Grow the reach of your posts on X.
An AI agent ships one change at a time, reads the result in the X API, and keeps only what moved impressions without resonance falling.
macOS · Runs on your ChatGPT Plus sign-in
Impressions per post
- Measured in
- The X API, via the xurl CLI
- Also reads
- resonance (likes + bookmarks per 1k) · profile clicks · link clicks
- Guardrails
Resonance must not fall while impressions rise
No post asks for replies or retweets to buy reach
Two runs. Two verdicts.
Every run states a hypothesis before it acts, and the real number settles it afterwards. Here are two of them side by side. One held. The other did not, and that is the half that stops the next run spending another month on the wrong thing.
Hook quality
Format
“Opening original posts with a specific number from a shipped result, instead of a general claim, lifts average impressions per post from 5,200 to 7,000 or better across the next 10 posts, without resonance falling below 9 per 1,000.”
“Quote-tweeting the highest-velocity trending posts of the week, instead of posts already in-topic, lifts average impressions per post from 7,400 to 12,000 or better across the next 10 posts, without resonance falling below 9 per 1,000.”
Rewrote the opening line of the next ten original posts to lead with a specific number from a shipped result.
Quote-tweeted the ten highest-velocity trending posts of the week instead of posts already in-topic.
- Impressions per post (avg)
- 5,2007,400
- Likes per post (avg)
- 3861
- Bookmarks per post (avg)
- 915
- Resonance (per 1,000 impressions)
- 9.010.3
- Impressions per post (avg)
- 7,40015,800
- Likes per post (avg)
- 6154
- Bookmarks per post (avg)
- 156
- Resonance (per 1,000 impressions)
- 10.33.8
Impressions cleared the bar and resonance rose instead of falling, so the hook was the constraint, not the audience. Hook quality moves toward proven, and the next batch of originals leads with a number too.
Impressions more than doubled and resonance fell to a third, so all that extra reach did not care. The guardrail trips, and quote-tweeting outside your topic drops back to candidate.
Hook quality
“Opening original posts with a specific number from a shipped result, instead of a general claim, lifts average impressions per post from 5,200 to 7,000 or better across the next 10 posts, without resonance falling below 9 per 1,000.”
Rewrote the opening line of the next ten original posts to lead with a specific number from a shipped result.
- Impressions per post (avg)
- 5,2007,400
- Likes per post (avg)
- 3861
- Bookmarks per post (avg)
- 915
- Resonance (per 1,000 impressions)
- 9.010.3
Impressions cleared the bar and resonance rose instead of falling, so the hook was the constraint, not the audience. Hook quality moves toward proven, and the next batch of originals leads with a number too.
Format
“Quote-tweeting the highest-velocity trending posts of the week, instead of posts already in-topic, lifts average impressions per post from 7,400 to 12,000 or better across the next 10 posts, without resonance falling below 9 per 1,000.”
Quote-tweeted the ten highest-velocity trending posts of the week instead of posts already in-topic.
- Impressions per post (avg)
- 7,40015,800
- Likes per post (avg)
- 6154
- Bookmarks per post (avg)
- 156
- Resonance (per 1,000 impressions)
- 10.33.8
Impressions more than doubled and resonance fell to a third, so all that extra reach did not care. The guardrail trips, and quote-tweeting outside your topic drops back to candidate.
What it can try on your account.
A lever is a class of change the agent believes moves impressions and resonance together, not one at the other's expense. These four ship with the template.
Topic selection
Draw each topic from a shipped result or a named audience pain instead of a general take.
Hook quality
Lead the first line with a concrete number or result instead of a general claim.
Format
Choose original, reply, or quote-tweet by which one drives impressions versus link clicks for that piece of content.
Reply targeting
Reply into posts still gaining traction on outcomes or measurement, not launch announcements, ranked by impression velocity.
And whatever your posts point to next
Four is where the bank starts. As verdicts land, the agent tests levers your own posting history suggests and drops the ones that keep trading resonance for reach, so what it tries next is earned by your account, not fixed by a template.
Point it at your account.
What you connect
- X API access via the xurl CLI, authenticated with your OAuth2 user context. Impression counts on your own posts are only readable as the authenticated account, so this connects as you, not as a public reader.
- The analytics behind your bio or in-post links, optional. It confirms which posts send real clicks instead of just impressions.
When you will see something
A single post's impressions are largely settled within about a day, faster than any other template here. But one post is noise: a lever needs several posts to run before its verdict means anything.
macOS · Apple silicon and Intel · No credit card