 39 points
HN discussionI ran some tests from the watermarking literature: https://johnjwang.com/post/2026/08/12/how-claude-watermarkin... Short version: it's likely a SynthID style statistical watermarking scheme that Anthropic is using. There's likely no hidden unicode and no large stable green/red scheme. Longer version: I found some really statistically significant outcomes when using Gloaguen et al.’s ( https://www.sri.inf.ethz.ch/blo…
Comment linking /post/2026/08/12/how-claude-watermarking-probably-works/
on “How Claude marks AI-generated content”Submitted Megafauna Extinction and Startups as a story · 3 points
HN discussionSubmitted What the megafauna extinction can teach us about AI startups and incumbents as a story · 2 points
HN discussionSubmitted Learnings from OpenAI's open source Codex Repo as a story · 2 points
HN discussionSubmitted Cheap software won't make engineering cheap as a story · 2 points
HN discussionSubmitted How Claude's watermarking (probably) works as a story · 1 points
HN discussionSubmitted How Claude's watermarking (probably) works as a story · 1 points
HN discussionSubmitted Number of tokens shouldn't be the only metric as a story · 1 points
HN discussion