TITAA #53: Quantum Angel Art Cards, Oh-Oh.
Simberg in Constellation - AI NPCs & GDC - Binarization - DuckDB Recipes - Prefect Dreyfus - Cosmic Wheel Sisterhood
This month’s TV love for me was Constellation on Apple TV+. It’s a weird parallel universe tale about astronauts coming home to perhaps the wrong world. This doesn’t do it justice. Jo the astronaut is on the ISS, there is an accident, and she barely makes it home to a world where things are slightly wrong. There is a sequence in it, the centerpiece of the series, in which Jo and her daughter Alice run in and out of dark cabins that may or may not be the same one. Once the cabin has a painting on the wall by Finnish artist Hugo Simberg, “The Wounded Angel," and another time is has his painting “The Devil By the Pot” (or is it “The Changeling?” — the show proposes). “The Wounded Angel” is also the title of episode 1 of the series.
Simberg (late 1800s) was a classically weird painter, “symbolist”, depicting folkloric themes and inventing his own mythos of devils, angels, demons, skeletons, and fairytales. He was a student of Galen-Kallela’s, who was also quite fantastical. Simberg’s devils are often “poor” or sad. In another of his famous works, skeletons tend to gardens (left),
While on the right, a skeleton guide takes a man to the gates of death. In the white door scene, laid out like a playing card, is it the door to Heaven? The man holds a deck of cards and one has fallen to the ground, the ace of spades. Presumably he ends up in Hell on the right.
It’s quite hard to find good texts in English, at least I haven’t yet. Gabrielle Kezia, focused on his fantasy elements, quotes him writing to his brother,
“I’m sitting on something big and shapeless; I don’t know what it is, it takes me somewhere, I don’t know where… The journey gets brighter and brighter, and I wake up more and more. Eventually I am fully awake, and I see something big and shimmering that will soon disappear. You know, that’s exactly what I feel when I watch the clouds pass over the moon on a moonlit night. If you experience something like this when you get to see my picture, I have succeeded.”
There’s a very nice VR Quest App Lab app, “The World of Hugo Simberg,” which came out end of last year, focused on some of his mythological work. You get some nice environments based on his paintings, animated a bit, with more paintings, text about them, plus a bit of his writings—too little translated into English, though.
Some of his work is influenced by theosophy and spiritualism. One of the curatorial texts in the app says, “Simberg too adopted the idea that art can uncover a path to secret truth and knowledge.” And on the undated “The Moon Is Rising” work: “It reminds us of the artwork’s significance as a vision-like flash that reveals the profound essence of reality. An artist must continually develop their senses, so that they might become a seer of secret worlds.”
You can browse a giant collection of his black and white photographs, writings, and artwork in this Finnish digital collection viewer. In the media recs companion post to this issue for the paid subs, I added screencaps of some of the curatorial texts in the VR app. In any case, I suspect it’s worth looking harder at the connections to the show Constellation, and hope someone writes something excellent on this. Meanwhile, Lu Wilson’s javascript mind-bender art sketch captures the vibe of Constellation for me.
Onto the news and recs! Copious as ever!
TOC (links on the web page/app view):
AI Art News (Video, 3d, Misc with Think Pieces)
Narrative News (Including comics, story evolution, fairytale data…)
NLP & Data Science/Vis - Embeddings, DuckDB, Tutorials & Tools etc
Recs (Books, TV, Games with VR) — longer form in the paid separate mailing.
AI Art Tool News
AI Video - A Few Notes
More OpenAI Sora footage came out from small artist-invited video collaborations. Much discussion ensued online both admiring and pooh-poohing. Here’s a nice piece in MIT Tech Review about the construction of three of them, “Air Head,” Paul Trillo’s trippy one (I love infinite zoom/dollies, and he did no post-processing), and Don Allen Stephenson’s impossible animals.
AI generated TV streams - I thought these were over, but according to Adam on Garbage Day, there are 2 running strong, AI Peter (from Family Guy) and AI Sponge for Sponge Bob. They have discords and fan communities and fan-directed generation. So, for instance, you get a Discord vote for:
And this happens, whatever this is (grabbed from a scrub thru the video):
YMMV but it is surreal and often as vulgar as you’d expect.
The code and demo for DynamicCrafter is out, which is meant to interpolate movement between two image frames, for say, a guy smiling. I am all about the weirder interps, though. I immediately tried with Simberg paintings, on the top of my downloads, and got this jaunty cooking devil disappearing (see top article) and cutting to maybe becoming the angel?
