iiinova — where your digital self keeps evolving

Kevin A. Bryan @afinetheorem.bsky.social · 24/08/2026

How should we modify our university teaching given AI? It's not enough to just "stop cheating" - we also should use these tools to improve our teaching and evaluation. Here's 8 rules I've been using in my U Toronto courses and with colleagues around the world. kevinbryanecon.com/teachwithai....

030

Kevin A. Bryan @afinetheorem.bsky.social · 15/12/2025

(Also, we set a new one-class weekly record for use this week! This is a big Principles class at a large public univ, but 39k uses of Intelligent Quiz (an AI-driven adaptive quiz) + 3k direct questions in one week. Wild. Across term- with exams + your good setup- 40-80 uses/student is normal.)

000

Kevin A. Bryan @afinetheorem.bsky.social · 15/12/2025

Something I didn't expect that we can see in www.alldayta.com data: study effort is so correlated with exams. Across courses, if use on Finals week is "1", relative use on midterm week is .7, and on non-exam weeks .05-.1. For courses with no exam, use always sticks at the lower level.

141

Kevin A. Bryan @afinetheorem.bsky.social · 06/10/2025

For students, *I don't care* if they get right answers. Why? If wrong, they have to explain to AI what they were thinking before moving on. Cheating doesn't save time b/c I don't grade on correctness, just whether you work through the quiz! Try it here app.alldayta.com/university-o... 3/4

100

Kevin A. Bryan @afinetheorem.bsky.social · 06/10/2025

Toss my lecture audio, slides, handouts into a module. A crazy AI workflow pulls out learning goals. AI then spins up questions (you can approve or not) from your docs plus context on why students might get them wrong. 2/4

100

Kevin A. Bryan @afinetheorem.bsky.social · 06/10/2025

I know it's my company, but All Day TA AI-driven quizzes are so good. Student cheat on all take-home work. How do you get them to learn? Do even better than we used to by having them learn *as they do low-staked hw*. Here's use just this week in a Texas univ course - students really use this. 1/4

150

Kevin A. Bryan @afinetheorem.bsky.social · 17/09/2025

New class on Progress starting tomorrow - I'm amped! Trying to put some rigor from economics, economic history, and philosophy on a topic very much in the air. It will be awesome. (And first class running slides in my all html browser-based slideshow program - details soon!) 1/2

160

Kevin A. Bryan @afinetheorem.bsky.social · 08/09/2025

Perhaps of interest to folks with social science PhD programs: at Rotman, we added an experimental 3 session "tech stack" training in addition to the math boot camp. My lecture was "how to do reproducible, open, quick research", aka version control, LaTeX, AI. 1/2 kevinbryanecon.com/techstack.html

191

Kevin A. Bryan @afinetheorem.bsky.social · 05/09/2025

PS - An awesome dev of ours was testing the featureand told me "I got it wrong on purpose at first for testing, but then forgot to divide by 2 for expected value until the system brought me there!" Exactly. Imagine this help for the student, and then summed up & reported back to you for each hw!

010

Kevin A. Bryan @afinetheorem.bsky.social · 05/09/2025

When the students do their assignment, and get a question wrong, the AI forces them to explain their logic, then uses your lectures, handouts, and so on to try to correct mistakes. We then use another AI system to report back to you precisely where students have been going wrong *and why*. 4/6

100

Kevin A. Bryan @afinetheorem.bsky.social · 05/09/2025

That's exactly what we built. Our system already interpreted the learning goals of your course, topic by topic. For question banks, we propose these using our AI, and once you edit and approve, we spin up question banks of varying difficulty. You can manually add, edit or kill these, of course. 3/6

100

Kevin A. Bryan @afinetheorem.bsky.social · 05/09/2025

I know lots of skepticism about AI here, but let me show you something we put out which I think is a huge improvement for university assignments. This is "Intelligent Quiz", a feature on All Day TA (www.alldayta.com). Assignments now have tons of cheating + little feedback to us or the students. 1/9

1112

Kevin A. Bryan @afinetheorem.bsky.social · 02/09/2025

I'd separate Trump from general public trust (incl. outside US); some unique features there. Even there, though, note that Senate NIH appropriations are up for 2026. I think the medical-related field with the trust deficit is public health, for reasons 1-8. But also - all this below was a mistake.

