Ryan Moulton writes occasional accessible articles on technical subjects such as deep learning, color, MinHashing and ranking, plus nature topics.
Commenters say he posts rarely but well, praise the mix of tech and nature, and link his color gamut post and other articles.
As a long-time reader of blogs (eg. slatestarcodex.com, kalzumeus.com, etc) I've sometimes struggled to keep up with all the posts that I'd like to read. I often don't make the time to sit down and plough through some of the longer ones.
I noticed that I listened to a lot of podcasts when commuting, so I made this website to convert interesting blog posts into a podcast. The tech component is mostly glue: the trafilatura scraper + AWS Polly TTS.
Given I find this site pretty useful, I figured that some HN readers would enjoy it as well. Here are some of my favourite articles from the past ~18 months with the accompanying audio:
Doing Business In Japan
https://www.kalzumeus.com/2014/11/07/doing-business-in-japan/
https://media.blogreader.com.au/media/e096c7623b305cb129c4dd5ecca8cd75.mp3
A Deep Dive into K-pop
https://dormin.org/2020/09/06/a-deep-dive-into-k-pop/
https://media.blogreader.com.au/media/741f2612536071e911f9a964f5adb65c.mp3
Peer review: Troubled from the start
https://www.nature.com/news/peer-review-troubled-from-the-start-1.19763
https://media.blogreader.com.au/media/4712ffd8d05dd1b51df2b7fedfc3fb82.mp3
The Real Reason Your City Has No Money
https://www.strongtowns.org/journal/2017/1/9/the-real-
reason-your-city-has-no-money
https://media.blogreader.com.au/media/1f92611929679af2763b348b29deaa15.mp3
A Software Engineer’s Guide to…
I wonder when there will exist camera and (laser) projection systems - and with that, movies - that cover even more of the human visible color gamut
https://arstechnica.com/gadgets/2015/10/imax-with-laser-superb-contrast-4k-resolution-and-huge-colour-gamuts/ mentions that with IMAX Laser:
> Colour-wise, the full Rec. 2020 gamut/space is available—though we weren’t shown any films/trailers that actually made use of this larger colour space; they all used the DCI’s standard P3 colour space.
https://en.wikipedia.org/wiki/Rec._2020
I have no idea how well the analog film compares there
https://moultano.wordpress.com/2026/06/19/where-to-find-the-colors-your-screen-cant-show-you/
http://moultano.wordpress.com
I write accessible articles on technical subjects when I have some particular insight to share that I haven't seen written about before. That ends up meaning that I don't post very much, but I think the posts are very good when I do.
Here are some of my best posts.
Logs, Tails, Long Tails: https://moultano.wordpress.com/2013/08/09/logs-tails-long-tails/
Minhashing: https://moultano.wordpress.com/2018/11/08/minhashing-3kbzhsxyg4467-6/
Why companies with unbounded resources still have terrible moderation: https://moultano.wordpress.com/2019/10/02/why-do-companies-with-huge-resources-still-have-terrible-moderation/
It is common to switch to a conservative mode of thought upon entering one's domain of competence;
The question is, though, if the expertise and intuition developed during, say, running XGboost classifiers at scale in AdTech really of much relevance when thinking about large transformer models trained with a self-supervised objective and RLHF?
If you try to study this in depth, these models can do something the usual "datascience"-tier ones commonly cannot: https://arxiv.org/abs/2205.10343 https://moultano.wordpress.com/2020/10/18/why-deep-learning-works-even-though-it-shouldnt/
This kind of talk can be steelmanned, but even that version doesn't have reassuring answers to the likes of
> Okay, you’ve all told us that progress won’t be all that fast. But let’s be more concrete and specific. I’d like to know what’s the least impressive accomplishment that you are very confident cannot be done in the next two years.
(from https://intelligence.org/2017/10/13/fire-alarm/)
Just today I was rather astonished by https://moultano.wordpress.com/2021/07/20/tour-of-the-sacred-library/ -- try digging up something comparable from mid-2019.
most stereograms are designed to look correct when you cross your eyes This is how I look at stereograms (looking nearer than the page), but at least some of the images on this page seem like they're designed for the other way around (looking in the distance, beyond the page).
This one looks weird when I look at it cross eyed, but fine when I look at the other way.
https://moultano.wordpress.com/wp-content/uploads/2025/02/17405320419888595852203168102922.jpg