The Dough toolkit (OS code) is kind of an interp and video control framework, which allows you to make a video from a bunch of guidance images and possibly use a video (even just for camera motion) to style an output shot. (H/t Dreamingtulpa.) I’d like this kind of feature in all the video generation tools with UIs — provide image steps and interpolate between them.
Misc other Sora-like efforts: VSTAR, which doesn’t have code yet but coming soon, says it does longer generation. Examples do have multiple scenes. Same with StreamingT2V. But it’s hard to be very excited about anything that just generates meh video, having seen Sora output. I recognize these are stepping stones. Similarly, Mora is an OS attempt at Sora.
3D Teeny Sample
In 3D Gen, there have been approximately a hundred 3D gaussian splat improvements over the past 2 weeks. Dreamingtulpa’s newsletter is following them more closely, and you might want to watch the X account of MrNeRF. Here’s a couple cool ones — this ThemeStation project is like a 3d style generator, given some examples, it will generate items in the same style, combining attributes. Kind of proc-geny?
The GRM (Gaussian Reconstruction Model) 3d splat project, with another winner of a weird demo video, has added code and a HF demo. It’s surprisingly good! Check out my crow from text only! I asked for ragged, with yellow eyes, and an open beak. I got a mesh. It will also do image (with bg removal) to 3d. The demo will even generate an image for you to do the image2model part.
Misc AI, Think Pieces, Procgen
🥘 🎵 New AI art experiments from Google friends: Food Mood and Musical Canvas from colleagues in Google Arts & Culture’s creative lab: Food Mood uses AI to create menus for new dishes with ingredients from different countries (you pick!); and Musical Canvas lets you sketch a picture and then generates music inspired by it—plus it says nice things about your (my) awful sketch.
Music gen: I was going to skip this, because I have to draw the line somewhere and I’m not musical, but then I noticed that everyone’s darling Suno.ai was founded by people I used to work with at Kensho in Cambridge, Mass. (Rolling Stone piece on them/it.) Go Kenshens! Suno.ai will generate music and lyrics for you. It’s a nice UI, i.e., it’s not a damn Discord— but no, really, it’s cool. I found the lyrics the weakest part. Here’s its attempt at a song about quantum worlds and identity:
[Verse] In the shadows, we exist Parallel worlds, intertwine and twist Lost in the haze of a twisted reality Searching for answers with no clarity [Verse 2] Identity blurred, lines become thin Aching hearts, longing to begin A voyage through the depths of the mind Unveiling secrets, we long to find [Chorus] In this parallel existence we sway (oh-oh) Lost in the melancholy of our own dismay (dismay) Tripping on the beats of a twisted groove (groove) Acid jazz whispers, a trip for the soul to soothe (soothe)
Terrible. But a hard task. 😅
Are Large Language Models Creative Problem Solvers? Dataset too, from AllenAI. “In contrast, LLMs, exposed to a variety of specialized knowledge, attempt broader problems but fail by proposing physically-infeasible actions.” The data is pretty cute:
Image gen: Fouriscale does large, high res, any proportion sizing image generations. They look super.
Hey, Cables.gl for webgl and procgen, via node diagrams, is still going strong. Gotta try it properly someday. After I learn ComfyUI.
A couple good think-piece readings:
📷 “On the Pictures Generation and AI Art,” by Kazys Varnelis (h/t Mark Nelson). This is a very thought-provoking and wide-roaming take on AI art and originality, with a heavy dose of French philosophy, art history, and a little bit of weird, citing Stranger Things. Some of his touch points include things I’ve covered myself, like the fact that many of the phone pics we take are already “AI” enhanced (my issue starting with “authenticity”) and David Salle’s work with AI in NYT: “Is It Good Enough to Fool My Gallerist.” One thing Varnelis grapples with is Instagram’s over-saturation of the landscape and tourist photos taken by amateurs (my word). He pulls on the photography movement of the “Pictures Generation” as a frame that may be useful right now.
“Rather than being seen as the singular source of meaning, artists of the Pictures Generation positioned themselves more as curators or commentators, utilizing the visual languages of their time to critique cultural norms and values. This reflects a move away from the modernist emphasis on the artist’s unique vision toward a recognition of the complex, contextual nature of art-making and interpretation.”
