iiinova — where your digital self keeps evolving

John Laudun @laudun.bsky.social · 07/10/2026

Neither Trump nor Musk have any sense of how science works. Musk doesn't even understand how science fiction works. It's don't think any of us were prepared for not only the Republicans to look to destroy science but also the tech bros.

030

John Laudun @laudun.bsky.social · 07/10/2026

Except for the Baley-Olivaw novels. And speaking of human-AI pairings, there's always Eliasz and Paladin in Newitz' "Autonomous."

100

John Laudun @laudun.bsky.social · 07/10/2026

Dear @jstor.bsky.social: why is there no persistent journal-limit setting? It's absolutely crazy-making that this isn't an option.

000

John Laudun @laudun.bsky.social · 30/09/2026

It’s even better when they provide the data to you and then ask you to put it in the form. Faculty here do that every year. They call it a workload document, but it’s really a load of work for which there is no reward.

011

John Laudun @laudun.bsky.social · 23/09/2026

I teach a class on narrative games. I don't need students to become either game devs, critics, or scholars. I want them to continue be, or become, live-long, and conscientious, gamers. If I can also get them hosting games with friends, then my job is done.

010

Reposted by John Laudun

angela zhou @angelamczhou.bsky.social · 09/09/2026

blog: I argue that we need auditable AI now, because current AI developments threaten auditability of AI systems for ordinary regulatory compliance, and we can't figure out “what to do next about AI” without common knowledge about AI (mis)alignment. separatedhyperplane.substack.com/p/we-need-au...

separatedhyperplane.substack.com

We need Auditable AI to figure out “what to do next about AI”

Mandatory, not voluntary, third-party governance, accountability, and transparency

1306

John Laudun @laudun.bsky.social · 06/09/2026

I recently came across @typst.app, a markup language that successfully splits the difference between the ease of writing markdown and the power of LaTeX. I was up and running with lecture notes within an hour or two, having come up with my own template, working with an LLM to brainstorm dimensions.

050

John Laudun @laudun.bsky.social · 31/07/2026

I have to admit that Gemma 4 (E4B Instruct) comes awfully close to frontier models on most tasks for which I seek LLM input. It runs comfortably on an M5 MacBook Air with the stock 16GB of RAM.

010

John Laudun @laudun.bsky.social · 28/07/2026

That's a lovely site you've put together! I need to follow your example and place larger materials in Knowledge Commons -- I'll admit to being lazy here. Right now, everything is sitting in the GitHub repo.

010

John Laudun @laudun.bsky.social · 20/07/2026

Way to be customer-oriented, @nationalcarrental.bsky.social: a 169-page user agreement?

000

John Laudun @laudun.bsky.social · 18/07/2026

FTR, Qwen3 4B is better in this regard, though it did not know how to read/understand "d66." On the other hand it did distinguish, and label, the difference between d6 x d6 vs d66 as being multiplication versus concatenation, which was nice.

000

John Laudun @laudun.bsky.social · 18/07/2026

My guess is that Llama doesn't understand dice roll notation. Its response was "2d6 / d6 x d6: 36 d66: 66." Claude got it right: 11, 18, and 36.

000

John Laudun @laudun.bsky.social · 18/07/2026

I want to believe that local LLMs can be as productive as so-called frontier models, but, dang, did Llama 3.1 fail hard this morning when I prompted it to “provide the total number of possible unique outcomes for the following dice rols: 2d6, d6 x d6, and d66.”

200

John Laudun @laudun.bsky.social · 15/07/2026

The amount of cross-posting the modern internet wants is crazy-making: Academia.edu, @researchgate.bsky.social, @linkedin.com, @orcid.org … The same goes across multiple domains. I really long for the federated conception of the internet. Even @doctorow.pluralistic.net ends up cross-posting!

000

John Laudun @laudun.bsky.social · 11/07/2026

I honestly don't think it should be this hard to view recent transactions on an Apple card from your Mac.

020

John Laudun @laudun.bsky.social · 04/07/2026

FTR, Qwen 4B does a better, and faster, job than Qwen 3.5 9B, which seems to get stuck in verbose loops—3 minutes into another parse and it was still talking to itself. Qwen 3 4B also bested Claude by adding the DOI.

000

John Laudun @laudun.bsky.social · 04/07/2026

Qwen 3 does a nice job of taking the title page from a JSTOR PDF and turning the citation information into a bibtex file you can import readily into your reference manager. (I use @zotero.org.) That saves some serious time. (Claude also does this easily, FTR, Qwen 3 4B runs easily on my laptop!)

120

John Laudun @laudun.bsky.social · 03/07/2026

This is very cool. I wish I had this a few months ago when I was trying to run verbatim searches on phrases pulled from LLMs — essentially using the "problem" of memorization as a feature. I'm going to try it out to see if I can verify results against the Common Crawl. Very exciting!

000

John Laudun @laudun.bsky.social · 27/06/2026

I went through some of Frank Dobie's archives at UT-Austin's Ransom Center: I was surprised to find that he took an astonishing amount of notes on the back of grade report cards. Sturdy stuff and readily available. I imagine a lot of creative thinking occurred on the back of such things.

000

Reposted by John Laudun

Operation Democracy ✊🏼 Free DC! @democracyop.bsky.social · 20/06/2026

Wow. I mean, this shit isn’t really surprising coming from Texas A&M, but it’s still disturbing.

