iiinova — where your digital self keeps evolving

Marco Giancotti @marco-giancotti.bsky.social · 13/10/2026

As a non-mathematician, I've been trying to understand the current situation with AI math. Yesterday I read this joke and I think I get it. OpenAI is releasing a ton of jokes no one can understand, and it's not helping joke-dom

000

Reposted by Marco Giancotti

Saleh @mskayyali.com · 04/10/2026

New Video on Interface Studies: What “Infinite Scrolling” was for. Infinite scrolling was built in 2005 to make exploring image search results easier. Twenty years later, politicians want to ban it and the designer most associated with it keeps apologising for it. youtu.be/eDDI9OSWY9k?...

youtu.be

What "Infinite Scrolling" Was For

YouTube video by Interface Studies

0223

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

Read the full post: aethermug.com/posts/pleas...

000

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

"What makes a message "personal"? Is the effort you put into writing the thing that matters, or is that just a proxy for something deeper, like attention and care?"

100

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

"The part that worries me is not people applying their own filters; it's people trying to popularize a stigma against anyone who shares AI-written text, period. They want to actively shame people for doing this."

100

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

"Using an LLM well for anything, including writing, is hard work and requires a skillset—metacognition—that many people don't exercise much." "Could it be that we just haven't yet learned to harness LLMs well for writing?"

100

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

"When you start learning to play the violin, there is an initial period in which all you can produce is horrible screeches that send everyone running. Maybe we're in the screeching stage of AI writing."

A kneeling man with eyes closed and arms spread in a striken expression, a winged cherub playing a violin beside him, a skull nearby

100

Marco Giancotti @marco-giancotti.bsky.social · 01/10/2026

New blog post is up: Please Don't Shame the AI Writers. Some excerpts. "AI slop is bad. The stop must slop." "But should well-meaning people be allowed to use LLMs to put their thoughts into words?"

A bearded man in a laurel wreath plays a violin while a young assistant beside him writes notes

100

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

Read the full post: aethermug.com/posts/backg...

000

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

"You know you've reached fluency in a language when your sense of what's strange in it matches that of native speakers."

100

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

"If you decide to learn Japanese from Dragonball, you're in for a lot of fun. You'll also have to dodge some funky language bullets here."

Anime illustration against a blue sky: a spiky-haired blond fighter in an orange gi leads a group of allies in fighting stances, facing a white-and-purple alien with a long tail and his soldiers in armored battle suits.

100

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

"The advice to never learn from fiction is wrong, but you do need proper backgrounding to avoid the pitfalls. The lies are there, you just need to discount them." "You must learn to notice what the author doesn't need you to notice."

100

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

"I see most people absolutely murdering their own motivation by forcing themselves to slog through sterile textbooks, baby-step speech lessons, and "business-speak" crash courses."

100

Marco Giancotti @marco-giancotti.bsky.social · 18/09/2026

New blog post is up: Backgrounding My Way to Japan. Some excerpts. "The truly strange part for me was that [my first arrival in Japan] didn't feel nearly as strange as I thought it would: there were no major surprises or "whoa" moments."

Black-and-white woodblock print: a shaggy, wide-eyed monk-like creature crouches at the window of a wooden outhouse, one clawed hand raised, and blows a straight stream from its mouth with a small cuckoo flying out at the end of it. Everything around it is ordinary household detail: a tiled roof, a robe on a hanger, New Year pine-and-straw decorations, and a large lidded jar on a pedestal. Vertical Japanese text runs down the right side.

110

Reposted by Marco Giancotti

Lawrence Wilkinson @lhwilkinson.bsky.social · 11/09/2026

A new way into Wittgenstein: "The Foldable Tractatus" @marcogiancotti.mastodon.social.ap.brid.gy (Plus- Theodor Adorno) Beyond linearity: roughlydaily.com/2026/09/11/w...

023

Marco Giancotti @marco-giancotti.bsky.social · 06/09/2026

I've said this elsewhere, but the argument that "if you didn't put effort in writing it, why should I read it?" is not generally solid. Someone could put a lot of effort (in fact, even more effort) in writing something with an LLM just to improve the quality vs their own.

