Full Rank

Less chatty judges

How my LLM judge advice carries over to Jev (or other ‘decision models’)

Oct 5 • Ryan Sloan

1

September 2026

COGS in the machine

Is Opus 5.5 really 40% cheaper? Only your own measurements can truly say.

Sep 28 • Ryan Sloan

If at first you don’t succeed, try something else

Agentic tasks that fail once often fail again. “Just retry” isn’t a strategy.

Sep 21 • Ryan Sloan

Better good than lucky

The winner’s curse is coming for your evals

Sep 14 • Ryan Sloan

Read your judges’ dissenting opinions

Using disagreement when a panel of LLM judges represents competing priorities

Sep 8 • Ryan Sloan

August 2026

Should PMs Own Evals?

An eval program is six jobs, assigned to the team you have

Aug 31 • Ryan Sloan

The Illusion of Precision

Probabilities from LLMs are a charade, ask for what they're good at instead.

Aug 24 • Ryan Sloan

A Million-Token Bad Habit

Leave the junk drawer for your kitchen

Aug 17 • Ryan Sloan

The LLM is a jackhammer

Overkill if your problem is a nail.

Aug 10 • Ryan Sloan

Not Every Ticket Is Your Opus

Bigger isn't always better, and you don't have to guess which model you need.

Aug 3 • Ryan Sloan

1

July 2026

Don't make the model do the math

Well-designed agent tools guard against hallucination

Jul 27 • Ryan Sloan

1

You can do better than retrieval

Using graph+vector structure to unlock new patterns in data

Jul 20 • Ryan Sloan