PostPilot
for X
What's out
The Voice Profile box finally explains itself, a no-card Pro trial landed, and the compose box still scores media, names AI slop, and shows a reach range from your own posts — not a scheduler, not account access, not writing the post for you.
STILL LOCAL · STILL NO X LOGIN · YOU HIT SEND
Just shipped (v0.9.1) Sep 3, 2026
What actually goes in a Voice Profile
The Voice tab's text box used to be a blank prompt: "paste your voice profile." Now it explains what each section (niche, hook preferences, length, tone) is for and why, makes clear it's optional — five pasted posts already gets you Voice Match — and has a "Load Example" button that drops in a real, worked profile you can edit into your own. It's a fully populated example, not a placeholder: it exercises every signal PostPilot's parser actually reads.
v0.9.0 Sep 2, 2026
7-day Pro trial, no card required
Try the full Pro tier — 3 AI rewrite variants, Voice Match, the Full Breakdown — for a week without entering a payment method. Starts once, from Settings; it doesn't reset by asking again.
A preview of what the Full Breakdown looks like
Free users with 5+ scored posts now see a preview of the Pro Analytics breakdown instead of a locked, empty tab — so the upgrade decision is based on what's actually on the other side of it.
Rewrite-scoring fixes
Two bugs that could make an AI rewrite score no better than the post it was rewriting: a hook-length cliff in the scoring curve, and a fabrication check that was stripping credit too broadly. Both fixed at the source, not patched around.
v0.7.5 Aug 28, 2026
Link penalty softened, on the evidence
We went looking for a real link penalty in X's published ranking algorithm and didn't find one — checked the visibility filters, the bot rules, the spam classifiers. The only two public link weights are both positive. So instead of guessing a bigger number, we moved our own conservative default closer to neutral (-3 → -1). Your own learned link data still overrides this the moment you have enough posts either way.
v0.7.4 Aug 28, 2026
Video scored separately from images
X's own published ranking algorithm (xai-org/x-algorithm) weights a video open differently from a photo expand — so now we do too, instead of folding video into the image score like every version before this one. Your video posts get their own cold-start default and, once you have enough of them, their own learned boost in Settings → Analytics. The image and link priors stay untouched — the public algorithm doesn't publish a link penalty, so we're not inventing one.
v0.7.2 Aug 24, 2026
Weekly score trend chart
Settings → Analytics now plots your last 13 weeks of scores next to your 7-day average, so the two numbers always agree. It stays visible between posting weeks and tells you honestly when there's not enough recent history yet — built entirely from scores stored on your device.
v0.7.1 Aug 19, 2026
New hook only
Swap the first line, keep the body you already wrote. Free gets one opener; Pro gets three angles. Replies stay a full rewrite — a reply isn’t a first-line-plus-body job.
Voice Match shapes rewrite layout
Rewrites already used your vocabulary and person. They now match how you actually break lines — short lines, stanzas, fragments — even when this draft is a single block.
Start here in Settings
Scoring works with no setup. Free: go write. Pro Voice Match: add five posts, then Analyze. Governor and hook knobs sit under Advanced. Leave them alone unless a score feels wrong.
v0.7 Aug 17, 2026
Import your X Analytics CSV
Settings → Analytics. Download the tweet-level export X already gives Premium users (Tweet id / Tweet text — not the by-day summary). Drop it in. Insights fill in without waiting for 20 scraped posts. Profile scraping still runs in the background.
Images and links move the score
Attach a photo or paste a link and the badge updates. Same as the text — scored on your device, using what actually worked on your account when we have the data.
“Is now a good time?” + a pre-publish checklist
A last look before send: hook, flagged phrases, length, media. Warns. Never blocks.
AI-slop warnings
When a draft sounds like ChatGPT — bland templates, telltale phrasing — PostPilot names it. The algorithm punishes robotic copy. So do we.
A reach range from your posts
Not a made-up “14,200 people.” A range based on how similar posts of yours actually did, with the sample size shown. Hidden until there’s enough data to be honest.
Compare two of your drafts
Short vs long. Question vs claim. See which one scores higher and why, side by side.
Does this post invite a reply?
Stopping the scroll and starting a conversation are different jobs. You’ll see if the draft is a broadcast or a prompt — without rewarding “thoughts?” bait.
A self-reply prompt after you post
One suggestion for a follow-up tweet. You still paste and send it. PostPilot never posts for you.
Share this
Want something that isn’t here? Reply on X or open an issue. Scheduling is still off-thesis — PostPilot never posts for you.