diff --git a/docs/re/f6-unit1-238-is-an-exit.md b/docs/re/f6-unit1-238-is-an-exit.md
new file mode 100644
index 00000000..124ab97d
--- /dev/null
+++ b/docs/re/f6-unit1-238-is-an-exit.md
@@ -0,0 +1,62 @@
+# F6 unit 1 — `238…250` on the sweep is an **EXIT**, not a start
+
+**Status: ✅ decoded ⟨disc⟩.** 2026-09-02. One unit, per `PROTOCOL.md` §"Work in
+units a human can check in a minute".
+
+**Question:** on the sweep elements, is `238…250` a start or an exit?
+**Look at:** their declared alpha. Rising 0→255 = start; falling 255→0 = exit.
+**Not covered:** what the game does at runtime; the plate's timing; F5.
+
+---
+
+## The answer
+
+`sylpheed-cli screen info --build 4 --geometry`, GP_TITLE:
+
+```
+pteff03.t32    0:a=0   70:a=0   100:a=255   238:a=255   250:a=0
+pteff03a.t32   0:a=0   70:a=0   100:a=255   238:a=255   250:a=0
+ptcopyright    0:a=0  138:a=0   160:a=255   238:a=255   244:a=0
+```
+
+**The sweep's parents rise `70→100` and fall `238→250`.** So `238…250` is a
+**fade-out**, and the brief's own warning was right: it is the same shape
+`ptcopyright` uses, and `ptcopyright` starts nothing.
+
+> **The sweep is declared visible from `t = 100` — 136 units BEFORE the plate
+> reaches full alpha at 236 — and it is fading out by 238.**
+
+## 🔴 So the declared data contradicts the report, and that is the finding
+
+The human reports the sweep *"only starts when the plate appears"*. On the parents'
+alpha alone it should already be on screen for the whole build-in. **Both can be
+true**, and the reason is the next unit:
+
+`ptloop01`/`ptloop02` are `.rat` **leaves with their own timeline**, and that
+timeline is a **translation**: `ptloop01` runs `x = −639 → 1521` over a **600-unit**
+loop ([`data/sweep-leaf-ramp.txt`](data/sweep-leaf-ramp.txt)). At its own `t = 0`
+the sweep sits at `x = −639` — **off the left edge of a 1280-wide screen.**
+
+**The parent's alpha says "drawn". The leaf's position says "where".** A quad can
+be fully opaque and entirely off-screen, and this one starts that way.
+
+## Next unit (not done)
+
+**When does the leaf's clock start, and where is the sweep at the moment the plate
+appears?** That decides whether the human is watching an alpha gate (answered: no)
+or a quad travelling on screen. It needs the leaf's clock origin, which is the
+same open question as the title's own — so it may need the capture, not the disc.
+
+⚠️ **Do not act on this unit alone.** It says `238` is not a start. It does **not**
+say the port should show the sweep from `t=100` — if the leaf is off-screen there,
+both renderers can be "correct on alpha" and differ entirely in what is visible.
+
+## Refutation attempt
+
+**Target:** the brief's lead that *"236–238 is a synchronisation point in the
+declared data and a human just reported a behaviour change there."*
+
+**Result: the synchronisation is real, the reading of it is REFUTED.** `pteff02` is
+keyed at 236 and `ptlogo_back2eff`/`ptcopyright` at 238 — so 236–238 *is* a shared
+instant. But for the sweep it is where things **end**, not begin. A cluster of
+keyframes marks a moment; it does not say which direction anything moves through it.