Privacy & AI

No ads, no cookies¹, no Google Analytics, no LinkedIn Insights, no Meta Pixel, no Pinterest Tag, no Substack spam. No tracking of any kind.² No need, turns out, for a privacy policy. I'm solely responsible for all content on gwagner.com.

If you happen to navigate to a page with an audio or video clip, you may be asked to accept a third-party cookie (e.g. from NPR or YouTube), but that's it.

If you sign up for my mailing list, I will email you occasional updates sent directly from gw@gwagner.com. These emails will always be free. Nobody else has access to the list. Nobody else is writing these updates. That also means no AI.

No AI

If I didn't bother writing it, you shouldn't have to bother reading it.

No text on this site has been generated by AI. If I put my name on it, I wrote it. That includes my correspondence and all my social media posts, whether BlueSky, Mastodon, or LinkedIn.

I do work with plenty of co-authors. So yes, guilty: not every word is indeed written by me; lots are written by other humans. All know about my no-AI writing policy. Or perhaps better: I work with co-authors because of their expertise, or their writing abilities—not as a shortcut of sorts to add more words to my site.

Yes AI

There are plenty of reasons for AI in research and especially in coding. It helps remove some of the tedium, allowing for more creative thought. (Yes, that's the opposite from AI in writing, where AI removes the creative thought.)

No Claude Code, no coproduction.io, nor globalscope.io. (Ping me for the password; worth it.)

There are also good uses e.g. for random forest models and other tools. Even though everything seems to be called "AI" these days, such machine-learning algorithms are fundamentally different from large language models.

LLMs, too, have good uses, both for research, and for otherwise inconsequential tasks. (I was an early adopter of x.ai circa 2017(!), when the site was a semi-useful scheduling tool.)

Writing is not one of those inconsequential tasks. But now I'm going in circles. (Yes, editing, too, is a task that takes special expertise. Some of my best writing has benefited from excellent editors with whom I've had the fortune of working over the years.)

AI in Teaching

This is the presentation I showed at the beginning of my latest class, videos of Warren Buffett, Andrew Tate and all: AI.pptx. (Feel free to use and adapt with credit.)

There are, of course, good uses for AI in teaching: If you teach coding, a "no AI" policy will only go so far. Something similar goes for applied data analysis. Not teaching about AI in that case is educational malpractice.

In my Climate Risk class, I encourage students to use AI for their final group project presented in class. AI here, ideally, means tools like Claude Code, to help explore data sets and creative quantitative analyses that would otherwise be impossible in a 6-week course.

I am also currently experimenting with using "CAiSEY" in my courses, a platform that allows students to explore business school cases while chatting with a bot.

For Business and Climate Change, the most popular elective at Columbia Business School taken by 500 of 800 MBAs each year, I give students the option of using either CAiSEY and/or shorter Financial Times mini case studies. (The jury is still out, but it seems like most students prefer the FT mini case studies: why spend 15 minutes chatting with a bot to 'explore' a 15,000-word case study, when you can just read 800 words?)

In the end, teaching and learning how to think means just that. Andrew Tate may disagree, but there are no shortcuts to readings and writing.

¹No cookies and dark mode

The no-cookies policy, turns out, does come with some challenges. Take the dark-mode toggle on the top right of each page. It works just as advertised without cookies, with one exception: If you have set a default on your browser, and you toggle away from it, you may now see a small flash every time you navigate to a new page.

The smoothest fix: a cookie. The browser reads those on the server-side before loading the page. There apparently is no way to avoid that flash without a cookie. Tradeoffs.

If you do know of a fix, please email web@gwagner.com.

²No cookies and usage statistics

I use plausible.io to count visitors to this site. Their data policy: "We don’t use cookies, we don’t generate persistent identifiers and we don’t collect or store personal data that can be used to identify individuals. All data is aggregated."

The fun part: Plausible tells me that this page is among the most viewed on my site: ~1k visitors/day. Time on page: 1 minute. Scroll depth: 80%. (Hi! 🙋) Bounce rate: <20%. My other preoccupations: gwagner.com/preoccupations.