We found a way to make AI run complex computer code instantly, in one go, without ever having to teach it how the language works.
April 13, 2026
Original Paper
Loom: A Scalable Analytical Neural Computer Architecture
arXiv · 2604.08816
The Takeaway
This architecture bridges the gap between traditional software (C code) and neural networks by deriving weights analytically rather than through learning. It turns AI into a predictable, symbolic computer that still benefits from the efficiency of the Transformer design.
From the abstract
We present Loom, a computer architecture that executes programs compiled from C inside a looped transformer whose weights are derived analytically. The architecture implements a 22-opcode instruction set in 8 transformer layers. Each forward pass executes one instruction; the model is applied iteratively until the program counter reaches zero. The full machine state resides in a single tensor $X \in \mathbb{R}^{d \times n}$ of fixed size, and every step has fixed cost for fixed $d$ and $n$, inde