Leonard Richardson's Blog, page 2

My PyCon US 2024 talk

I've put up a transcript of the talk I gave at the PyCon US Maintainers Summit last month, about the lessons I learned while being the solo maintainer of Beautiful Soup, over 20 years and through two periods of professional burnout:

How to maintain a popular Python library for most of your life without with burning out

The quick takeaway is that strong boundaries are important: both the software boundaries provided by published APIs and packaging dependencies, and the decision as to where your volunteer open source work ends and the rest of your life begins. I have some suggestions for the ways the two interact, and an anecdote about how we mentally rewrite our memories of our struggles to make ourselves more active participants. If you're the maintainer of an open source project, I recommend checking it out!

Tapes And Transcripts Are Available!

I've updated The Schickele Mix Online Fan Archive with... transcripts! All of the 150-ish episodes archived by fans now include transcripts of Peter Schickele's wisdom and silliness, cross-referenced to the corresponding timestamps on the Internet Archive. Here's a random example: the transcript of Episode 84, "Clarinet Plus".

Some of these transcripts were created by running Whisper on my computer; others I created by paying someone else to run Whisper on their more powerful computer. Now that I've put it all up, one transcript per page, it doesn't seem that impressive, but it's a solid [runs script] 63.80 hours of transcribed text; that's after all the music was filtered out.

I've also updated the dataset with some previously missing information, thanks to Reddit user kiyyik. Remember, if you've got any Schickele Mix recordings, I'll take 'em!

Although the .srt files available for download are the originals as they came out of my/someone else's Whisper process, I wrote some code to tidy up the transcripts for the HTML views. Apart from cleaning up common hallucinations such as transcribing orchestral music as "����" or "Thank you.", I caught and corrected forty different ways to misspell Peter Schickele's name. Here they are:

Chicelet
Chick-Alee
Chick-fil-A
Chickalay
Chickaly
Chickelet
Chickley
Chickly
Chik-fil-A
Cicholet
Schiccoli
Schick-Alee
Schickel
Schickeli
Schickelman
Schickely
Schickley
Schickli
Schickly
Schiekely
Shicabley
Shickeley
Shickely
Shickily
Shickley
Shickly
Shiggly
Shigley
Shikali
Shikely
Shikily
Shikley
Shikoli
Shikolik
Shinkley
Sickily
Sickle-ee
Sickley
Sickly
Sickely

Who could forget Captain Picard taking on the Shikolik?

Schickele Mix Archive

To commemorate the recent passing of composer Peter Schickele (R.I.P.), I've created The Schickele Mix Online Fan Archive. This site presents all of the extant information online about Schickele's amazing music-education radio show Schickele Mix, which stopped airing in 2007 and has been in copyright clearance hell ever since—an inevitable but undeserved fate.

I've scraped the now-defunct official Perl CGI that gave out Schickele Mix listings, and reformatted the listings with links to archived recordings of all the episodes that have been saved by fans. About 130 of about 180 episodes total have been archived, thanks entirely to two people, both of whom show up in this Reddit thread to take credit. Thanks, Frodo_Picard and gattgun, from a grateful world.

All this metadata is now available as a big JSON file (so no one else has to download those old listings from the Wayback Machine and rewrite my scraping code), and I also created a podcast RSS feed that lets you experience Schickele Mix in its haphazard original broadcast order. (This is different from the benofsky.com Schickele Mix podcast mainly in that it includes the gattgun archive, not just the Frodo_Picard archive.)

I've got one more big piece of this project planned, but this is enough to tell the world, I think. The web page makes it clear which episodes of Schickele Mix are still missing fan-archive recordings. If you think you might have one of the missing episodes on an old cassette tape or something, please email me at leonardr@segfault.org.

The 2023 Crummy Review of Things, Part 2: Books and Games

OK, I've finished looking over my records and I'm ready to do book and game recommendations. As often happens, the Crummy.com Review of Things will proceed in fits and starts this year, but I'll eventually get it all done. I've discovered that this is now how I keep track of the passage of my life, rather than blogging the things I do right as I do them like a normal person.

The Crummy.com Book of the Year 2023 is Happy Snak, because it's the only book I wanted to do its own blog post about. Other good books I read in 2023 include Rescuing Prometheus: Four Monumental Projects That Changed the Modern World by Thomas P. Hughes, and the conceptually very similar The Backroom Boys: The Secret Return of the British Boffin, by Francis Spufford.

The Crummy.com Game of the Year is Mr. Sun's Hatbox by solo developer Kenny Sun, which sets roguelike combinatorics in a platforming environment with a slapstick style to create a comedic, customizable Spelunky-style experience. Other games I really enjoyed include: The Legend of Zelda: Breath of the Wild, Mosa Lina, Factorio and Void Scrappers.

Finally, Sumana and I added a few new games to our irregular daily rotation:

Diffle, a Wordlelike that, apparently, fits with the way we think about words.
Connections from the NYT.
Metazooa, Metaflora, and Birdle, games of cladistics.

Next up: something else!

Happy Snak

While making a list of books I read in 2023 I realized there was one book that I wanted to give its own post: Happy Snak by Nicole Kimberling, published in 2018. I was recommended this book by, I believe, Liz Henry, who said that she recommends Happy Snak and Constellation Games as a package deal, since anyone who likes one will like the other. As the author of Constellation Games, I had to test this hypothesis.

