Tom 7: Badness 0 (Three ways)

Wishes do come true https://news.ycombinator.com/item?id=40063261

That also struck me as a great way to make software more whimsical. Imagine if you randomly got an "Achievement unlocked!" message if you churn through more than 10 GB of data in a single invocation of `grep` for the first time or something. So many possibilities!

There are quite a few errors in the presentation that seem intended to annoy detail-oriented people. One that stood out to me was the use of backslashes in his website URL.

I read the Epsom's Version PDF before the video was released, and the typesetting was absolutely terrible.

https://imgur.com/a/wjK5br5

After the opening on how he's bothered by minor errors, I thought it had to be part of a joke, and it was a setup for a punchline to be revealed later in the paper. But eventually I realized, no, the PDF must not being displayed the way it was intended. I was reading it in Evince at first, but muPDF gave much better output.

The paper has the fifth occurrence of “Wikipedia” in lower case, which greatly irritated me in the Knuth version.

You're getting a lot of glib answers, but in all seriousness, this is one of the amazing things about tom7 videos. It sometimes starts by appearing to be about nonsense or triviality, but as he weaves all these weird stories, they start to come together into observations that are absolutely brilliant and funny and he writes code to demonstrate what he's saying. I highly recommend "Harder Drive: Hard drives we didn't want or need" as an intro to his style, his humor, and the absurd lengths he will go to in order to prove a point. It's in three chapters and the middle chapter still blows my mind.

Tom7 is a somewhat well-known mad computer scientist who specializes in technically ambitious projects of limited utility.

In this latest escapade, he takes inspiration from Knuth's line-packing algorithm used in typesetting beautiful print documents using the TeX typesetting system and intensifies it by training a large language model to automatically rephrase the input text until it is worded to minimize the "badness" of the page layout. Thus, "Badness 0".

It's that but also much more ambitious, impressive, and stupid.

The "three ways" are that you can see his technical writeup in its original words with layout using Knuth's algorithm in TeX, a variant of that article using his own layout engine (and thus slight variations on the wording that, plausibly, maintain the meaning and semantics), and then a video version.

You have taken the first step towards understanding Tom7.

Don't worry, by the end he has fully justified all of the diversions.

It's about aligning text to fit a specific length using language models.

You are then in the correct headspace to experience it.

Truly great works cannot be summarized

The video explains the main idea(s) behind the two papers which are linked at the top of the page.

It's about many things

skip to 7:22, he explains what the video is about.

Don't feel bad. It's not for everyone.

You didn't see the point?

There a BUG. Try reading in Chrome.

Retyping from memory, I guess.

> Hard Mode: Any revealed hints must be used in subsequent guesses

As implemented, hard mode doesn't count gray letters (letters nowhere in the solution) as hints. You're allowed to reuse gray letters, but true hard mode should prevent that.

In his example he knows 3 of the colours as indicated by yellow but he also knows that those letters are not in those positions and hard mode let's him make bad guesses using those letters in the same positions again when it shouldnt