docs/getting-started/why-attaform.md pitches undo/redo twice: in the page description ("built-in field arrays, history, wizards, and devtools") and as a headline bullet at line 80 ("Undo / redo with bounded history, opt-in per form").
Measured cost of the feature: 309 B gz on a consumer who never uses it, 498 B more on one who does. That is cheap, and the plugin seam behind it is good design: the runtime rides the consumer's own attaform/history import, so the core never links the history internals.
The problem is not the code, it is the billing. Undo/redo has zero uses across 136 useForm call sites in the only real consumer, three months after the decision to promote it. The decision's own recorded caveat predicted this: "undo/redo demos better than it gets used in the wild; the risk is under-discovery." Dogfooding settled which it was.
A headline bullet is a claim about what a reader will reach for. This is not that. It should read as a capability that is there when wanted, alongside the other opt-in entries, rather than as one of the reasons to pick Attaform.
Keep the feature and the docs/cross-cutting-state/undo-redo.md page. This is about the pitch only.
Found in the 2026-09-18 feature audit (main@b4a64ef9, eager baseline 33,204 B gz). Every byte figure is a measured ablation, not an estimate.
🤖 Generated with Claude Code
https://claude.ai/code/session_01PnAVwFjpvKQNppAMkSvoiH
docs/getting-started/why-attaform.mdpitches undo/redo twice: in the page description ("built-in field arrays, history, wizards, and devtools") and as a headline bullet at line 80 ("Undo / redo with bounded history, opt-in per form").Measured cost of the feature: 309 B gz on a consumer who never uses it, 498 B more on one who does. That is cheap, and the plugin seam behind it is good design: the runtime rides the consumer's own
attaform/historyimport, so the core never links the history internals.The problem is not the code, it is the billing. Undo/redo has zero uses across 136
useFormcall sites in the only real consumer, three months after the decision to promote it. The decision's own recorded caveat predicted this: "undo/redo demos better than it gets used in the wild; the risk is under-discovery." Dogfooding settled which it was.A headline bullet is a claim about what a reader will reach for. This is not that. It should read as a capability that is there when wanted, alongside the other opt-in entries, rather than as one of the reasons to pick Attaform.
Keep the feature and the
docs/cross-cutting-state/undo-redo.mdpage. This is about the pitch only.Found in the 2026-09-18 feature audit (
main@b4a64ef9, eager baseline 33,204 B gz). Every byte figure is a measured ablation, not an estimate.🤖 Generated with Claude Code
https://claude.ai/code/session_01PnAVwFjpvKQNppAMkSvoiH