110

Kevin A. Bryan @afinetheorem.bsky.social · 01/09/2025

Eight rules to regain trust in universities. All can be done today. They're absolutely not the core beliefs about the role of the university for some faculty right now, but are essential if we're to contribute to the trusted production and diffusion of knowledge. kevinbryanecon.com/trust.html

83155

Kevin A. Bryan @afinetheorem.bsky.social · 25/08/2025

Getting "big" things done and maintaining existing things are not the same. But why? What causes path dependence? What role do liberty, or particular individuals, or cities, or science, or incentives play? 2/3

120

Kevin A. Bryan @afinetheorem.bsky.social · 25/08/2025

Incredibly excited for my brand new class at Rotman this fall: Progress! Econ history + theory + history of thought + philosophy on why rare orgs at rare times in rare places accomplish new things. Trying to put rigor onto an idea that is very much in the air. kevinbryanecon.com/Bryan-Progre... 1/3

1164

Kevin A. Bryan @afinetheorem.bsky.social · 08/08/2025

To create the XML, just drag and drop pdf or latex/bib into the browser, wait a minute, and you have things all set. Occasionally a couple minor manual edits needed, and you need to place your figures in the same folder. Otherwise, that's it - just upload to your website! 4/5

110

Kevin A. Bryan @afinetheorem.bsky.social · 08/08/2025

Because the entire thing is just raw XML, it's easy to build AI tools on top. I have two: an auto "layman's summary", and a really good AI chat with sourcing. Here, it does a calc I don't explicitly do in the paper. And it's using only free models! 3/5

100

Kevin A. Bryan @afinetheorem.bsky.social · 08/08/2025

This code, which runs on one .html file, takes your pdf (or latex/bib), rewrites the whole thing in a pure text XML format, adds in-line references and footnotes, zoomable images, and retains clickable navigations. Plus prettier, and adapts view for phones. But also... 2/5

200

Kevin A. Bryan @afinetheorem.bsky.social · 08/08/2025

I've built a bunch of tools this summer to move my whole workflow to using things that are "pure text" so I can use AI on top. Here's the first: a modern paper reader. PDFs suck. Ugly, terribly interface, big files, fixed. I want papers that are adaptive, interactive, pretty. 1/5

281

Kevin A. Bryan @afinetheorem.bsky.social · 09/07/2025

I'm worried about universities. We *have to* be seen as neutral, truth-seeking, unbiased, rigorous, or public support will continue to fall. Floating around today was Columbia's core curriculum "Contemporary Civ". Here's the entire post-1865 reading list www.college.columbia.edu/core-curricu... 1/10

161

Kevin A. Bryan @afinetheorem.bsky.social · 12/06/2025

Editing our writing is a common academic task. I *hate* how spell/grammar checks annoy me by thinking proper nouns or LaTeX code are misspellings (yes, I want `` ''). I also want *style* tips like a good editor would give me. Problem solved. 1/5

1155

Kevin A. Bryan @afinetheorem.bsky.social · 23/05/2025

Here is Harvard. (Many) more faculty "very liberal" than anything from moderate to right. And 23% "moderate or conservative" almost all hard science. How is this sustainable? CEOs, industry researchers, top doctors, even Harvard students (2nd image): all smart, look *nothing* like this. 5/6

240

Kevin A. Bryan @afinetheorem.bsky.social · 12/05/2025

80 images (plus a holdout set), all not online, from private collections, some 20 years old, globally distributed, all metadata removed. 95% have no text or road markings of any kind. 2/x

100

Kevin A. Bryan @afinetheorem.bsky.social · 12/05/2025

"Where in the world is this place" is a surprising skill of multimodal LLMs. But as Geoguessr fans know, road markings, words on signs, etc. make it much easier. I made my own benchmark to fix this. 1/x kevinbryanecon.com/HardGeoBench/

262

Kevin A. Bryan @afinetheorem.bsky.social · 04/05/2025

In the Star today (Canada's biggest paper) on why Canada can't redirect trade away from the US without making the country much poorer, why the (many) attempts to do this historically failed, and what to do instead: www.thestar.com/news/insight...

