Late-Jan Review

Reviewing progress since Lydia-2026, setting new goals!

Let’s review the most recent three weeks.

Good parts

• I met a bunch of new people!

  • Esp. good was ppl from internet

  • & highly filtered events

• I wrote 16 blog posts (& many Notes):

  • 4 Math

    • 1 Metric Spaces mega-post

    • 2 Probability posts

    • 1 meta: “Notes on ‘When We Cease…’”

  • 3 Macrostrategy

    • The Good-Faith Actor Principle

    • On Redistribution

    • We need to make ourselves people the models can…

  • 2 Econ of AI

    • Why falling labor share != falling employment

    • Response to Decker re monopoly v. market competition

  • 1 technical ML

    • Model AI decision-making through Markov chains…

  • 3 ‘Living better’

    • Against commentary as full-time occupation

    • I live in interest-space

    • A tiny amount of force applied in just the right place

  • 1 tiny research-adj

    • Bitter Lesson Eats SPOwML

  • 1 fun-CS / algorithms

    • Joint Optimization at Lunch

  • 1 fun-bio

    • Stereocilia + mechanotransduction

• I edited a paper to completion, hosted vibecoding soc x2, submitted main summer applications, launched a Disc server,

• Read a book cover-to-cover

  • When we cease to understand the world

• Sampled five books:

  • Society of Mind (planning to continue)

  • Half of a Yellow Sun (preferring science to society)

  • Summa Technologiae (just started, seems rly good)

  • Deep Utopia (just started, seems…for popular audience?)

  • True Names (new genre)

• Found a new thinker to enjoy

  • Drexler!

Things I want to lean more into

• The few X/Twitter summaries of technical ML papers I posted ended up influencing my thinking a lot! I did that in December, not yet in January, and I think it’s time to get back into it. I really enjoyed the process. :))

• I feel good about roughly all my blog posts, but Celeste is pushing me to do more tiny-research write-ups — should probs aim for ~2-4/month

• I feel a bit sad I only wrote one technical ML post, and it feels very much half-baked. In contrast, I’m outputting a lot of macrostrategy/econ-of-AI stuff; it feels easier/more accessible but also less rigorous/interesting. I think it’s partly because I read all the FHI-people. I should read more Redwood-blog (for a nice blend of strategy & technical), OAIA-blog; Ant AliSci + Interp blogs; who else? Want to find small/underrated :)

Bad parts

• Things I should’ve dispatched quickly took me a pretty long time.

  • What’s more, they expanded to fill the time available.

  • I didn’t feel the crunch, because I’m on leave.

  • I feel pretty conflicted about this. On the one hand, 12-18 Jan was ultra-productive (socially), 1 week/year necessary-cashing-in (application-wise), and 19-22 Jan was semi-productive (Overleaf-jitsu & burnout prevention). OTOH, I didn’t adhere to ‘2h deep work before checking any devices’.

    • I think I can re-adhere to that on Weds 28 + Mon 2 onwards Tue 27 onwards!

      • If I consider any events in the 2 Feb — 22 Apr span, I’ll run them past careful checks — either wholesome/restorative with friends or interest/theme-based like Second Order or…

• I went to two Mox events that left me feeling drained.

  • I should start to conceive of events a bit as ‘satisfying vs. unsatisfying use of time’

• Trying to do ~anything else at the same time was ~incompatible with doing university math.

  • But that’s OK. I’ll definitely be back into it from Feb 2 Jan 27! onwards. Right now, it feels like ML research/experiments are too fun…(& easier to phase in and out of with light focus)

• I notice that my baseline satisfying-fun should be ‘the amount of satisfying fun I had doing ‘Bitter Lesson Eats’ & if I’m having less fun than that at an event, I should just find a quiet spot & switch to research

  • currently doing this at wsparc + happy :)

Live CoT

• In the process of writing this post, I realized I needed to, again, take my mnestics and cancel the California Zephyr (a 55h train ride) next weekend¹.

  Leaving a party an hour early to get to sleep on time, saying no to meeting a school friend, saying no to another conference, remembering to take decaf in the afternoon, reading that article, organising that book club, promptly responding to important messages.

  These are things I can only do reliably when I'm in regular emotional contact with a picture of what actually moves me, ordinarily too big to fit through the door of my mind at short notice. But if one takes a massive shot of space with which to dwell on it, you can compress that picture down into a mnestic you can take daily.

  This is happiness-inducing; it frees $350 and four days. Four days! I can start the rest of my life early! See, when I have a trip coming up, I often feel like “oh, I can start the rest of my life after this trip.” And when I’m really excited for that beginning, well…why not start ASAP?

• (I notice so much happens in SF every weekend that some non-SF event has to be really, really good to get away for. ~Almost everyone I need to meet comes through SF at some point. Any weekend I’m away from SF, I miss something good. Some people travel to SF just for a weekend — my default weekend is SF; I am blessed almost beyond measure).

Give yourself less to process

Canceling arduous trips applies a broader principle: giving yourself less to process. Some activities are inherently fairly dysregulating — fragmented sleep, low-quality food. If you’re trying to do good intellectual work, you should be careful about your exposure to dysregulating settings, especially ones that take recovery afterwards. I am really enjoying settings where someone else is the designated organizer, explicitly responsible for food/housing/etc. — each time you go to an event like that, you are getting their time for free. Well-planned reading groups are a win on this frame.

Fortnight-plan

with the rest to be filled out — metaphysics, etc. — as we go :)

1. I have three main modes: 1) doing mathematics, 2) doing research, 3) exploration. all of them are much better with wifi/claude.