Coach
Score the post before it ships
Estimates Phoenix action probabilities from the text, then applies the published production weights. Micro, short, long, article, and thread are not the same candidate. The Chrome extension does the same thing on x.com.
Reach index
75A
Scene · Micro
One breath. Reply/quote or it dies.
Scene post. Fat-tail OC. Lands if the room already has the context. Not a template. Scoring on reply (5 × p≈21.2%).
Do this before you post
- now
Do not explain it in the same post
The replies exist because the thought is unfinished. Spell it out and you get a blog sentence. Answer in the replies — each one is weight 5.
ReplyWeight / QuoteWeight — scene prior
- next
This only hits a room you already trained
Fat-tail OC. Phoenix scores per viewer. Mutuals who have been eating your last 20 posts will reply. A cold OON viewer (×0.75) sees a fragment. Don't industrialize this. If you sell to operators, steal the one-breath shape — not the in-joke.
OonWeightFactor 0.75 + per-viewer Phoenix
- now
Cut it until it feels unfinished
The NK-shape is one breath. A paragraph of context is Operator voice wearing a Scene costume.
Scene prior vs dwell-on-thread
- next
Stay for the conversation, don't escalate
Loaded topics lift reply and quote priors. They also lift NotInterested and Mute a little. The published report head is −234 — do not turn this into an insult thread.
ReplyWeight vs NotInterested / Mute / Report
Phoenix predicts P(action) per viewer from their engagement history. Outrank estimates those probabilities from the draft, then multiplies by the published production weights. It is a coaching instrument, not the live ranker.