Catch-Up Thursday

Everything that happened in AI coding this week

How models are served and how that impacts your bill

I rented a GPU and served an open model to understand how prices work

Why new models get cheaper

Opus 5.5 and GPT-6 Sol cost about half as much per task.

I moved my coding agents to a remote server

My laptop is no longer the bottleneck

Which model to use?

Fable, Sonnet, Opus and now GPT 5.6. Which one to use?

Someone put ads in the Claude Code spinner

Plus: understanding AI code, and how /fast works

How to Use Claude Code and Still Understand Your Own Code

How Claude Code Fast Mode Works

Why it's 2.5× faster, why it costs 2–6× more, and when to turn it on.

Fable 5 is offline. The government pulled it.

Here's what we know.

Two weeks with Claude Fable 5

and what is Loop Engineering?