I just finished my umpteenth reread of AFUtD. It’s one of those books I can pull off the shelf every few years and enjoy from beginning to end.
The Cradle series by Will Wight, starting with Unsouled. Kind of hard to describe, but it’s an sci-fi/Asian fantasy martial arts cultivation epic. If you liked Dungeon Crawler Carl, it may be of interest.
If you haven't read about his (technical) system for writing, do so. Very geeky. He described it (including notes) in my ebook copy of Fire Upon The Deep
My latest trick (literally from yesterday) is to just ask it to write according to ISO 24495-1, the standard for plain language:
> [This standard is] for anybody who creates or helps create documents. The widest use of plain language is for documents that are intended for the general public. However, it is also applicable, for example, to technical writing, legislative drafting or using controlled languages.
Keep in mind all this kind of stuff can make the model less capable. If it has to think in "plain" English, it may well be squashing quality of code etc output.
I'm not sure how true this is, but when using "forced" json output it def had a big drop off in quality - https://arxiv.org/html/2408.02442v3.
I think you're better not fighting it with hacks like this and find a different model.
I would not overgeneralize from paper. Firstly: forcing JSON output is, in my opinion, a bigger change than asking it to match the above style guidelines, and secondly, as is always the case with these kinds of papers, what was true for the model tested in the paper may either be completely false, or greatly reduced, in later models. That paper is almost 2 years old and models today have been trained in very different ways (or more accurately post trained in very different ways) and are in general far more capable.
Based on that paper, I would maybe try to check if it was true for a modern use case, I would very much not assume it was still true.
Tell that to all the CLAUDE.md lines across dozens of repos I have to write to get them to understand that git commit standards and PR description standards are different.
Actually, only the first few pages are available there (introductions, Sections 1-3.8). The meat of the document, Section 5 ("Guidelines") is completely absent.
See:
> Only informative sections of standards are publicly available. To view the full content, you will need to purchase the standard by clicking on the "Buy" button.
I doubt you have to do anything more than that very prompt. I suspect the model has the full standard in its training set, which means it also gets the gist of it due to not just the general associations it makes but also the amount that it's been cited wherever it's been used.
It's somewhat old, but NASA has its SRTM height-map dataset. I wrote a project for reading this in Python many moons ago, in case it is of help to anyone:
They have a special place in my heart. Vinge didn't write much, but what he did write I think is very special indeed.
I've also taken to buying hardbacks of my favourite books from ebay.
https://www.goodreads.com/en/book/show/77711.A_Fire_Upon_the...
https://www.goodreads.com/book/show/226004.A_Deepness_in_the...