News for hackers. Stuff that matters.
Re:Understanding Computer Memory Architecture and SSD Internals by jeffrallen on Tuesday September 08, 2026 @06:36AM (#49608400)
Pangram says the first 100 words are 100% AI slop.

(Don't shoot the messenger, just trying out Tangram for the first time.)

Re:Understanding Computer Memory Architecture and SSD Internals by visarga on Tuesday September 08, 2026 @08:09AM (#49609228)
Author might have spend days researching it, selecting what to include and what not to, revising the language and graphics, then comes Pangram and slaps 100% AI on it without looking into the substance.
Re:Understanding Computer Memory Architecture and SSD Internals by fxwin on Tuesday September 08, 2026 @09:22AM (#49610005)
Unfortunately, given the sheer magnitude of text on the internet, we need heuristics to gauge whether something is worth reading or not. For authors/blogs that I don't yet know, LLM-generated text is a clear anti-signal for me, and Pangram has been very useful in that regard.
Re:Understanding Computer Memory Architecture and SSD Internals by jnkl on Tuesday September 08, 2026 @09:41AM (#49610236)
What false positive rate would you regard as still acceptable for your use case?
Re:Understanding Computer Memory Architecture and SSD Internals by fxwin on Tuesday September 08, 2026 @11:06AM (#49611436)
False positives (Pangram labels something slop that isn't) I'm not that worried about, since the worthwile articles generally circle back to me some way through the communities I'm involved in, so I like to err on the side of not instantly reading something anyways.

For false negatives, that depends on the type of article at hand. If an error leads to me spending 2 minutes reading an article that ended up being not worth reading I'd be fine with an order of magnitude more than if false positives happen on articles that take 30 minutes to go through. Maybe on the order of 1-3% overall?

Re:Understanding Computer Memory Architecture and SSD Internals by adrian_b on Tuesday September 08, 2026 @07:46AM (#49608981)
Pangram is wrong.

The text does not have any of the superfluous words typical for Claude/OpenAI talk.

It is written like an answer to "Explain to me like I am 5", but it contains appropriate information for such a context.

Only towards the end of the article, there are some signs that indicate AI editing, like a too frequent use of the word "simply", e.g.:

> "The one-clock-cycle delay introduced by an RDIMM is not simply a penalty. It results from pipelining the memory interface. "

This would have been better just: "The one-clock-cycle delay introduced by an RDIMM results from pipelining the memory interface."

Overall, it seems that the AI contribution was minimal.

The problem is that for scientific/technical texts it will always be impossible to determine whether they are written by a human or by an AI, in the case when they are correct.

For now, we detect easily the AI generated texts even for scientific/technical information only because they are bad, by including a lot of non-necessary words or by making unexpected mistakes. When the AI-generated text describing some scientific/technical fact will be flawless, it will not be distinguishable from what a competent human will say about the same content.

TFA might have been 100% AI-generated, but in that case it was a better AI than what has generated the immense quantity of slop that I have seen on the Internet lately, which is very easily recognizable as AI-generated, after reading just a few sentences, unlike the article linked by this thread.