Bolton Bailey (@fogofchess) header
Bolton Bailey (@fogofchess) profile picture

Project Numina

Datumplane
Joined June 2009
Photo by Bolton Bailey (@fogofchess)Photo by Bolton Bailey (@fogofchess)Photo by Bolton Bailey (@fogofchess)Photo by Bolton Bailey (@fogofchess)
Astrophotography attempt #1. I need a sturdier way to mount my camera to the telescope.
3
45
youtube.com/watch?v=4rkvDJlk… A podcast I did with @j_schwartzz on some research I did in grad school was released today! You might like to listen if you have an interest in quantum computation or blockchains.
2
6
509
Bolton Bailey retweeted
🔥Releasing NuminaMath-LEAN, a large-scale dataset of 100K mathematical competition problems formalized in Lean 4, with more than 20K human annotations. Paired with Kimina-Prover, Kimina-autoformalizer, CombiBench, we hope these will advance open-source AI for formal math !
7
105
332
18,702
Bolton Bailey retweeted
Happy to introduce Kimina-Prover-72B ! Reaching 92.2% on miniF2F using Test time RL. It can solve IMO problems using more than 500 lines of Lean 4 code ! Check our blog post here: huggingface.co/blog/AI-MO/ki… And play with our demo ! demo.projectnumina.ai/
6
45
11
274
28,658
Here's another post, this time on quantum computing for bitcoin mining open.substack.com/pub/thequa…
2
164
Here is a post I wrote as an introduction to quantum pseudo-telepathy. Speculation on future applications included! open.substack.com/pub/thequa…
1
1
3
711
New fascinating work by @or_sattath and Bolton Bailey exhibits a new quantum attack on Bitcoin (and PoW in general), that only requires a single "relatively small" quantum computer. The gist of that attack is that the attacker cultivates a side chain where they *increase* the difficulty by a factor of X, increasing the weight of their blocks by X while only increasing the work they put into each block by sqrt(X), allowing them to create a heavier chain despite being computationally inferior. This attack does not contradict previous results showing quantum miners only have quadratic advantage, since these works assume constant difficulty whereas the new attack abuses difficulty adjustment. The paper leaves mitigating the attack (by means better than repeated hashing) an open problem. More details here: arxiv.org/abs/2403.08023
22
24
6
157
20,584