iter.ca update #13
Welcome to this week’s instance of my weeklyish newsletter. I hope you enjoy it!
stuff I’ve done
- I started working on shutting down an old service I can’t maintain anymore
- I participated in and hosted Writehaven, where you had to do a lot of writing every hour
- I vibed up What's your favorite letter?
I wasn’t that productive this week; hopefully I’ll be more productive over the coming week!
posts
- I updated my About me page
- Bypassing the GitHub workflow permission with nested Git tags
I'll publish some more things from Writehaven soon.
future plans
What should I do in the rest of 2026? I don’t think there are any fellowships or similar happening this year that I could still apply to, so I’ll probably have to do something self-directed. Here are some quick thoughts on the main things I’ve been thinking about working on:
NLA stuff
I’m still pretty excited about natural language autoencoders (I’ll write another post about my general NLA thoughts soon); I think that area of research might be useful to work on. Probably it would be best for me to try to think of ways to make something that's like an NLA but better?
I think some of my previous work on NLAs was too unambitious. It was useful for helping me understand NLAs (and how to train them) better, but it wasn’t very useful for actually using NLAs better. But at this point I should try to directly make meta-models that are really good, and doing that probably looks like making something that’s pretty different from the existing stuff.
I’m doing Writehaven as I write this and I like writing about a bunch of stuff; it seems like it would be helpful to write about my thoughts and current status as I go. I think I might actually do the “Thinkhaven” thing this month.
other things I could do
Here are some other things I’ve considered doing:
pivot to governance/advocacy?
My vague impression is that there might be some low-hanging fruit here! It might be worth spending about a week looking at the current status of the AI governance and advocacy world and trying to figure out if there’s anything that seems very useful to have done that isn’t being done by anyone. (It would be pretty sad if I could actually meaningfully contribute here, since it seems there are many more people who have the ability to contribute here. I hope governance/advocacy isn’t under-resourced!)
get a job?
Maybe I should get a job at some org that does cool AI safety stuff in the SF Bay Area; there are several such orgs! I think it would be hard for me to get such a job right now, though.
pivot to ops?
Doing ops stuff is actually pretty cool and fun! I’ve talked to some ops people in the past, and it seemed like they have really cool jobs. I’ve read stuff like Adventures of the Lightcone Team and thought it seemed pretty fun. I think I should consider maybe doing ops stuff.
where
Assuming I don’t get a job (which would tie down my location), I don’t have any hard constraints as to where I can do stuff from. So probably London, unless I can figure out a way to live in SF? Currently I’m planning on being here in London until December 14 (give or take a few days), at which point I'll go back to Toronto for a bit.
london
I haven’t actually done that much in terms of London-maxxing, but I’m pretty sure I will over the next week!
Add a comment: