Long-form posts on craft, product decisions, and what we're learning from beta writers.
A word diff proves exactly what changed. It cannot tell you whether the rewrite kept the things the screenplay needed to stay true.
Read postA character answers a question nobody asked them. A gun left in a car is back in a pocket. Each reads fine in its own scene, and only breaks against everything before it.
Read postUse Ask for the judgment. Use Agent Mode when the judgment and the editing need to develop together.
Read postGuild floors tell you the legal minimum on a covered job. BLS and Guild reports tell you what the market looks like around that floor.
Read postAcross 994 parsed scripts, the median per-script dialogue turn is seven words, while the median per-script 90th-percentile turn is twenty-four. Those are reference points for rhythm, not line-length rules.
Read postRaw cast size counts names. Speaking load shows how many of those names actually carry the conversation — and the gap gets much wider as casts grow.
Read postPage count is one number. Pacing is where that number goes. Measured across thousands of features, screenplays do not pace evenly — and the middle is not supposed to look like the open.
Read postA 105-page feature is not a 105-minute movie. It is a cold-read estimate. Treat the number as a signal, then check it against practice.
Read postName the job. Stay if versions are not the pain. Import if they are.
Read postCoverage judges the read. A breakdown tags the shoot. The useful first pass is an inventory, not a score.
Read postWhy Canvas bets on lossless markdown instead of locking prose in a proprietary graph.
Read postPlot for the room’s bible. Canvas for Notion/Docs-shaped knowledge around the show.
Read postBundled, Anthropic, or OpenAI — who holds the key, who pays the provider.
Read postFast, balanced, frontier — what each tier is for, and what a pass draws.
Read postPlot for screenplays, Canvas for everything else — and how to tell which one you need.
Read postA practical guide to taking one structured screenplay from White pages through production handoff.
Read postSkills edit nodes, not chat transcripts.
Read postImport once, name every read, export when the room asks — not before.
Read postHow we prioritize what ships while Plot is young.
Read postChoose the AI workflow by the writing job—not by which chat box happens to be open.
Read postThe habit of naming a moment matters more than the button that creates it.
Read post