Not a chance. Far too funny, too well written, too terse while being densely packed with wit. I see zero signs of it being LLM-generated and lots of stuff LLMs have no way of doing.
If I am somehow wrong I would salivate at a chance to see the input.
And actually I see it clearly now, it has a bunch of signs I have called out multiple times myself. (It is entirely made out of lists of various types, and never states an opinion.)
Just my ego getting hold of me because I didn't realize it on my own.
You don't even need to read past the first timeline entry. The name "Marcus Chen" is literally a meme within AI creative writing circles due to how often Claude defaults to that exact name when naming fictional characters.
I never used Pangram before today, but since I've seen it mentioned many times on HN and I enjoyed reading the OP, I decided to try it. I am only using the free plan so let me know if I'm missing something. I am assuming the parent was referring to the tool hosted at pangram.com and not some other tool of the same name.
Pangram indeed claims the OP is 76% AI-generated. It has "high confidence" (EDIT: some parts are "medium confidence") that the early portions of the text were created by AI, and "medium confidence" that some of the later potions were written by a human. EDIT: I was especially dismayed to see that the dog might have been an AI creation :(
When I use the "supporting evidence" option, the main piece of evidence Pangram provides is the frequent use of em-dashes. Each timestamp is followed by an em-dash. Personally I think the em-dashes could be a copy-pasted em-dash or inserted by a markdown to HTML converter. nesbitt.io is apparently using Jekyll [0] - any Jekyll users know anything about this??
Pangram's "supporting evidence" feature also considers → and € to be "unusual Unicode".
Personally, to me it looks like the "supporting evidence" feature still needs some work because Pangram's AI detection is probably a lot more sophisticated than a grep for Unicode symbols. In fact the feature even has a notice claiming that "These patterns aren't used to determine our AI score; they help you see why AI text often reads differently."
As for the rest of the OP's content, it would be interesting to compare the Pangram results to a timeline of a real vulnerability. I tried doing so, but exhausted my free "Pangram credits" - apparently the first 1000 words of this article [1] about the log4j vulnerability is considered 100% human.
Not a chance. Far too funny, too well written, too terse while being densely packed with wit. I see zero signs of it being LLM-generated and lots of stuff LLMs have no way of doing.
If I am somehow wrong I would salivate at a chance to see the input.
The author suddenly began writing a post per day around November 2025. They’re all tongue-in-cheek. I believe you are wrong.
Huh, neat. I will take a look at those.
And actually I see it clearly now, it has a bunch of signs I have called out multiple times myself. (It is entirely made out of lists of various types, and never states an opinion.)
Just my ego getting hold of me because I didn't realize it on my own.
1 reply →
You don't even need to read past the first timeline entry. The name "Marcus Chen" is literally a meme within AI creative writing circles due to how often Claude defaults to that exact name when naming fictional characters.
Probably being used to enhance the humor, intentionally.
I never used Pangram before today, but since I've seen it mentioned many times on HN and I enjoyed reading the OP, I decided to try it. I am only using the free plan so let me know if I'm missing something. I am assuming the parent was referring to the tool hosted at pangram.com and not some other tool of the same name.
Pangram indeed claims the OP is 76% AI-generated. It has "high confidence" (EDIT: some parts are "medium confidence") that the early portions of the text were created by AI, and "medium confidence" that some of the later potions were written by a human. EDIT: I was especially dismayed to see that the dog might have been an AI creation :(
When I use the "supporting evidence" option, the main piece of evidence Pangram provides is the frequent use of em-dashes. Each timestamp is followed by an em-dash. Personally I think the em-dashes could be a copy-pasted em-dash or inserted by a markdown to HTML converter. nesbitt.io is apparently using Jekyll [0] - any Jekyll users know anything about this??
Pangram's "supporting evidence" feature also considers → and € to be "unusual Unicode".
Personally, to me it looks like the "supporting evidence" feature still needs some work because Pangram's AI detection is probably a lot more sophisticated than a grep for Unicode symbols. In fact the feature even has a notice claiming that "These patterns aren't used to determine our AI score; they help you see why AI text often reads differently."
As for the rest of the OP's content, it would be interesting to compare the Pangram results to a timeline of a real vulnerability. I tried doing so, but exhausted my free "Pangram credits" - apparently the first 1000 words of this article [1] about the log4j vulnerability is considered 100% human.
[0] https://github.com/andrew/nesbitt.io
[1] https://www.csoonline.com/article/571797/the-apache-log4j-vu...
[dead]