Gemini Deep Research is so so good dude. This isn’t news - I’ve been using it pretty much daily for probably around 2 months at this point, but it just keeps striking me how amazing it is.

(I’ve used ChatGPT deep research too, in fact I paid £200 for a month of uber-pro or whatever it’s called back when DR was only available to that tier. It may be silly, but I think the “export to google docs” functionality of Gemini is a seriously killer feature, and I find I haven’t liked chatgpt’s reports as much. Also this may be dumb as hell but I’m like “eh, Google already has so much of my data, I don’t mind using Gemini a bunch”, whereas I feel kinda icky giving OpenAI my data)

I love the modern-ass workflow of reading a chapter of a book on my reMarkable and repeatedly scribbling “make a DR report!!” in the margins next to various things.

For example, in Frans Osinga’s book on John Boyd, he has brief sections in chapter 3 (which is about the zeitgeist Boyd was immersed in) for Karl Popper, Thomas Kuhn, and (first name) Polanyi (sorry for not remembering your first name Mr Polanyi. Thank you for the concept of tacit knowledge).

They’re good short primers, but very constrained by the necessity to keep the book from being an insanely digressive many-thousands-of-pages epic. So it leaves you with a bit of a sense of who e.g. Thomas Kuhn was and what his key contributions were, but not really enough for it to embed properly in your head.

So now for me learning if often this kind of insane recursive thing where a book chapter may spawn 10-30 deep research reports (e.g. one on Kuhn, one on Polanyi, Popper, the Ilya Prigogine, Ludwig von Bertalaffny (sp?), a primer on “the new physics”, a primer on “the epistemological revolution of the ~1960s”, and the cognitive revolution, and etc etc etc. And of course, each of these reports may lead me to make 1-5 more reports, e.g. the cognitive revolution one led to a big-ass rabbit-hole of stuff because it was also super relevant to learning about the history of AI (GOFAI, Herb Simon and Allen Newell and Dartmouth 1954 and RAND and LT/GSP and symbolic AI vs new approaches and AI winters and booms and John McCarthy & etc. Claude Shannon and Norbert Wiener. these are all topics that I can ramble about now, that I could engage in a conversation about. I’ve done a similar thing with the basics of nuclear too.

In a way, this feels insanely mad and inefficient, in another way it feels wildly efficient, like I can really rapidly (especially with Speechify and Anki) go down the exact correct rabbit-holes and hoover up a bunch of stuff real quick. Like e.g. idk 2-7 days of this (+ a week or so of flashcard reviewing) would lead to a genuine basic foundation in stuff that you previously knew truly nothing about. That’s wild!!!

So yeah, feeling grateful for Deep Research today.

It’s mad to think that this just didn’t exist like, what, 6 months ago?

I remember getting my first internal operations/knowledge management job, and not having read much actually stuff (instead being powered by my own taste/intuition), and there was a sense of “I should read some proper texts”, but also a sense of “oh god, where to begin, and how to ensure I’m getting a proper overview rather than reading a dry book on one tiny insignificant slice but I don’t know enough about the gestalt to know what to read, it’s all unknown unknowns”. Vs today I ran a report on “so hey, what actually is internal operations, what are the key knowledge centres, what are related fields? Like, I have a sense that systems thinking could be a really valuable thing to learn about for internal operations, is that a decent thing to learn next, or are there more obvious low-hanging fruit to tackle first? And I got a beautiful report, and of course maybe you can’t trust it all, but as a basic primer, the lack of friction and abundance of exciting rabbit-holes is just so insane.

The end! I love you Arch Linux ThinkPad. clicky clacky