Subquadratic came out of stealth with a large claim: it says it has made progress on a mathematical limitation that has constrained large language models for years. MIT Technology Review notes that the company initially offered few details, drawing skepticism from researchers.
The potential significance is scale. If the approach holds up, it could affect how future LLMs handle computation, context, or efficiency constraints that have become central to model development.
The story is still early, so the strongest read is not that the bottleneck is solved, but that a new technical claim is now entering public evaluation.