In this channel's material, a burn the ships rewrite is one of the hands-on vibe checks the Every team runs when evaluating a new AI model — a test of whether the model will make a bold, all-in rewrite of a piece of writing rather than hedging with light edits. The allusion is to committing so fully that there's no option to turn back: the original draft's structure is thrown out, and the model has to rebuild the text from scratch. It matters because it exposes a real behavioral difference between models. When Dan Shipper, Brandon Gell, and Katie Parrot tested Claude Opus 4.7 live, they found a model that was more literal and more detail-oriented than its predecessor — strong at planning, spreadsheet-and-writing work, and long-horizon background tasks — but one that proved hesitant to execute bold rewrites, preferring careful, incremental changes to a full, committed rebuild. In this channel's material, a burn the ships rewrite is a vibe-check test in which a model is asked to make a bold, all-in rewrite of a piece of writing — discarding the original draft rather than lightly editing it — to see whether it will commit fully or hedge with incremental changes.