Currently unlisted, will rewrite yearly review and goals in a smarter form soon.

Looking back on 2022

If I had to summarize the year in one word, it would be “social”. I worry that I didn’t grow or study as much as I could have. In particular, I didn’t do any other learning projects comparable with teaching myself real analysis and writing a solutions manual for Abbott.

I also abandoned a lot of things I’d started. I didn’t finish Hammertime, didn’t finish functional analysis or have any major outputs from it, stopped reading Penrose’s road to reality, stopped reading Quantum computing since democritus

A lot of this is overscheduling and not giving myself enough slack, a problem I aim to fix by saying ‘no’ to most things in future.

Overall I give myself a C. I’d give myself a D if it wasn’t for my large increase in agency. Academically my powers didn’t grow much.

Month-wise summary:

• January - April: Readding abbott and axler (real analysis and linear algebra), writing a solutions manual, fragments of solutions to axler as well. (eventually stopped & just proved things to myself)
• April 2022: Death with dignity destroyed the last remnants of my trust in civilizational adequacy. Decided I’ve got to start doing something now.
• June: Functional analysis reading group, Atlas Fellowship. Made many friends, didn’t stay in touch much.
• July: Prometheus ELK Contest with a friend from Atlas, got into the OG shard theory doc, our proposal sucked but my models of alignment changed a ton. We got a generous honorable mention. Here’s my postmortem of the experience.
• August: Vacationing in Italy and ESPR. Many friends! made sure to stay in touch this time, started a discord server for my espr+atlas+other nerd friends. Very successful.
• September: FTX regrant to visit the serimats symposium, met seri mats scholars, met John and Alex decided to apply to the next MATS. Here’s my journal entry for that day.
• October, November: Experimental group house in Berkeley with some Atlas & ESPR people, accepted into serimats under Alex & John, had a small crisis since I liked them both so much, ended up going with Alex while still doing the John workshops.
• December: SERI mats research sprint, not happy with my result, though I think I did improve alignment models & research taste a bit. Journaling on hackmd for a bit, mostly holidays socializing after research sprint.

Plan for 2023

(Or rather, optimization targets for 2023.)

I want to grow as a person (meditation / mental healthiness / rationality / productivity), study a ton (especially ML/stats/linalg), and publish some useful alignmenty-things (likely distillations).

Daily I definitely want to:

• Sleep 10pm-7am, spend 7-9 efficiently doing daily stuff
• Exercise intensely (~20mins/day) (plank + pushups + handstand things)
• Meditate 1h/day
• Anki reviews (~10mins)
• Memory training (~20mins)
• Budget of 15mins discord time till 5pm, don’t check in the morning

Murphyjitsu for daily plan

Things that’ll obviously go wrong by default:

• I do it for a bit, something happens, I stop for a few days, I don’t restart
• Sometimes I need to contact people over discord for work things
• I get roped into rationalist social events / parties, stay up late, have low willpower the following day, my schedule collapses.
• I have a deadline and skip my routine to work on it more
• I decide my routine is bad and I need to change it

Plans to fix those:

• Precommit (with a friend) to never skip more then one day in a row, and no more then two days per week. (by “skip” I mean do less then all the things listed above.)
• When you want to share links note them down and share later, when you must use discord text someone quickly and tell them to text you back using phone num or slack.
• Don’t go to rationalist social events that go late. No more scarcity mindset!
• Factor the fact that you won’t skip days into your procrastination-planning. You still have plenty of time while factoring that in
• Precommit to being able to change routine, but with a five (successful) day wait period (similar to beeminder), though I’m allowed to add stuff at any time.

Things that’ll go wrong now:

• Friends stop reminding and I silently stop checking in.

Plans to fix those:

• Make the default action for an update to be sent. E.g. via a crontab script that sends messages to discord. Like a dead mans switch of some sort. (edit: I’ve implemented this)

Rough calendar/timewise plan (will edit, not done):

• January, March: SERI MATS. Distill causal incentives research, get good at ML experimenting.
• March forwards: Study intensely, saying no to opportunities that restrict my freedom/slack. Goal is to do an Ultralearning + quality speedrun of John’s study guide, get good at using tutors & learning rapidly, etc.
• While I’m studying intensely setup a frequent blogging habit / sharing and distilling what I’m learning. I’d like to make framing exercises at some point too.