FEX-Emu β A fast linux usermode x86 and x86-64 emulator
https://fex-emu.com/
https://fex-emu.com/
Fex-Emu
FEX-Emu β A fast linux usermode x86 and x86-64 emulator
#todo write custom metal kernels in MLX
Target bottlenecks: Fused RMSNorm + ROPE + SWIGLU, quantΓzed matmul, or attention variants.
https://ml-explore.github.io/mlx/build/html/dev/custom_metal_kernels.html
Target bottlenecks: Fused RMSNorm + ROPE + SWIGLU, quantΓzed matmul, or attention variants.
https://ml-explore.github.io/mlx/build/html/dev/custom_metal_kernels.html
Building low-level software with only coding agents | Lee Robinson
https://leerob.com/pixo
https://leerob.com/pixo
Leerob
Building low-level software with only coding agents | Lee Robinson
520 agents, 350M tokens, and $287 to build a Rust-based image compressor.
[2512.24880] mHC: Manifold-Constrained Hyper-Connections
https://arxiv.org/abs/2512.24880
https://arxiv.org/abs/2512.24880
arXiv.org
mHC: Manifold-Constrained Hyper-Connections
Recently, studies exemplified by Hyper-Connections (HC) have extended the ubiquitous residual connection paradigm established over the past decade by expanding the residual stream width and...
If you're going to vibe-code with an LLM, pick problems where the LLM's blindness to edge cases is catchable by automation.
The Principle
Novelty = (importance of problerm) (gap from current best) x (verifiability)
Novelty = (importance of problerm) (gap from current best) x (verifiability)
Build GPUI + Ghostty without writing code
https://xuanwo.io/2026/01-gpui-ghostty/
https://xuanwo.io/2026/01-gpui-ghostty/
xuanwo.io
Build GPUI + Ghostty without writing code
Achieving Data Freedom Through Open Source and Rust