17922

John Laudun @laudun.bsky.social · 19/06/2026

Yes, please! I'll DM you my mailing address! So awesome!

000

John Laudun @laudun.bsky.social · 19/06/2026

Apparently @americanairlines.bsky.social was okay flying a plane into Cincinnati with the incorrect safety cards but not out again. The upshot is my wife is now stuck in Cincinnati because that delay meant the flight is now delayed by weather.

000

John Laudun @laudun.bsky.social · 07/06/2026

As the son of an architect, the basis of the metaphor was something I saw ignored at the peril of the building being "born." Also I am tickled when a phrase or word I use comes into common usage. (And saddened a bit when it passes out!)

020

John Laudun @laudun.bsky.social · 02/06/2026

Wow, just wow. As the parent of a recent college graduate, I worry about the lost years both for their cohort but also for businesses who buy the misapplication/hype because they think they don’t have time to think it through. The divot in the economy is already emergent. It’s a perfect storm.

000

Reposted by John Laudun

AAUP @aaup.org · 27/05/2026

“What we’re highlighting is that there’s a broad number of faculty who have significant, significant concerns about their ability to teach, research, do public-facing work without punishment or restraint.” —Daniel Martinez HoSang, Yale AAUP President @yaleaaup.bsky.social @aftct.bsky.social

insideinvestigator.org

Survey finds Yale faculty self-censoring, fear political consequences

Read More at Inside Investigator >>

02011

John Laudun @laudun.bsky.social · 28/05/2026

I got a manuscript from JAF: the proverb scholar, R1, wanted more about proverbs. The LLM scholar, R2, wanted more about LLMs.

010

John Laudun @laudun.bsky.social · 27/05/2026

I am not that smart, but somehow I find it easier to set up and use API keys for LLMs in Python than in any of the GUI apps I have downloaded and tried. I suspect it might be that most apps want you to subscribe to their infrastructure as opposed to using stuff to which you already subscribe. Sigh.

000

John Laudun @laudun.bsky.social · 11/05/2026

This. I forget where I read it or heard it, but I recently came across the claim that da Vinci was not a scientist: he wrote in a secret code and kept his discoveries secret. That is not science. That's just business. Science is built on the idea of engagement.

010

John Laudun @laudun.bsky.social · 11/05/2026

And I imagine that applies to a lot of people in the humanities, so the return to CW, as you put it, makes a certain sense. It's the resources available to them. It's also the case of how much time are you willing to give up to educate students on just AI as opposed to what the course is about? 2/2

020

John Laudun @laudun.bsky.social · 11/05/2026

Well said. In the English department where I work alongside literary scholars, compositionists, linguists, and creative writers, it is largely like watching a flock of birds scatter to the wind when a predator appears. To be fair, almost all of them are underpaid and under-resourced. 1 / 2

110

John Laudun @laudun.bsky.social · 03/05/2026

Slightly diminish a book: Starship Auditors

010

John Laudun @laudun.bsky.social · 25/04/2026

For those who hadn't already seen this: Quanta has a nice oral history, or at least reporting with lengthy quotations from people on the ground, about how LLMs took over NLP.

quantamagazine.org

When ChatGPT Broke an Entire Field: An Oral History | Quanta Magazine

Researchers in “natural language processing” tried to tame human language. Then came the transformer.

011

John Laudun @laudun.bsky.social · 25/04/2026

These kinds of mandates are coming—it's another way of making it easier to police, I mean oversee, I mean manage, I mean enhance student learning. Canvas is a for-profit LMS, but they are all perfectly awful to work with(in). (Maybe Sakai is not? I'd like it if one of these things was decent.)

000

John Laudun @laudun.bsky.social · 11/04/2026

I get the Vance move to support Orban, both want to players in the threatening global oligarchy, happy to get scraps from the high table. But what's up with Albania making Eric Adams a citizen? That came out of nowhere.

000

John Laudun @laudun.bsky.social · 10/02/2026

Sometimes I agree with Reddit.

000

John Laudun @laudun.bsky.social · 03/02/2026

I know how long and hard you've worked on this. On behalf of users yet to enjoy KC Works, I thank you.

010

John Laudun @laudun.bsky.social · 31/01/2026

In the complete opposite direction of Haxen is Loner RPG. Honestly, I think, if you keep notes about characters in the margins, Four against Darkness is probably playable on an airplane tray table. Maybe? (I like the game because it's simple and it doesn't take up a lot of room while playing.)

110

John Laudun @laudun.bsky.social · 31/01/2026

Well so much for replacing humans with AI because AI never sleeps …

000

John Laudun @laudun.bsky.social · 29/01/2026

No, no it's not. Clueless leaders, managers really, love dashboards, it's information as pre-chewed food. I don't know higher ed is going to work its way out of the corporatization foisted on it by conservatives, but we are going to need to find new footholds for engagement.

000

John Laudun @laudun.bsky.social · 29/01/2026

But this isn't surprising given the MBA-ification of university administrations, no? Consultants are the gift that gives to both execs and consultants: if/when things go wrong, execs blame consultants for plan and consultants blame execs for not following the plan.

100

John Laudun @laudun.bsky.social · 29/01/2026

When @dmimno.bsky.social crosses over, he crosses over big:

010

John Laudun @laudun.bsky.social · 22/01/2026

Sometimes you find yourself caught in a loop…

000