Source

Branch

1/ @zeuxcg

@TheGingerBill

Missing from this list - your recent blog post seemed, at best, very heavily LLM assisted. Unsure what the process there was but it was painful to read.

2/ @TheGingerBill

@zeuxcg

My blog post was not LLM written nor assisted.

I have been trying to write better, since my old style as not great, but I clearly failed in that regard if many keep thinking I used an LLM for it. I guess I’ll go back to my poorer style even if it is worse.

3/ @zeuxcg

This is quite surprising; there were so many tells that I was very sure, and Pangram is very sure too. Sorry if that’s not the case. I’m specifically referring to “typed assembly” post; the previous posts have a very different writing style that Pangram recognizes as human written. You’re saying there was no LLM involved, even as copy editor?

4/ @TheGingerBill

I used an LLM only to fix typos and grammar. Nothing more.

I have even passed my article through those LLM tests since you are not the first to bring it up. Most of it is nonsense and focused on my use of em-dashes or mixing conversational and technical language.

It is frustrating that I am being accused of something I didn’t do.

5/ @zeuxcg

@TheGingerBill @max_spero_

kinda curious where things fall short; the post in question is https://www.gingerbill.org/article/2026/08/20/designing-odins-inline-asm/ - pretty much any set of paragraphs here Pangram claims to be “100% AI” (but “paraphrased”?)

6/ @Wassimulator

@zeuxcg @TheGingerBill @max_spero_

it’s kind of annoying seeing this passed around, Bill just writes like this, and I can’t imagine him ever publishing something an AI wrote.

7/ @zeuxcg

@Wassimulator @TheGingerBill @max_spero_

No, all earlier posts do not read the same way to me. I really like Bill’s writing and this was the first post I bounced hard off. Again, apologies for implying something that was not true.