000

Marco Giancotti @marco-giancotti.bsky.social · 06/09/2026

But simply ostracizing everyone whose text has the signs of being AI-written is the wrong thing to do, and it blocks genuine progress. In the worst cases, this feels like envious gate-keeping on the part of the purple group to keep the contributions of the majority out.

100

Marco Giancotti @marco-giancotti.bsky.social · 06/09/2026

That's a separate thing from the problem of AI slop, people using AI to spew tons of near-meaningless and often-wrong words with minimal thinking on their side. That practice should be actively discouraged.

100

Marco Giancotti @marco-giancotti.bsky.social · 06/09/2026

If this distribution is even approximately true (Sturgeon's Law suggests it is) then we should welcome LLM-written text as a way to include the people who, until now, did not have the courage/means to share their thoughts online because of writing skill issues.

110

Marco Giancotti @marco-giancotti.bsky.social · 06/09/2026

My current perception of the situation. It looks like LLM text is a clear step down from human writing only because you don't usually see the 80+% of human writing that would be worse, if people were bold enough to share it.

100

Reposted by Marco Giancotti

Saleh @mskayyali.com · 04/09/2026

Starting a new series on Interface Studies, speaking with people who are building new kinds of interfaces and reimagining old ones. The first episode features @danielpipk.in and Omar Rizwan from folk.computer. Hope you enjoy it. youtu.be/uXHHnps4tU8?...

youtu.be

Conversations 01: Omar Rizwan & Daniel Pipkin from Folk.Computer

YouTube video by Interface Studies

1135

Marco Giancotti @marco-giancotti.bsky.social · 02/09/2026

There is value in finding that someone has put to words—even poorly and approximately—something that you knew already but could never put to words, or never even thought to try.

000

Reposted by Marco Giancotti

Fernando Rosas @frosas.bsky.social · 29/08/2026

What is information, really? Instead of arguing for a single valid interpretation, we propose a pluralistic view where rigorous alternatives coexist — each with affordances and limits Finally published: “Four faces of information in natural systems” 🧵

210628

Reposted by Marco Giancotti

SMBC Comics @smbccomics.bsky.social · 26/08/2026

This is of course an SMBC psyop. COMIC ◆ www.smbc-comics.com/comic/psy PATREON ◆ www.patreon.com/ZachWeinersm... STORE ◆ smbc-store.myshopify.com

1-panel SMBC comic update. A man and a woman are shown talking. The woman says "thanks, man. I really appreciate your insights about modern politics." The man points at her, asking "ha! nice try, KREMLIN! Who paid you?!" The caption below the comic reads "How come nobody ever assumes the people who agree with them are a foreign psyop?"

1329139

Marco Giancotti @marco-giancotti.bsky.social · 01/09/2026

5. You won't be tempted to blame it for your mistakes (not more than you can blame your keyboard for your words, at least)

000

Marco Giancotti @marco-giancotti.bsky.social · 01/09/2026

4. You start thinking about it for what it is, a tool you can be expert or sloppy with.

100

Marco Giancotti @marco-giancotti.bsky.social · 01/09/2026

3. Getting angry at it becomes as hard as getting angry at a car: not impossible, but more easily recognized as useless.

100

Marco Giancotti @marco-giancotti.bsky.social · 01/09/2026

2. The idiosyncratic language of LLMs stops being annoying, because it's not a personality thing, it's a signal you learn to parse. "Hating" Claude's "load-bearing"-speak is like hating the color of your thermometer's mercury column.

100

Marco Giancotti @marco-giancotti.bsky.social · 01/09/2026

Why you should not treat AI as a person: 1. When you are alone, you should know you're alone

100

Marco Giancotti @marco-giancotti.bsky.social · 28/08/2026

5. Thus, Clarke's Third Law is tautological.

000

Marco Giancotti @marco-giancotti.bsky.social · 28/08/2026

4. Then you can define magic as "technology that defies understanding."

100

Marco Giancotti @marco-giancotti.bsky.social · 28/08/2026

3. Magic is only magic if it defies your understanding of how it could physically work. Otherwise it would be engineering.

