Distributed Thoughts
  • Home
  • About
Sign in Subscribe

inference

A collection of 3 posts
We Rented the Mainframe Back
ai

We Rented the Mainframe Back

Two AI assistants went dark in forty-eight hours last week. In neither case did the model break. The thing that broke was the wire in front of it, and we built that fragility back on purpose.
25 Jun 2026 4 min read
The Token Got Cheaper. Your Bill Didn't.
ai

The Token Got Cheaper. Your Bill Didn't.

An enterprise client of an AI consultant accidentally spent half a billion dollars on Claude in a single calendar month. The story broke in late May. It is not an aberration. It is the bill the per-token pricing chart has been hiding for two years.
22 Jun 2026 7 min read
The Inference Inversion
distributed computing

The Inference Inversion

For the past three years, the AI industry has operated under a simple assumption: more centralized compute solves everything. Bigger clusters. Bigger data centers. Bigger power contracts. The logic was intuitive and, for training workloads, largely correct. Training a frontier model requires tightly coupled GPU clusters with high-bandwidth interconnects,
06 Apr 2026 6 min read
Page 1 of 1
Distributed Thoughts © 2026
  • Sign up
Powered by Ghost