Also see Lev Manovich’s Artificial Aesthetics article collection on AI art. Chapter 7 came out in March. It reads as a personal essay, an attempt to come to grips with the new tools (and a lot of reflection on his own past work and images he’s making in Midjourney). The end feels a little poignant.
Because AI is so vast and endless in its knowledge and skills, you needed to work on the micro-scale. Very narrow. So narrow that AI can’t quite get there. Through the needle eye. Only in this way can you compete with superhuman generative AI.
Games News: GDC, AI NPCs and More
Tommy Thompson of AI & Games was on the org committee for the AI & Games Summit sessions at last week’s Game Developer Conference (GDC), and he has a great summary overview piece. He tallies up the general interest and fears, including the risk aversion of most AAA game devs wrt tools used (tooling has to last for 5 years or so!)
AI NPCs were discussed a lot at GDC (and I guess at the concurrent NVIDIA conference). Here’s a very nice, thoughtful Aftermath article by Nathan Grayson summarizing his take on the demos from “Nvidia, Convai, Unity, Inworld, Microsoft, and more.” (Not Ubisoft but see below.) Lots of them are partnering with Inworld.ai. One smart observation before a lot of deeper thoughts about narrative limits:
“This is the second, less-discussed uncanny valley of AI NPCs: If they can respond to anything you say, you quickly come to expect that they should be capable of carrying out any corresponding action. That, however, requires an entirely different scale of production than simply hooking an NPC up to a bunch of large language models. Suddenly, they need to be able to recognize discrete items in the environment and even interact with them.”
Tommy Thompson said it as, “If I can have a chat with a character, why do they have zero spatial awareness? If they can tell me the backstory of the world, why can’t they help me with my objective?”
Hollister at The Verge played with the Ubisoft NPCs, and said “One of them was slightly better written thanks to an actual narrative designer scripting their backstory, but delivery was still awkward, a little stunted, with lag before replies and occasional vocal stutters.” He still enjoyed some of them. When I played the Kraken game with LLMs on rails, the delivery was easily the worst part, to be honest. And here’s a Verge piece on NVidia’s efforts here.
There’s a lot in that Aftermath piece by Grayson about possible ways the work of a narrative designer changes if they use these tools all the time in their current state. There are a few scenarios for their use, but “to what end?” is always an open question right now, especially given the expense of the smartest models running “live.” But, fine article aside, at least one dude at a16z who does games investing thinks the pawns in Dragon’s Dogma 2 are “AI” (X thread, sorry) and they are of course the best thing about it, according to everyone. Although I might be sold on the real time hard travel open world, myself.
Regarding more general AI uses, Mike Cook wrote a series of 4 articles about AI and Games in RockPaperShotgun; here’s the last one with links to the others. Recommended as a clear intro to the issues and topics in debate. Not all AI in games need be LLMs talking real-time, it could be used for other algorithms or for static content generation or even testing.
Speaking of AI in games, I forgot to cover in this last issue — a well-liked mystery deduction game free on Itch, playable in your browser, “The Roottrees are Dead” (AV Club article on it), was made by a single developer for a game jam and he used AI art (MJ) to illustrate it. This put some folks off, but for others, it contributed to some creep-factor. He has announced it got quite popular and he is hiring artists and remaking it for Steam release, with some expansion. Arguably the game would not have been possible for a single person to make without AI art generation assistance? AI is a great prototyping tool, in any case.
Rosebud.AI (h/t David Ha) - a woman-led company offering to take you from text description to tiny game. Here’s an interesting blog post about a beta tester’s games — they’re really simple, tending towards 2d, pixel art, and text, unsurprisingly. But it’s a good read. She has some videos (and here) on the sfka Twitter showing how she edits and creates in the tool.
Onto the less AI-ish games stuff…
Minecraft with insane shaders.
One of Ukraine’s most famous mines has been preserved in Minecraft - via RockPaperShotgun, as a fund-raiser:
With the help of French devs Endorah and miner Stepan Bandrivskyi, who worked in the mines for a decade, Ukraine has recreated the Soledar salt mines and part of the surrounding town as a playable Minecraft level dubbed Minesalt. With physical access to the real-life mines impossible due to the Russian occupation, the devs spoke to former workers and used a combination of existing photographs, videos and even hand-drawn maps to make a faithful recreation.
Progress Quest (wikipedia) — via Triptych - I did not know about this. It is officially awesome. “It is loosely considered a zero-player game, in the sense that once the player has set up their artificial character, there is no user interaction at all; the game "plays" itself, with the human player as spectator. The game's source code was released in 2011.” HAH: “Critics have commented that, despite the automatic progress, Progress Quest was an enjoyable experience.”
“‘I wasn’t sure it was even possible’: the race to finish 80,000 levels of Super Mario Maker” — best recap I’ve seen of this, in the Guardian. Nintendo is shutting down the servers, so a bunch of folks tried to finish all the created levels. (Here’s ArsTechnica’s article.) I would read this as a novel. Even the twist ending.
The appropriately named The Last Dance was a maze of spikes, ghosts and spinning blades, punishing any missteps. It also requires stamina, demanding flawless play for more than two minutes. “It was on nobody’s radar before probably the last 10 levels,” Fritzef says. Meanwhile, Trimming the Herbs was uploaded in 2017, and despite more than 200,000 attempts, it had never been beaten. It should take only 17 seconds to complete, but to do it, a player must pull off a string of frame-perfect jumps. It should be possible; its creator even shared a video of how to do it.
Interesting experimental games roundup from a session at GDC. Also a good overview of pieces on the IGF winners in GameDeveloper. I’ve played none of them 😢 not even Venba, the lovely-looking Indian cooking game. But wow, this interview with the writer of Mediterranea Inferno made me buy it immediately.
👣 Art Movements in Video Games, Justwalkingism, by Oscar Barda in 2014 - an old piece that I really, really like. Focusing a lot on walking sims and open world games, or “toys,” he has a ton of great reference games, some of which qualify as “weird.” “We can think of games like colouring books: the outlines are there but the finished piece is still yours to make.”
🎨 Level up: exploring the artistic side of video game design by Angelica Frey in It’s Nice That. This is a great look at custom art (and music) in games with hand-illustration, your Dordognes, Venbas, Sables, Neon Genesis Noirs, Journeys, etc. Really terrific piece. “Existing in parallel to a world of hyperrealistic CGI is a growing segment of gaming using illustration and painterly art to make mesmerising and charming visual experiences. Angelica Frey explores this world of video game design and speaks to designers at the creative forefront.”
Conferences! Narrascope schedule is up (it’s this summer)! Same with the ThinkyCon Conf schedule for puzzle game talks, which is April 4-6. The Wordplay Workshop at NLP conference ACL 2024 is open to subs related to interactive narrative including text games till mid-April. I gave an invited talk there a couple years ago about some stuff I find fun in texty games and games in general.
Dog Poo Golf (in the browser). Via waxy.
Narrative
“Cultural evolution in populations of Large Language Models” by Perez et al. With code. They simulate the sharing of information in stories over time in a social network via a multi-agent LLM system. It’s basically a telephone game. There’s even a proto GUI. 😅
“It allows organizing LLM agents into networks wherein each agent interacts with neighboring agents by exchanging stories. Each agent can be assigned specific personalities and transmission instructions, serving as prompts for generating new stories from their neighbors’ narratives. … We also provide quantitative insights from the generations of stories by tracking the evolution of similarity between new generations of stories and the the initial one, as well as the evolution of similarity within generations and with successive ones. We also analyze the evolution of measures such as positivity, subjectivity and creativity across generations.”
And a zinger in their findings… “It also suggests that results from studies of human-generated culture may apply to machine-generated culture.” Why not.
Neural Text Rewriting (thesis by Lai): “In this thesis, we focus on Natural Language Generation and zoom in on controllable text rewriting, exploring how to model and manipulate the style of the input text, automatically generating a new text.” Are you interested in text style? I’ve been tracking a few of these for reasons, let me know.
Mythalu is accepting AI-co-authored stories and art for a special issue, due mid-April. (Via Amit Gupta on X.)
The AI-Gen Comics thing is still going strong, even stronger with the opening up of Lore Machine after a year of dev (MIT Tech article). And Julian Bilcke’s Hugging Face Comic Factory now does multipage. Julian made a list of other comics-gen tools here.
The “Inspo: Writing Stories with a Flock of AI and Humans” paper has been updated.