100

Marco Giancotti @marco-giancotti.bsky.social · 28/08/2026

2. Magic is a tool made by hypothetical humans called wizards, witches, sorcerers, etc.

100

Marco Giancotti @marco-giancotti.bsky.social · 28/08/2026

1. A human-made tool of any kind is technology.

100

Marco Giancotti @marco-giancotti.bsky.social · 27/08/2026

Pre-made threads, export to PDF or Markdown, no login or registration. Annotate it to death, and preserve your work over days and weeks. Whereof one cannot speak, thereof one must be silent, so go try it: foldabletractatus.aethermug.com Why I made it: aethermug.com/posts/tract

000

Marco Giancotti @marco-giancotti.bsky.social · 27/08/2026

You can click any technical term anywhere in the text to instantly highlight all other occurrences of the same term and its variants. Then you can optionally pin them all in one click: instant new thread you can read, save, and share.

100

Marco Giancotti @marco-giancotti.bsky.social · 27/08/2026

A set of controls lets you expand any statement to reveal its direct children, "pin" specific statements that you want to get back to, and add annotations with your own thoughts. Fold all again, and what's left is a custom "cross-section" of the book.

100

Marco Giancotti @marco-giancotti.bsky.social · 27/08/2026

So I made this Tractatus foldable, like a map. The book starts in its fully-folded state, only showing you the seven-statement root of the conceptual tree (or maybe I should call it a forest of seven trees). foldabletractatus.aethermug.com

100

Marco Giancotti @marco-giancotti.bsky.social · 27/08/2026

Wittgenstein's Tractatus is written like a tree. It's structurally very unbooklike. What is, on the surface, a single paragraph of seven assertions, is fractally expanded into five hundred and twenty-five points at different depths of detail, snowflake-like.

100

Marco Giancotti @marco-giancotti.bsky.social · 21/08/2026

We need a new term. AI slop = output of a model at the first or second try, shared without close checking or significant corrections ??? = output of a model after several turns of steering and corrections, accurately reflecting the user's intentions, preferences, and taste

000

Marco Giancotti @marco-giancotti.bsky.social · 18/08/2026

Claude's "load-bearings" and "worth your attentions" don't bother me that much, but this tendency to leave vestiges of the conversation in permanent docs is annoying!

000

Reposted by Marco Giancotti

Nothings Monstered @nothingsmonstrd.bsky.social · 16/08/2026

The Great Abyss — Frederick Judd Waugh, 1909

Painting of the surface of a dark green sea waves Rolling Along under clouds under a pale green sky trying to convey the darkness and mystery of the sea above the depths

211919

Marco Giancotti @marco-giancotti.bsky.social · 16/08/2026

The fact that some people seem to stop listening the instant they hear that AI was used in any way suggests cause all thinking.

020

Marco Giancotti @marco-giancotti.bsky.social · 16/08/2026

Correlation and causation do tend to be confused often.

121

Marco Giancotti @marco-giancotti.bsky.social · 13/08/2026

Treating AI like a kind of person makes thinking about these and other questions viscerally difficult. Think of AI as a lens instead. Read the full post: aethermug.com/posts/gen-a...

aethermug.com

GenAI Is a Lens into Humanity

not a person!

010

Marco Giancotti @marco-giancotti.bsky.social · 13/08/2026

Under this framing some hot controversies melt away. Is AI conscious? Not more than a looking glass has good taste for what to show you. Is AI slop more base than human output? Of course not, it is as human as it can possibly get, for better or for worse.

110

Marco Giancotti @marco-giancotti.bsky.social · 13/08/2026

The metaphor leads to some parallels with optical aberrations: bad focus → vague, generic output chromatic aberration → bias blind spots → unused knowledge distortions → the jagged edge artifacts → hallucinations We've refined glass for centuries. AI lenses are brand new and very rough.

110

Marco Giancotti @marco-giancotti.bsky.social · 13/08/2026

Training a model starts by building a semantic map of everything it was fed: novels, flame wars, papers, source code, cries for help, ads. You can't read that map yourself, so it comes bundled with an instrument. A prompt is just the coordinates you point it at.

110