Ryan @_ryanbloom

Any sufficiently advanced technology is indistinguishable from magic ryanbloom.xyz San Francisco, CA Joined August 2018

Tweets

193
Followers

176
Following

935
Likes

946

Ryan @_ryanbloom

10 months ago

@DavidSHolz Interested!

0 0 0 32 0

View Details

The hardest part of coding with LLMs is resisting wishful thinking about their capabilities. It's so tempting to just type a simple prompt and sit back to watch; it's also rewarding sometimes, like a slot machine. But it's a pretty ineffective workflow. For now.

0 0 1 186 0

View Details

Ryan @_ryanbloom

11 months ago

@tensor_rotator If you mean globally I think I would have to pick money

1 0 1 29 0

View Details

Ryan @_ryanbloom

a year ago

Woah, it got the second one!

0 0 1 226 0

View Details

Ryan @_ryanbloom

a year ago

I wonder what capability Cursor will gain first: automating my entire job, or opening a folder when I drag it onto the dock icon

1 0 5 359 0

View Details

Ryan @_ryanbloom

a year ago

@urbanistvc @EdgeEsmeralda Happy birthday!

0 0 1 48 0

View Details

Ryan @_ryanbloom

a year ago

Agency is mostly just curiosity about the best way to do something

0 0 1 237 0

View Details

Ryan @_ryanbloom

a year ago

When reward hacking gets in the way of AI progress

0 0 5 299 1

View Details

Ryan @_ryanbloom

a year ago

A Few Things I Learned About Evals ryanbloom.xyz/posts/a-few-th…

0 0 2 183 1

View Details

Ryan @_ryanbloom

a year ago

@KfirDolev @alexinfanger Glad you liked it!

0 0 1 23 0

View Details

Ryan @_ryanbloom

a year ago

The weights for the autoencoder come from this paper by Aditya Cowsik, @alexinfanger, and @KfirDolev: arxiv.org/abs/2410.12101

0 0 0 134 0

View Details

Ryan @_ryanbloom

a year ago

Watch it here: youtu.be/q1aB91-6tx8

1 0 1 201 0

View Details

Ryan @_ryanbloom

a year ago

I made a video! It introduces superposition in neural networks, and shows how to construct weight matrices for superposition in a simple autoencoder. (1/3)

1 1 9 518 0

View Details

Ryan @_ryanbloom

a year ago

that feeling of dread when o3-mini botches an edit for the third time and you have to actually read your codebase

0 0 2 151 0

View Details

Ryan @_ryanbloom

a year ago

Just updated my homepage for the first time since “What (if any) are the limits of large deep learning models?” was an interesting question

0 0 2 166 0

View Details

Ryan @_ryanbloom

a year ago

Few things make me as grateful for the internet as seeing someone thrive in a niche that couldn’t possibly have existed without it

0 0 3 132 1

View Details

Ryan @_ryanbloom

a year ago

Welp there goes my 2025 FrontierMath forecast

0 0 1 133 0

View Details

Ryan @_ryanbloom

2 years ago

“Invent a new hyper-entity” is my new favorite eval for reasoning models

Michael Nielsen @michael_nielsen

2 years ago

New essay: "How to be a wise optimist about science and technology?" Link in next post

15 47 273 79K 233

0 0 0 162 0

View Details

Ryan @_ryanbloom

2 years ago

Challenges for Impact Markets ryanbloom.xyz/posts/challeng…

0 0 2 83 0

View Details

Ryan @_ryanbloom

2 years ago

Ok here's another: there's a base model inside all of us ❌ How can I motivate myself to get more exercise? ✅ How can I prompt myself to get more exercise?

0 0 3 242 0

View Details

Ryan @_ryanbloom

3 years ago

One way LLMs have changed my perspective on psychology is that I’ve started to picture subconscious processes as happening “inside” or “between” conscious ones, rather than just “under”