🧚 FairyTaleQA repo and paper: my new favorite dataset. (From 2022.) Related dataset building on it: FairytaleCQA: Integrating a Commonsense Knowledge Graph into Children's Storybook Narratives, 2023. “A follow-up experiment shows that a smaller model (T5-large) fine-tuned with FairytaleCQA reliably outperforms much larger prompt-engineered LLM (e.g., GPT-4) in this new QA-pair generation task (QAG).”
I’ll be covering the weird Claude backrooms generation effort of world_sim in my mid-month weird & esoterica newsletter, behind the paywall (last one opened with some of the precursor work here).
NLP & Vis & Data Science
A couple data viz projects:
Models All the Way Down - by Christo Buschek & Jer Thorp, a datavis scrollytelling illustration about the construction of LAOIN 5B’s dataset used in most image generation training. It’s good if long; I’d like to note that all the serious image gen orgs have continued to collect new aesthetic ratings from their users as part of their training process (yes, probably still WEIRD but evolving).
Also see Santiago Ortiz’s vis of branching responses from ChatGPT at “Mind.” This is an interactive explorer of possible completions, in 3d and hyperbolic branching tree. Dizzying.
Binary embeddings are all the rage this past week. You can evidently massively speed up embedding search if you quantize them (and some are even nestable). This means you may not even need a database behind them.
Cohere’s BinaryVectorDB code. For use with their offered binary embeddings.
Evidently with Nomic’s you can just binarize by converting to 0 and 1 from their regular embedding (per their reply to Simon Willison).
Related, Matryoshka embeddings (blog post).
Thom Wolf at HF is doing some tutorial material, here’s his Little Guide to Building Large Language Models video. To pair with deep dives from Karpathy, I guess.
Cooking with DuckDB — the inestimable energetic hrbrmstr Bob Rudis (newsletter) is posting a great cookbook on DuckDB. Including tons of CLI tips and bits on using WASM on websites with it, so far.
A useful overview of from the JinaAI founder of some of the gotchas and confusing bits in DSpy, a lib I’ve covered before which attempts to find a good prompt for your desired output behavior by a trial/eval method.
OCR & PDFs in the browser, via energizer bunny Simon Willison. He also wrote a cool CLI to AWS’s textract service.
There were lots of great LLM releases this past couple of weeks, feel free to ask on the comments if you want picks. I suggest subscribing to the AI News summary roundup newsletter that summarizes activity in the main discussion forums if you are serious about trying to keep up. It’s a job on its own to read it though.
Reminder that the media recs appear in a longer form (with pictures) in the paywalled, optional special separate posting of recs only. If you like the recs, feel free to upgrade!
Book Recs
⭐️ Aurora Rising, Alastair Reynolds (sf). Prefect Dreyfus book 1, which is a tour de force of space opera police procedural with some fantastically creepy AI characters. The cruel Clockmaker will live in my nightmares! There are interlocking mysteries involving a strange sculpture, the destruction of a planetoid, and corrupt voting, all set in a system of free satellite habitats, the Glitter Band. In the Glitter Band, citizens all live democratically under whatever local rules and incarnations they like, answering only to their voting body the Panoply. Fab action and derring do.
Elysium Fire, Alastair Reynolds (sf). Prefect Dreyfus book 2, a bit less complex, still with a good mystery, one involving weird abandoned space stations, bloody nightmares, and challenges (again) to democracy and the Panoply. A return visit from an old AI friend.
Machine Vendetta, Alastair Reynolds (sf). Prefect Dreyfus book 3, wrapping up. Another set of interesting mysteries: why are hate crimes being committed against the satellite inhabitants of the hyperpigs? And there’s another fascinating abandoned mansion in space. The Clockmaker re-appears. I did not entirely buy the resolution, but it was another fun read.
The Hunter, by Tana French (mystery/thriller). Fantastic sequel to The Searcher, in which Chicago cop Clay in rural Ireland deals with the return of Trey’s father and his con schemes aimed at a gullible London visitor. Some very tense, slow build character tension as Clay tries to protect Trey in a hands-off manner, while navigating the local gossip network.
The Tainted Cup, by Robert Jackson Bennet (fantasy). Excellent mash of genres and odd characters in a strange land of genetically modified people and plants. It’s kind of an alchemical science mystery, I guess? An eccentric detective who never leaves her home sends her young assistant with perfect recall to investigate strange deaths — people who died with trees growing out of their bodies. Meanwhile, leviathans from the ocean depths are testing the walls that protect the inner lands. Really great read.
TV Recs
⭐️ Constellation on Apple TV+ (sf). A very good “nordic noir” sf mind-bender… there’s an accident on the space station, and astronaut Jo barely makes it home — after seeing some weird things from the corner of her eye. But home seems different. Her car is the wrong color, her husband is estranged, her daughter smells wrong. We gradually suspect there is some kind of universe slippage or quantum event at play and she’s not the only one affected. A key episode features a cabin in the snowy north with paintings by Hugo Simberg (see opening article) that shift each time she goes in the door. The ending, with a very creepy last scene, left room for a new season; and I am seriously crossing my fingers.
Three Body Problem, Neflix (sf). I did not love as much as the rest of the internet — book 1’s 4 episodes felt rushed. I liked more towards the end when we got into book 2 and it seemed to slow down a bit. The main character from the book is split into several young folks in Oxford (I did not love them), and mostly it takes place in England with a bit of Yu Wenjie’s flashback to the observatory in China where it all began. (Mega CW: suicide, bloody violence.)
Girls5Eva, Netflix (comedy). I binged this, it’s fun. A former girl group gets back together and we get something like “Flight of the Conchords” but with (sadly) fewer musical numbers. The dynamics among the characters are fun and
The Gentlemen, Netflix (uh, Guy Ritchie). I binged this in a hotel room on a rainy weekend in Marseille. I survived but felt like smoking and shooting things after. If you like Guy Ritchie’s violent clever crooks and fools, this is for you. (CW: lots o’ violence.)
Delicious in Dungeon (Netflix) continues but in a much, much weirder direction? Eeee?!
Game Recs
⭐️ Cosmic Wheel Sisterhood. Strong rec, shortish (this is good). It’s a pixel art interactive narrative game about space witches, only some of whom are human. You are Fortuna, a fortune-telling witch stranded by your coven leader on an asteroid prison for 1000 years. You construct a new deck of fortune-telling cards from sf and fantasy backgrounds and symbol components. The game offers some surprisingly moving story elements and a weird universe with some deep questions about reality. Much more detail in the separate recs posting plus pics. [ETA: I relented and put one here, because of the title I chose for this post.]
(There are mirrors in Constellation too.)
Paranormasight: The Seven Mysteries of Honjo. A visual novel of ambition and breadth and creep-factor. There are 7 cursed objects in town, with the ability to kill and store souls. Enough souls, and you might be able to perform the a rite of resurrection. But there’s a lot going on including an awful serial killer and possible suicides. Obviously mega content warning despite being non-animated. I like the folkloric paranormal storyline and hope I can find a way to survive it with one character. 👻
VR
Kayak VR, a casual kayak simulator with scenes in tropical and northern waters. It’s not big, and there is a weird amount of ocean trash in the Norwegian setting? But it’s pretty and relaxing.
The Red Stare — This is free and old (2017)! The scenario is that you are an agent looking for Communist spies in the building across from you in the 1950ies. You take pictures of people’s meetings and try to figure out who the spies are. It’s a bit mystery, a bit thriller. I haven’t played for long but the idea is smart and cute. The little camera and fax are great.
World of Hugo Simberg on App Lab for Quest. There’s a little video there if you like. I wrote about this in the intro article above, and there are pics in the separate recs post.
A Poem
This happened long ago when blood root bloomed, the dazed spring still holding onto makeshift railings. We sloshed around winter’s old fields in poor man’s shoes, bought large to grow into. We heard the stubble breathe, caution, caution, saw something white crumple and fall from the sky. A heron? Wild swan? We ran toward it. A wingéd thing, a heap of feathers we carried home, her feet too odd for any shoes. That was the year an angel lived in our kitchen, recuperating on the bench beside Mother’s oven. She isn’t like us, Mother said, when we’re tired or hurt. She won’t put up any fuss. That was the year we learned about earth and its gravities, how they hold some of us down, but free the unearthly. From the kitchen’s back stoop we three watched the angel unfurl her wings one morning and barefoot take flight into the blue, infinite sky.
—Sharon Chmielarz, Commonweal (2015)
Drop me a note if you’re enjoying this! Feel free to share! And happy spring holidays, whatever form it takes for you. Into the blue, infinite sky.
Best, Lynn (@arnicas on the sfka twitter, mastodon, and bluesky and Threads)
+1 text style; Would be interested in seeing more about that.