Latest news & articles.

Relying on hosted machine learning models introduces a security risk.

Paying people to read and grade AI answers gets slow and expensive.  Because of

Deploying a massive frontier model to process every single incoming user query is ec

Traditional spoken dialogue systems rely on a cascaded architecture.

Standard AI models are a bit short-sighted.

A lot of early agent engineering relied on massive scaffolding frameworks.

Running an AI model takes a ridiculous amount of power.

The attention mechanism in transformer models is notoriously memory-bound.

The execution speed of large language models during inference is rarely limited by r