Well, I can tell you that Happy Snak is a lot of fun, and definitely has a similar feeling to Constellation Games. It's got weird aliens, cultural exchange focused on low culture, and a snarky main character whose reaction to first contact is to start a business—all things I love reading and writing. It gave me a fun "You're probably wondering how I got into this situation" feeling, so check it out!

The Crummy.com Review of Things, Part 1: Film

2024 already, eh? Well, I'll do the easy part. I've updated Film Roundup Roundup with the best 25 films I saw in 2023. Actually 27, since I'm counting the entire Before trilogy as one entry. After 10 years of Film Roundup, there are nearly 300 movies in my "recommended" list. Amazing!

And as long as I'm messing around with this spreadsheet, here's a top ten of the films I first saw in 2023:

Before Sunrise/Before Sunset/Before Midnight (1995/2003/2013)
Godzilla Minus One (2023)
When Harry Met Sally (1989)
Polite Society (2023)
Barbie (2023)
Oppenheimer (2023)
The French Dispatch (2021)
Eat Drink Man Woman (1994)
What's So Bad About Feeling Good? (1968)
The Blind Man Who Did Not Want To See Titanic (2021)
Asteroid City (2023)

Whew!

Today, on the last day of NaNoWriMo 2023, I have met my (very) longstanding goal and finished a (very) rough draft of The Constellation Speedrun. Finally complete at 150,285 words.

When writing the final scene, I also came up with a brilliant game idea: a robot-programming game in which the programming language used by the robots is procedurally generated on every run.

The Hallucinating Detective

I've put up my 2023 NaNoGenMo project, The Hallucinating Detective . A simulated murder mystery environment is used to coax a variety of large language models into playing detective. Which models will have the guts to actually investigate the mystery, rather than continually saying "Hey, I'm just a text generator, don't ask me!"?

Surprisingly, the answer is—primarily—models trained on something other English prose. Of the twenty-four stories in The Hallucinating Detective, I think the most interesting is "The Adventure of the Rapid-Fire Squirrel". The rift-coder model, trained on Python and Javascript repositories and programming problems, starts off thinking that figuring out whodunit is as simple as writing a solve function, but then starts interrogating suspects and exploring the imaginary crime scene. Many of the other models can't get past asking the suspects why they just said whatever (randomly selected) thing they just said.

I tried to vary up the prose styles in the prompts, from hard-boiled to comedic, but rarely did a model pick up on the request. I will say that em_german_mistral—a model presumably trained on and designed to output German text—not only produced very good English, but picked up that Sherlock Holmes could be a character in a Sherlock Holmes-style mystery: see "Sherlock Holmes and the Draining Pen Affair".

Finally, I want to highlight a beautiful, melancholy poem I discovered in the random Project Gutenberg selections that drive the dialogue in The Hallucinating Detective: "To His Brother Hsing-Chien, Who was in Tung-Ch'uan", written in the year 815 by Bai Juyi.

You are parted from me by six thousand leagues;

In another world, under another sky.

Of ten letters, nine do not reach;

What can I do to open my sad face?

Thirsty men often dream of drink;

Hungry men often dream of food.

Since Spring came, where do my dreams lodge?

Ere my eyes are closed, I have travelled to Tung-ch'uan.

How to (Finally) Follow Instructions

Way back in 2012 I gave a talk about hypermedia and code-on-demand called "How To Follow Instructions". I've always thought of it as my "lost" talk, one that could have been as influential as "Justice Will Take Us Millions of Intricate Moves". I've also convinced myself that this didn't happen because I never put the text of the talk online. Unlike most of my talks, I didn't write the script ahead of time, and transcribing it was a huge/expensive job.

But ten years later, Whisper makes it cheap and easy to do basic audio transcription with a laptop. I've used Whisper to transcribe my talk and edited it into what it should have been. Some of the talk has aged poorly: the same underlying technology that transcribes the audio also makes it possible for a computer to follow some of the human-readable "instructions" I mention in the talk. But I think it was pretty prescient at describing what was happening in the world of APIs and where we've gone over the intervening decade.

Sock breakthrough!

For about four years now I've been low-key searching for a replacement for my beloved Muji recycled-yarn socks, and I'm happy to report a breakthrough, thanks to another Japanese retail brand, Uniqlo. Here's the extremely detailed report:

The closest match I've found to the old Muji recycled-yarn socks are now Uniqlo's "melange socks"; a mix of cotton (80%), nylon (16%), polyester (3%) and spandex (1%). They don't feel as heavy as the previous champion (Muji right angle pile short socks), and I wore them through a recent heat wave with no problem. They even look like the old Muji socks, with a gradient of yarn colors, which makes me think they're manufactured with the same process.

Uniqlo melange socks are available as short socks and the misleadingly longer half socks. They are a little larger than the old Muji versions, which is okay with me as I always thought the Muji "short socks" were a little too small for my feet. Apart from that, the only real difference is the cotton-dominant fabric mix, where the recycled-yarn socks were mostly polyester.

Thanks to Sumana for dragging me into a nearby Uniqlo; otherwise I would not have found these.