Learning to think, day 1
Topic: Learning to think
Date: Jun 11, 2025
As per my blog post from today, I wanna start predicting way more things, particularly other people’s world models.
I have a sense that this is gonna be kinda painful/embarrassing at first. But I can totally imagine it becoming fun and addictive like Defender of Basic says it became for him.
Question/problem 1 → who to tweet at? Tpot?
The main question for me right now is – where to begin?? What world views to try to predict??
Do I choose a tpot-y, post-rat-y person that I follow and dig through their tweets? Do I find someone ranting about politics? Do I find someone who has been tweeting about stuff I’ve been learning about recently?
Ok… I’m definitely noticing that, looking through my list of tpot people who I follow, that no one really feels like a good fit… Either they’re a big account who won’t reply, or they’re someone who I’m like “idk, because they’re post-rat, they may be more vibes-based anyway, where they won’t have a clear world model/stance”. This is probabyl dumb because I guess everyone has justifications for things, even if they’re hazy. There’s also a sense of not wanting to “shit where I eat”, re: becoming an annoying tpot reply guy who asks tpot people “why do you think that?” all the time, lol…
Damnit how do I find people. I think I need to leave the tpot zone, I’ve half-typed a few replies to tweets by a few people who I follow and just felt too insufferable lol. B/c tpot is vibes-based, replies like “do you think that because x” does feel very annoying (I’ve also received them and been like ‘dude shut up idk’)
Breaking free of tpot/post-rat scene
Ok so how to break free…
I guess I currently feel like I don’t have burning questions. Like, I don’t want to, idk, wade into “Israel vs Palestine” twitter and guess at people’s takes. Or find people who have been tweeting about Heidegger and ask for clarification…
Maybe I should make an alt so that I can have a different algorithm…
Maybe I need to follow more rationalists. It feels like tpot people aren’t saying stuff that is ammeanable to followup questions, more aphorisms and pithy sayings and shit.
Ayyy ok I stumbled across Defender of Basic showing off his good faith alt account, I’m gonna copy it
Making an alt and asking ChatGPT who to follow
I’m gonna set up an alt and follow in his footsteps
Ok an hour or so has passed. I made an alt, asked ChatGPT for some big people to follow (like Elon Musk), so that I can read their replies and find people to engage with. I read into some replies on AOC’s tweets and just… this is hard!! It’s people who are very polarised, saying talking points that don’t seem back up with evidence (e.g. the 13.7 million people who will be removed from Medicare → it’s ok because they’re all illegals → no they’re not, illegal immigrants don’t get access to Medicare)
I have a sense that I’ve started right at the deep end. I think they’ll be a calibration period where I get better at picking topics and people… because so far I’ve just seen kind of mad, very polarised, MAGA people, who are just dunking on people. I can’t imagine them replying to me and giving an explanation. But maybe that’s a skill issue on my part!
First thing I engaged with!!!! AOC & the Boulder attack
Ok cool, I poked around in AOCs replies for a bit longer and this led to some updates, like (here’s my ChatGPT history): (from people being like “it’s the illegals, dumbass!” when she’s tweeted about how bad it is that Trump is cutting Medicaid to 13 million americans, and another tweet I saw where someone said the Boulder attack was her fault. Thinking “ok, why would someone claim that AOC is to blame? Because of Democrat immigration policies? Because she’s “pro-Hamas”? Etc):
OG tweet here
what happened with the 2025 boulder attack?
https://en.wikipedia.org/wiki/2025_Boulder_fire_attack
can you give me the years that various presidents were in office, from 2000 onwards
how do illegal immigrants get on medicaid? do they?
How do American states work with regards to leadership? Like, do you have a head of the state? Thinking about in England, you have an MP, Member of Parliament, but obviously counties are very small and they don't have their own laws. As I know in America, you'd have a state, which would be for a certain party, like California, I assume is Democrat. And then you have like a state leader who sets the laws and policies and stuff for that state, but there's also federal. I just, I don't know much about this.
What’s REALLY exciting here is that - it feels like the SALIENCE is way higher, when learning this way. It’s not “I’ll learn about economics because it’ll be useful one day” It’s - I have a PREDICTION! I think they’re blaming AOC because x! Is that right??? And then I feel super engaged, and i’m like a detective, and I feel surprise at the real answer, and excitement to immediately learn new things that it’s very salient that I didn’t know.
So, it feels like the new info slots into my world model way more efficiently – it feels like I won’t need flashcards to remember this, because I had the very salient and emotive experience of having uncertainty, making a prediction, and getting the “aha” moment when the answer landed.
This is WAY different than the feeling of “I guess I’ll read a chapter of this book on John Boyd, and then make flashcards about it, and one day it’ll be useful”.
It’s learning to solve for x (to borrow another thing from Defender), rather than learning because it might be useful one day.
Another one → “Delaware is toast!”
After poking around for a while longer, trying to find something to reply to where the person wasn’t like, super pissed off and clearly not gonna reply to me, I find this thread and there was a clear “prediction affordance” (that is, I was like “oh shit, I don’t understand this simple thing”).
It feels like sometimes you stumble across something that could be easily answered. Like, “why does this guy think AOC is to blame for the Boulder attack” is also juicy, but in this tweet, someone is saying that Ohio is gonna rise, and Delaware is toast. It’s like, there’s a single fact that I’m missing, and can predict. “Delaware is toast because x - what is x?”
So this was great because I could make a prediction (I guess Delaware is particularly pro-business?), and then get a very quick update to my model (ask ChatGPT and it’s like “Delaware has historically been the place to incorporate, but there are 2 activist judges and this Elon Musk payout block thing and etc and now people are wary and looking to move away)
So, super quick update, super quick additional depth to my model!
And this also gave me the fruitful thought of “oh shit, so obviously in America, (I’m english), as states have different laws, some can be more business-friendly than others!
And that made me think - “I wonder why Silicon Valley is where Silicon Valley is – why the West coast, why California?” - so I guessed at that (“I think it’ll be partially the weather, partially maybe New York was already populated by bankers and stuff and was expensive, and I guess also California must have had startup/tech-friendly laws?), and then chatgpt told me that also the West coast was further away re: bombing in WWII, and a uni (I forget which one) intentionally fostered tech startup-y culture in the 30s-50s, and also military investment, and also non-compete clauses didn’t exist (or something like that)
As you can see, I don’t end up with iron-clad facts → I haven’t been making flashcards when learning by this way. But I do get honed intuitions and resolution (“oh, it was something to do with military funding + the involvement of colleges, nice, that’s a thing that can have an effect, makes sense). It feels a lot faster than “ok I’ll make a deep research report on this, read it with speechify, make flashcards, review the flashcards, lock all the facts in.
I also tried to get a reply from someone here. I think unskillful because this doesn’t seem like the kind of take where someone is interested in explaining further…
End of session post-mortem
- I got to experience this new mode of learning, where I have a strong desire to answer a prediction (why are they saying AOC is to blame??), which makes the answers I get highly salient, so they slot into my world model in a very robust-feeling way, that is, not requiring flashcards !!!
- I didn’t get any replies on twitter,
- But, I set up the alt, poked around trying to find decent threads to reply to, set up this blog post, following people, etc
Also I tweeted about following in Defender’s footsteps and he follows me now, hell yeah
