You can modify the sample text

Victor Mono made it fairly late in my rounds, but the "@" character looked terrible as rendered in firefox. It looks a bit better in my terminal, which points out to a downside of this that others have observed.

Ended up with Hack, but it is missing the font I use[1]. I think I could live with hack.

It's amazing to me how many of the fonts in this bracket have lower-case ell and one as nearly identical (both with a line at the base and some protrusion to the left at the top).

On improvement would be to allow scalling different fonts to different sizes. With them all at (e.g.) 16pt, there is a wildly different amount of text on the screen.

1: https://github.com/belluzj/fantasque-sans

That's more than double what I'm paying for 3 phone lines now.

See also https://gittup.org/tup/build_system_rules_and_algorithms.pdf

Ninja is possibly the best example of the "Do one thing and do it well" philosophy. All it does is execute commands based on a static build graph.

It's syntax is simple enough that it's trivial to e.g. write a shell script to generate the build items if you need dynamic dependencies.

What does it say about me that when I run my writing through one of those "detect if AI" tools I seldom see a value of less than 70% confidence that the writing was AI generated?

I know this is a spicy take, but it probably just means you're more eloquent in your writing than most netizens...

And that's not really a hard bar to clear if you look at how people write comments online (including places like GitHub).

Anyone that uses punctuation, and capitalises words, probably automatically gets past the 70% confidence line.

It baffles me when I see ostensibly smart people refusing to click shift. Especially programmers. I know you can do it! I've seen you use curly brackets!

AI detectors don't work.

What it says (and this fact is not popular around here) is that you write better than the average person.

Simple: The derived variance in your word usage and sequences, is outside the mean distribution range, that would be labeled as AI generated, given this specific evaluation algorithm

It’s not nondeterministic

you can probably do the shannon entropy calculation yourself if you understand what the evaluation algorithm is

That said…if the evaluator is non-deterministic, then there’s no value in the estimate anyway

It probably means that your writing stylistically is close to the vector-space average of "good" writing, which is what AI produces.

FWIW, your comment history here does not look like AI at all to me, and I think I have a very (maybe too?) high sensitivity to AI slop.

I haven't tried my HN comments; I've only tried things spanning more than a few sentences and that I've put more effort into. I only discovered this when my son put an e-mail I wrote to his teacher that he was CC'd on into the tool on his school iPad.

> and they have nearly 0 false positives

I really don't see how this can be possible unless they're accepting abysmal recall? Perhaps I'm missing something fundamental here, but the idea that AI and non-AI assisted text can be separated with "nearly 0 false positives" just says to me that it's really just a filter for the weakest, most obvious AI generated text. Is that valuable?

Pangram's explicit pitch is extremely low false positives, accepting that a higher rate of false negatives is acceptable.

try it with something published before 2022. do you still get the same results?

I really doubt those tools are good for anything

I've not spent significant time with border collies, but I'd say that if I had to rank, multiple species of corvids are smarter than german shepherds (a breed I'm more familiar with).

So basically it's like half of all therapists.

My mom had that exact poster in her home-office.