iiinova — where your digital self keeps evolving

John Gallagher @johnrgallagher.bsky.social · 10h

The AI "slow down" is related to burn rates of training new models and is an economic play for IPOs.

Reposted by John Gallagher

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 16h

In our Nature paper, we introduce the first superhuman Stratego AI, which we built using general techniques that we developed for RL & test-time compute under imperfect information. www.nature.com/articles/s41...

nature.com

Scalable decision-making for games of imperfect information - Nature

Ataraxos, an AI for the board wargame Stratego, establishes a design pattern for reinforcement learning and search that is effective under large amounts of hidden information, a longstanding desiderat...

1019847

John Gallagher @johnrgallagher.bsky.social · 14h

Anthropic and OpenAI are going to be a massive asteroid that hits the US economy.

000

John Gallagher @johnrgallagher.bsky.social · 17h

The GPT videos are literally just templates. Watch 4 of them and you'll see the edges immediately.

100

Reposted by John Gallagher

John Gallagher @johnrgallagher.bsky.social · 30/09/2026

AI token costs are going to SKYROCKET after IPOs to the point that it's eyewatering. In order to break even, Anthropic would need to make 400 billion a year for 5 years (they made 4.6 this year). THE MATH DOES NOT MAKE SENSE. www.reuters.com/business/fin...

economictimes.indiatimes.com

Anthropic IPO explained: $2 trillion valuation, ballooning net loss, AI risks and other details in leaked IPO papers

Anthropic is preparing for a potential IPO as its revenue surges, but losses and infrastructure commitments balloon. The AI company has also warned of existential risks from advanced models and highli...

152

John Gallagher @johnrgallagher.bsky.social · 30/09/2026

The Atlanta Hawks are killing it.

000

John Gallagher @johnrgallagher.bsky.social · 30/09/2026

Retrieval practices are the best practice for learning and writing is the number 1 retrieval practice.

020

John Gallagher @johnrgallagher.bsky.social · 30/09/2026

"maybe you should lay off the raspberry and brie on sourdough" ~my doctor

200

John Gallagher @johnrgallagher.bsky.social · 28/09/2026

Younger offspring: What talent do you want? Me: to control time, reality, and the knowledge to control those two things Older: They're fairy powers. Like water or flight. Me: Water. Younger: what is color is your outfit? Me: purple & black... Younger: No it's blue. That's the color you have.

000

John Gallagher @johnrgallagher.bsky.social · 27/09/2026

I got asked in an interview once, "What do you want on your tombstone?" It was a office job.

030

John Gallagher @johnrgallagher.bsky.social · 27/09/2026

THIS IS LITERALLY HOW INSURANCE WORKS

030

John Gallagher @johnrgallagher.bsky.social · 27/09/2026

I have never seen the movie the Holiday. Been married for over a decade and my spouse today found this out and was outraged.

110

Reposted by John Gallagher

Matthew Guzdial @matthewguz.bsky.social · 26/09/2026

They’re so, so desperate for large scale real world data from consumer’s perspectives. They will do anything short of just paying people to get it.

0102

John Gallagher @johnrgallagher.bsky.social · 26/09/2026

Why a person even want to produce 200 papers in a year?

030

John Gallagher @johnrgallagher.bsky.social · 26/09/2026

Five hours later, inbox (and tasks in those emails) zero.

020

John Gallagher @johnrgallagher.bsky.social · 26/09/2026

Fifth grader made breakfast sandwiches for our family. Added smoked gouda to the eggs, cooked some pepperoni, add them on top of toasted sourdough. This is eudaimonia.

040

John Gallagher @johnrgallagher.bsky.social · 26/09/2026

A) I know this is bait B) THE EXPANSE

250

John Gallagher @johnrgallagher.bsky.social · 26/09/2026

absolute banger. "Nothing that happened in September makes understanding unavailable. It makes it, for the first time, optional: a theorem can now be certified without anyone’s having understood it."

000

John Gallagher @johnrgallagher.bsky.social · 23/09/2026

Free study idea: measure the participation inequality of bots and non-human content. Easy use of the gini coefficient. I'm curious how the participant inequality compares between human and synthetic content.

010

John Gallagher @johnrgallagher.bsky.social · 23/09/2026

Or the sad hope they read the book. Even if it’s not out yet.

010

John Gallagher @johnrgallagher.bsky.social · 23/09/2026

Thanks to @mikertrice.bsky.social and everyone at MIT for having me! Disclaimer: Equations are not mine. Lol!

