Bram Wasti (@bwasti) header
Bram Wasti (@bwasti) profile picture

work with fair, msl

NY
Joined January 2011
Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)Photo by Bram Wasti (@bwasti)
calling it now: the next google model is going to be named rutherfordium
1
10
483
what the careful and uncompromising development of formal verification over the past 40 years has enabled
1
15
732
Bram Wasti retweeted
Can LMs discover & fix bugs if you don't tell them what went wrong? To proactively maintain a repo, agents need to find problems before users do. SWE-sweep benchmarks this on 100 repos, 22 languages, 4k real bugs (numpy, php interpreter, lean kernel). Top models get <5%.
75
89
52
851
608,590
me attempting to rent a car in mid 2026
1
6
1,203
Bram Wasti retweeted
We made Muon run up to 2x faster for free! Introducing Gram Newton-Schulz: a mathematically equivalent but computationally faster Newton-Schulz algorithm for polar decomposition. Gram Newton-Schulz rewrites Newton-Schulz such that instead of iterating on the expensive rectangular X matrix, we iterate on the small, square, symmetric XX^T Gram matrix to reduce FLOPs. This allows us to make more use of fast symmetric GEMM kernels on Hopper and Blackwell, halving the FLOPs of each of those GEMMs. Gram Newton-Schulz is a drop-in replacement of Newton-Schulz for your Muon use case: we see validation perplexity preserved within 0.01, and share our (long!) journey stabilizing this algorithm and ensuring that training quality is preserved above all else. This was a super fun project with @noahamsel, @berlinchen, and @tri_dao that spanned theory, numerical analysis, and ML systems! Blog and codebase linked below 🧵
17
163
24
1,036
221,662
🚨BREAKING AI NEWS 🚨 A Cambridge study just dropped that PROVES you can exactly calculate the slopes of functions at an arbitrary point. This UNLOCKS gradient optimization that experts say is vital for AGI. Download our app for a daily AI digest delivered to your inbox
32
107
20
1,344
96,808
pretty much no AI can classify a hand painted CIFAR-100 clock. a very under-explored dataset!
1
5
787
gemini gets it tho. 👏
2
161
using vim to edit my prompt
7
237
when claude comes up with a root cause it already debunked three chat compressions ago
10
397
Bram Wasti retweeted
🥝 Meet Kimi K2.5, Open-Source Visual Agentic Intelligence. 🔹 Global SOTA on Agentic Benchmarks: HLE full set (50.2%), BrowseComp (74.9%) 🔹 Open-source SOTA on Vision and Coding: MMMU Pro (78.5%), VideoMMMU (86.6%), SWE-bench Verified (76.8%) 🔹 Code with Taste: turn chats, images & videos into aesthetic websites with expressive motion. 🔹 Agent Swarm (Beta): self-directed agents working in parallel, at scale. Up to 100 sub-agents, 1,500 tool calls, 4.5× faster compared with single-agent setup. - 🥝 K2.5 is now live on kimi.com in chat mode and agent mode. 🥝 K2.5 Agent Swarm in beta for high-tier users. 🥝 For production-grade coding, you can pair K2.5 with Kimi Code: kimi.com/code - 🔗 API: platform.moonshot.ai 🔗 Tech blog: kimi.com/blogs/kimi-k2-5.htm… 🔗 Weights & code: huggingface.co/moonshotai/Ki…
754
1,928
1,189
15,536
7,365,762
new side project, stay tuned :)
1
1
12
834
would be so nice to have claude venvs. `source my_claude_env/activate` without having to check your dir for the configs/mds cc @bcherny
1
3
346
Bram Wasti retweeted
prefill with a single token be like
3
2
121
10,391
Our new paper shows that RoPE—the positional encoding used in most modern LLMs like Qwen, Gemma, DeepSeek—has a fundamental flaw: it entangles "what" (content) and "where" (position) information. Our fix (PoPE) is simple but powerful. Paper: arxiv.org/abs/2509.10534
30
175
18
1,453
164,207
do we really need things like garbage collection in an AI-native world? how much of the overhead in high-level languages can just disappear with vibe-coded stacks? zig is looking more appealing every day
1
2
336
github:bwasti/binfer -- an experiment with fast inference serving using bun + cuda. trying to focus on speed / UX experiments, like language overhead and startup time. not trying to maintain this as a legit framework (use vllm or sglang)
6
859
is this just hidden states?
2
205