Welcome to Fresh Worktree

Fresh Worktree is a place where I test the frontier of the outer harness, and tell you the results.

At the time of writing AI coding is changing every week with new models, skills, plugins, methodologies, harnesses, and somehow more.

Loops, goals, graphs, multi lab, runners, labs telling us to strip back agent md files, orbs ? This is what I’m talking about.

There is a mountain of model benchmarks, and a small foothill on how to use those results in the outer harness.

Also, the frontier is moving too fast to cross check every combination of new releases.

So, by design, the testing will be high quality but unashamedly incomplete - prioritised by what's important to engineers in the moment.

To know what that is we will also interview engineers and publish their workflows  - DMs open.

Tom