050

Reposted by John Gallagher

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 22/09/2026

Now that LLMs are good at math, we can solve other long-standing verifiable problems like the law, institution building, community, and management

816111

John Gallagher @johnrgallagher.bsky.social · 21/09/2026

Are you in and around Boston this Tuesday? Come see me at MIT! I’m presenting findings about ways students perceive their own writing skills as compared to AI. And stick around for a framework on writing better than AI (and your past self)! cmsw.mit.edu/event/underg...

cmsw.mit.edu

John Gallagher: "Students still think they write better than the machines: A study of undergrad self-perceptions of Generative AI and writing skills"

Gallagher discusses research that shows GenAI erodes writing ability over time by making us less persistent.

282

John Gallagher @johnrgallagher.bsky.social · 21/09/2026

It’s very hard to fathom but we have intelligence without consciousness which our science fiction hasn’t prepared us for. Intentionless intelligence is weird.

6509

John Gallagher @johnrgallagher.bsky.social · 20/09/2026

Free preview of AI Through the Experts’ Eyes! (introduction) upittpress.org/wp-content/u...

upittpress.org

173

Reposted by John Gallagher

Donnie Johnson Sackey @donniejsackey.bsky.social · 20/09/2026

CCCC Leadership Response to NCTE’s De-Chartering of CCCC. Please read and sign. theorganizationformerlyknownascccc.com

12732

John Gallagher @johnrgallagher.bsky.social · 19/09/2026

You cannot build best practices from descriptive data.

000

Reposted by John Gallagher

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 18/09/2026

Weird request; we've had a fairly big breakthrough that we've been mostly keeping a lid on for the past year. We're looking to talk to some science / AI writers about this and I know a lot of y'all are out here on bsky...

815732

John Gallagher @johnrgallagher.bsky.social · 18/09/2026

I'm convinced of a solution: giving up doomscrolling and social media.

040

Reposted by John Gallagher

Gautam Kamath @gautamkamath.com · 16/09/2026

Nihar Shah did a heroic experiment for TMLR: he spent 20-25 hours over two weeks interviewing authors of seemingly low-quality submissions about their own papers. He confirmed what we all suspected: people submitting these papers have *no idea* what is going on in them.

5276110

John Gallagher @johnrgallagher.bsky.social · 16/09/2026

BJ Fogg’s research group studied how to make computers more “persuasive” in the 2000s. His students went on to found little known companies such as Instagram using those principles. Zuck enrolled in his classes. it’s all DESIGNED to be addictive.

110

John Gallagher @johnrgallagher.bsky.social · 16/09/2026

Here is where I’m actually pro-AI: I think they’re n-dimensional producers, which is objectively non human. And that’s fascinating.

000

John Gallagher @johnrgallagher.bsky.social · 16/09/2026

My blog posts have now been responsible for multiple invited talks and the blog posts have been translated into multiple languages. Seems like I just should this instead of fight peer reviewer comments.

010

John Gallagher @johnrgallagher.bsky.social · 16/09/2026

“Do I think this writing is any good?” Is still the best metric.

000

John Gallagher @johnrgallagher.bsky.social · 16/09/2026

I made baked cinnamon apples for my kids this evening and blew their minds it was both this easy and it somehow turns into Apple pie.

000

John Gallagher @johnrgallagher.bsky.social · 14/09/2026

Well the fund-raising strategy of "AI will kill us" seems to be working

000

Reposted by John Gallagher

Maria Leonor Pacheco @mlpacheco.bsky.social · 10/09/2026

Come join us in Boulder! Feel free to get in touch if you have any questions. jobs.colorado.edu/jobs/JobDeta...

jobs.colorado.edu

Tenure-Track Faculty in Artificial Intelligence

095

John Gallagher @johnrgallagher.bsky.social · 13/09/2026

Apple is going to collapse the AI market when they put an LLM into their devices. And their new vapor cooling system is the step in that direction. I'm telling you: OpenAI and Anthropic are going to be Yahoo and AOL of the 2020s and early 2030s. LLMs are going to be like browsers.

001

Reposted by John Gallagher

Andrew Zolides, Ph.D. @zolides.bsky.social · 13/09/2026

I just taught the Dot Com boom and bust on Friday, and boy was it a timely discussion given all this ‘news’ of delayed IPOs

012

John Gallagher @johnrgallagher.bsky.social · 13/09/2026

The reason openai and Anthropic are “slowing down” is because the stock market can’t handle their IPO. It’s nakedly obvious if we check the stock market once week.

150

John Gallagher @johnrgallagher.bsky.social · 13/09/2026

In the AI unit in my science and literature class (why am I teaching literature again? Who knows) we are reading Klara and Sun + The Mountain in the Sea. Quite interesting to read them in light of current events. At least I think students are reacting well!

030