The Jetty Blog: Ground Truth

Looping into a Game About AI Evals

Agents play-testing a game about AI Evals

Jul 21 • Jonathan Lebensold

1

June 2026

Tests are Boolean. Evals are a Gradient

Why a green suite is necessary but never sufficient

Jun 22 • Jonathan Lebensold

1

1

Verifiable Agents Engender Trust in Systems

Turning confident guesses into measurable claims

Jun 13 • Jonathan Lebensold

1

Valley-Dodging: The Half of Agent Reliability You're Not Optimizing

Why your best runbook lines were never about the score

Jun 8 • Jonathan Lebensold and Ade Oshineye

1

1

May 2026

A Pelican Learns to Ride

What happens when every iteration of a benchmark is on disk

May 22 • Jonathan Lebensold

1

Lights-Out Manufacturing Had a Brake Pedal

The lights are going off in software the same way they went off in manufacturing

May 16 • Jonathan Lebensold

5

1

Research Closes the Loop. Production Keeps Us In It.

Why we kept the outer loop open by design

May 8 • Jonathan Lebensold

1

April 2026

Patterns Were the Map in the Search for Beauty

Christopher Alexander told the patterns community they missed the point. Thirty years later, agents finally let us listen.

Apr 28 • Jonathan Lebensold

14

6

2

My Backend is 442 Lines of Markdown

We shipped a web app whose entire backend is a structured document

Apr 21 • Jonathan Lebensold

2

1

The Jagged Frontier Is an Evaluation Problem

Why your AI system breaks in ways your evals won't catch

Apr 13 • Jonathan Lebensold

1

1

March 2026

Visual workflows are procedural programming in a costume

Why outcome specs beat node graphs in production

Mar 31 • Jonathan Lebensold

2

1

Runbooks: what agents need to hill-climb

The Missing Layer Between “Call This API” and “Accomplish This Outcome”

Mar 27 • Jonathan Lebensold

9

1

3