000

Kevin A. Bryan @afinetheorem.bsky.social · 31/03/2025

A good measure of Abundance/Progress: how upset are you about reg restrictions slowing self-driving rollout? Self-drive cars drop serious crashes by 80%+, even though remainder almost entirely are caused by others. Waymo's swerve away from even many of these! www.understandingai.org/p/human-driv...

141

Kevin A. Bryan @afinetheorem.bsky.social · 20/03/2025

I cut video early due to private info in the final pdf, but what it gives you is a dozens of pages pdf identical to what I'd have spent a couple hours doing manually. Also: 1st project where I didn't manually write a line of code. 3 LLMs, I PRed each part & asked for corrections.

130

Kevin A. Bryan @afinetheorem.bsky.social · 20/03/2025

A fun AI demo. Spent five hours coding this - let's file expense reports in literally one minute, incl. cover spreadsheet, signing, converting currencies, ripping info out of underlying receipts, and reordering all receipts in one big pdf in the order the finance office likes it.

151

Kevin A. Bryan @afinetheorem.bsky.social · 18/03/2025

Really nice article in the Financial Times on @alldayta.bsky.social / alldayta.com. The team is adding lots of new features - better control of how direct/pedagogical system should be + more accuracy on harder math, rolled out this week. Reach out for your spring courses! www.ft.com/content/daa0...

041

Kevin A. Bryan @afinetheorem.bsky.social · 15/03/2025

Today is "housing costs are a regulation problem": family in Carmel worries about $2,000,000 houses. Buys a 5 acre lot for $400,000. My detached house is on an 1800 sq ft lot. 5 acres = 217k sq ft. So land cost for my lot in Carmel? $3300. Everything else is reg burden.

010

Kevin A. Bryan @afinetheorem.bsky.social · 10/03/2025

Ah, March 10. Mar 11 was "Tom Hanks and NBA cancels day". At my office, we had a buffet that night! I checked my thinking - honestly, not bad! "Did best we could given uncertainty" is cover for bad policy. PS - Still v serious post Covid trends, like decline in social interaction.

050

Kevin A. Bryan @afinetheorem.bsky.social · 02/03/2025

For obvious reasons, Canada has a surge of patriotism. That said, Build Canada folks are right: this election, have to solve economic stagnation and housing affordability. Real GDP per capita is lower than 5 years ago, and unchanged since I moved here 11 years ago. And during a resource boom!

170

Kevin A. Bryan @afinetheorem.bsky.social · 28/02/2025

From the Collison brothers, Irish legends who founded Stripe, and who quote O-Ring theory and Chad Jones growth papers off the top of their head: a wild piece of survey evidence from European founders. A strong European economy benefits the world, yet policy so counterproductive.

152

Kevin A. Bryan @afinetheorem.bsky.social · 25/02/2025

A fun (and surprisingly difficult!) new feature for @alldayta.bsky.app: don't answer homework questions! See below, from my venture finance class. It won't answer homework, even when the student is tricky (asking indirectly in French), but still answers related Qs students may have while studying.

120

Kevin A. Bryan @afinetheorem.bsky.social · 19/02/2025

Related to the forthcoming development research scandal, the fundamental problem (as in social psych) was this: THERE IS NO SUCH THING AS MAJOR CHANGES IN IMPORTANT THINGS FROM TINY INTERVENTIONS! *Every* paper making claims like this are wrong- you don't even need to read them.

