looped transformers
Beyond Market Intelligence keeps looped transformers in one place: 2 stories so far. The section currently leads with “Looped transformers share weights but memory limits keep them from staying on chip” and “Beyond the Token Stream: Architectures for Deeper Reasoning”. Shared weights sound like a free lunch, but memory bandwidth has other plans. The field is quietly shifting its focus from longer chains of thought to something more fundamental: reasoning that never touches language at all. Acme AI is the next-generation, AI-powered spreadsheet platform built to replace Excel and redefine how analysts, data scientists, and enterprise teams work… The list below is every looped transformers story on Beyond Market Intelligence, newest first.
Looped transformers share weights but memory limits keep them from staying on chip
Shared weights sound like a free lunch, but memory bandwidth has other plans. A chain of eight matmuls with one 72 MiB weight ran 24% faster than eight identical copies, reading almost nothing from HBM. Scale that weight to 288 MiB, larger than L2 cache, and the advantage vanishes. Eight iterations of a full Huginn Core took eight times as long and read eight times the data. Sharing saves storage, but the GPU still fetches those weights from HBM every single time.
Beyond the Token Stream: Architectures for Deeper Reasoning
The field is quietly shifting its focus from longer chains of thought to something more fundamental: reasoning that never touches language at all. The breakdown of latent reasoning into five distinct families, from Coconut's continuous hidden states to BDH-CQ's in-context recurrent memory, gives us a useful map of where the frontier actually sits. The real question isn't which architecture wins, but whether readable traces were ever the point.