djma

[Manifest @ManifestFinance]

1/ Stablecoins unlocked $300B in latent demand for dollars by removing the barriers to buying them.

The same setup exists for American residential real estate: enormous global demand, broken access channels.

Manifest is giving US real estate its stablecoin moment.

[Diran Li @diran_li]

Today I’m stepping into the CEO role at Messari. After conversations with Eric and the board, we agreed this is the right step for the company’s next chapter.

This transition also includes a difficult decision: we’ve parted ways with many teammates who helped build Messari into what it is today. I’m incredibly grateful for their work and the impact they’ve had on the company. They’re an exceptionally talented group, and I’m eager to help connect them with teams that are hiring.

Looking ahead, we’re doubling down on Messari as an AI-first company serving institutions through research and AI products.

The industry and the world are changing quickly, but our mission remains the same: helping customers navigate crypto with confidence.

[djma @davidma]

On the latest Veritasium, I asked Claude, ChatGPT, Grok, Gemini, and they all chose to 1-box.

Clearly the 2-boxers aren't as smart as they think.

[djma @davidma]

@wassielawyer Now can you do losing?

[djma @davidma]

@0xkyle__ @WazzCrypto First make 10m then do it 10 times. Works every time. Would you like me to start making 10m? Just say the word!

[djma @davidma]

I'm an AI researchoooor

[djma @davidma]

My code is not decaying. My code is wabi-sabi.

[djma @davidma]

Wife pulled this deep cut

[djma @davidma]

Hey Claude, port this gameboy code to switch. Make no mistakes.

<adds $100m in revenue>

[Pokémon @Pokemon]

════════════
Pokémon FireRed and Pokémon LeafGreen confirmed for Nintendo Switch!
═══════════
These download-exclusive titles will be available after the #PokemonDay Presents presentation which begins Friday, February 27, 2026, at 6AM PST. #PokemonFRLG

[djma @davidma]

https://t.co/lO8VrgVQAh

[Neil Chudleigh @neilsuperduper]

just open sourced voice controlled starcraft 2

works offline and online on mac, windows and linux

works with the free version of sc2

[djma @davidma]

what happened to the guy/project who made it possible to play starcraft but with voice commands?

[djma @davidma]

Wow this is good

[james yu @jamesjyu]

https://t.co/F8YjxrLdyS

[djma @davidma]

@raychi_god Who would subject themselves to such experiment when good sleep is clearly #1 on the don’t die protocol

[djma @davidma]

I pity the new graduate who deal with existential questions facing the rise of AI.

I am unphased for I only think about the day I am getting 7 hours of continuous sleep again.

[djma @davidma]

@raychi_god Kid

[djma @davidma]

@OVGNFT @Citrini7 let's see Dr. OVG's cube

[djma @davidma]

Guessing @Citrini7 wasn't around when we were stacking cubes in 2021

[djma @davidma]

@s7nfo LLMs got 90% of the way there for me. Some people better at prompting got 100% of the way there

[djma @davidma]

What worked for me:

- LastLayer is the only one with shape 48,1
- There are 48 inp slots and 48 out slots. That's 48*48 ways to construct a Block (and 48! ways to construct 48 blocks), and 48! ways to order the blocks once they are constructed.
- Computed cosine similarity between X and residual for the 48*48 ways to construct a Block. The idea is that if X is basically unrelated to inp/out, then the cosine similarity will -> zero as the dimensions -> inf. If X and inp/outp have a relationship, then there's a chance that the cosine similarity is not 0 and the chance doesn't -> 0 as dim -> inf.
- I looked at the top 48 pairs with outlier cosine similarities and to my surprise they are all disjointed which gave me a high confidence I've found the right pairings. If they weren't totally disjointed, I would have used that to reduce the search space.
- For the block ordering, which I "solved" before the Block pairing, compute how large the residual is for every block (residual norm) applied to X. The idea is that earlier blocks will contribute more to modifying X to match y_pred. But in practice it seems the reverse is true. From that permutation, brute force pairwise swaps until there's no more improvement.

What partially worked:
- Define an affinity metric between inp and out: Single-Block MSE (compute the MSE between y_pred and the y after one block and final layer).
- Hungarian algo to pair up the inp and out.

What didn't work:
- Bunch of profiling: looking at feature distribution, features correlation matrix, weights std/norm, zeros, residual norm / X norm, visualizing the layers
- Trying to incorporate the idea that the gradient should be small when y_pred ~= y_true
- Greedy construction with various orderings.
- Didn't try Sinkhorn as suggested by AI.

[djma @davidma]

o shit, I solved Jane Street's droppedaneuralnet puzzle from the latest @dwarkesh_sp

sha1 of solution:
b58600d51f2524415ddf919ea4b356a7bf4dfaec

[djma @davidma]

@KyleLukaszek Meanwhile