3104

Kevin A. Bryan @afinetheorem.bsky.social · 07/02/2025

California: center of econ universe run by incompetence, part 1562: one block from Meta hq (gray in background) and the "Zucktown" urbanist plot FB has tried to build for 8 years, stymied by NIMBY regulation. I can guess why housing is expensive. This should be cyberpunk paradise

050

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

Under the hood, this is full o3 finetuned on how to browse the internet, pull data, do calcs w/ python, draw embedded images/plots. o3 is very impressive, even compared to what we can all use on o3-mini-high. On "Humanity's Last Exam", o3 Deep Research is 26%. Current state of the art is 9%. 5/x

2111

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

It wasn't exactly what I wanted, and is quite far away from even one novel result (basically, it gave me a slight extension of Scotchmer-Green). But again, the path is very clear, and I've definitely used frontier models to help prove theorems already. kevinbryanecon.com/o3Innovation... 4/x

1120

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

I tried to spin up a theory paper as well. On my guidance on the problem, it pulled a few dozen papers, scanned them, then tried to write a model in line with what I said. 3/x

190

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

Related to tariff nuttiness, what if you wanted to read about the econ of 1890 McKinley tariff, drawing on modern trade theory? I asked Deep Research to spin up a draft, w/ a couple paragraphs of guidance, in Latex, w/ citations. This is one-shot, 10m of think. kevinbryanecon.com/o3McKinley.pdf 2/x

2120

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

The new OpenAI model announced today is quite wild. It is essentially Google's Deep Research idea with multistep reasoning, web search, *and* the o3 model underneath (as far as I know). It sometimes takes a half hour to answer. Let me show you an example. 1/x

68328

Kevin A. Bryan @afinetheorem.bsky.social · 31/01/2025

o1 and o3-mini both win on "evaluate this section of an econ paper. any errors of corrections needed?" Pasted in a few pages of a real paper with a mistake I found. No AI I'd tried found it - o1 and o3-mini did; R1, Gemini Thinking, Claude 3.5, 4o, o1-preview failed. Impressive!

041

Kevin A. Bryan @afinetheorem.bsky.social · 31/01/2025

o3-mini-low (let alone o3-mini-high) getting my favorite "somewhat hard problem that isn't online" totally correct. Actually, the proof it gives is better than what o1 and Deepseek and Google 2.0 Flash Thinking were giving. Very impressive for the price.

140

Kevin A. Bryan @afinetheorem.bsky.social · 20/01/2025

Also, if you're wondering, yes, of course, it's super politically gated. But open source, and we have seen it's pretty easy to jailbreak these models...

040

Kevin A. Bryan @afinetheorem.bsky.social · 19/01/2025

I'll doublecheck the official NBER call to make sure this info is included. I know at least when we've advertised things we've tried to mention this as many times as we can, e.g., below.

000

Kevin A. Bryan @afinetheorem.bsky.social · 13/01/2025

And see "Create Note" there? A new feature we just rolled out. Get a good Q after class, or don't like how the AI answered? Add in a note! You can even add to older Qs under 'Weekly Summaries' & the AI updates going forward. (Here, let's clarify that skunkworks are Lockheed) 4/x

100

Kevin A. Bryan @afinetheorem.bsky.social · 13/01/2025

The AI 1) won't answer anything unrelated to your course, 2) will defer to your documents as much as possible, 3) won't draw on 'inactive' documents, 4) will give less weight to 'optional' documents. 3/x

100

Kevin A. Bryan @afinetheorem.bsky.social · 13/01/2025

Excited to see All Day TA (@alldayta.bsky.social - www.alldayta.com), our in-class AI, in 100s of universities for the spring term! A quick thread of how I would recommend you set it up. Tldr: you won't need more than 30 minutes. 1/x

1103

Kevin A. Bryan @afinetheorem.bsky.social · 26/12/2024

Honestly, can someone at Princeton talk to KK? This is exact same kind of language he used about me, and in response to nothing. I don't get it, but "your well known professor is an asshole online" isn't good for the university. His colleagues are "termites"?! And he is a jerk going way beyond AI...

3150