Zero drafting a pilot, start to finish

The world as it stood

Established before the session, and all of it written down:

• Eli Vance, 47, keeper of the light at Cape Sorrow.
• Eli has built a signal that pulls the drowned back.
• His daughter Maren drowned at 9. She came back at 9. She has not aged in the eleven years since.
• Eli has kept the light for twenty years.
• The town knows something is wrong at the lighthouse. It does not know what.

That is a small, tidy canon — the kind that feels completely settled. Five facts, no obvious tension, all of them decided deliberately. Worth noting, because the failure below is not hiding in an obscure corner. It is in the middle of the premise.

Choosing the scene

Not the opening. The scene chosen was the one where Eli and Maren are alone in the lantern room and he asks her, directly, what she is now.

It was chosen because it collides the most: both characters, the signal, the eleven years, the location, and the town's ignorance all have to hold simultaneously in one room. It was also, quite openly, the scene that had been avoided for two weeks — which was the deciding factor. Avoidance is the most reliable targeting instrument available, and it was correct here.

Setup

Canon printed, one page, sitting to the left of the keyboard. A second blank page to the right, divided into two columns headed **Missing** and **Broken**. Timer set for ninety minutes.

The printout matters more than it sounds like it should. In an earlier attempt at the same scene, the canon lived in a document behind the draft, and checking meant switching windows. The checking stopped inside ten minutes and the session produced nothing — which is the standard failure, and the reason the method insists on physical visibility rather than merely having a record.

What the session was like

The first stall came about four minutes in, on what Maren calls him. Not a small question — if she still says "Dad" the scene is one thing, and if she uses his name it is a different show. It went on Missing, a placeholder went in the draft, and the scene carried on.

That is the rhythm: write fast, hit a wall, log it, keep moving. Roughly two pages in, the walls started coming faster, which is normally the signal that the scene has reached the load-bearing part of the world.

It stopped at sixty-five minutes, short of the timebox, at the point where the writing turned into an attempt to finish the scene rather than to find anything out. Three and a half pages. Every one of them deleted afterwards.

The Missing list

Things the world had simply never decided:

• What does Maren call Eli now?
• Does she eat, sleep, or breathe? Or does she only not age?
• How long did the signal take to work — one night, or years of building?
• Does Maren know what she is? Does she know she came back?
• Who else in town has actually seen her since?
• How far is the keeper's house from the lantern room?

Six items, none of them a crisis, all of them things a script would have had to answer eventually — and would have answered inconsistently across six episodes if six different answers had been improvised in six different scenes. That is the ordinary yield, and on its own it justifies the session.

The Broken list

One item. It was enough.

**Eli's twenty years at the light cannot coexist with his age and Maren's drowning.** The scene needed him to have been keeper when Maren drowned — she drowned off the point below the light, and his failure to reach her is the wound the whole show sits on. Maren drowned eleven years ago. Eli is 47, so he was 36. But twenty years of keeping the light puts his start at 27, which works — until the scene also needed him to have arrived at Cape Sorrow *with* a young family, and Maren to have been born there.

Written out as a sequence, it does not close. The arithmetic had been checked before, in isolation, and it passed in isolation. It only failed when a scene forced all three facts to be true in the same room at the same moment — which is precisely the property that makes a scene the right instrument and an outline the wrong one.

Nobody had been careless. Each fact was decided deliberately, months apart, and each one was individually correct.

Resolving, then deleting

Missing items were decided and written into the record, not into the draft. Maren uses his name — she has not called him Dad since she came back, and he has never once corrected her, which turned out to be the best thing the session produced. She does not eat or sleep. The signal took nine years to build. The house is a four-minute walk, uphill.

The Broken item forced a real fork: change the scene or change the canon. The canon changed. Twenty years became eleven — he took the light the year Maren was born, and she drowned nine years into it. This is a deliberate retcon rather than an accidental contradiction, and the distinction matters: it was made once, in writing, before anything was built on top of it.

Then the three and a half pages were deleted. Genuinely deleted. There was a good exchange in there and it did not survive, which is the correct outcome — keeping it would have made the next session an exercise in protecting that exchange.

What it cost and what it caught

Sixty-five minutes, three and a half discarded pages, six decisions made, and one premise-level contradiction found and fixed before a single script page existed.

Had it not been caught there, the likely discovery point was a table read, or a continuity note in episode four, at which point the arithmetic would have been load-bearing for several scripts and the cheap fix would no longer have been available. The version of that cost curve across a production is in why plot holes survive rewrites.

The honest caveat: this pass caught one class of error well and would not have caught others. A zero draft tests what a scene instantiates, so anything that only appears across a long stretch of reading — voice drift, escalation that stops escalating — is outside its range. The map of what it does and does not catch is in continuity failure modes.

Frequently asked questions

How long should a zero draft of a pilot scene take?

Timebox ninety minutes and expect to stop earlier. This session ran sixty-five minutes and produced three and a half pages. Stop when you notice you are writing to finish the scene rather than to find something out — finishing earns you nothing.

What if the zero draft does not break anything?

That usually means the target was chosen badly rather than that the world is sound. A scene that collides with little validates cleanly and teaches nothing. Pick the scene where the most established facts have to be true at once, or the one you have been avoiding.

Do you really delete a scene that had good writing in it?

Yes, and this session had a good exchange that did not survive. Keeping it would have made the next session about protecting that exchange rather than stress-testing the world. The freedom to break your own canon comes from knowing none of the prose is being graded.