AI Refine
It does not write your posts. It refits the one you wrote for each place it is going.
Generated-from-nothing social copy reads like generated-from-nothing social copy. The useful job for a model here is narrower: take your draft and make it fit LinkedIn, then X, then Bluesky, respecting each platform’s real limits and habits. You approve every variant.
Credits included on paid plans.
Capabilities
What you get with AI assistant.
- One draft in, one variant per channel
- A multi-platform refine is a single model call that produces a tailored version for every channel you targeted, rather than the same text padded or truncated to fit.
- It sounds like you, not like a model
- A voice profile keeps the variants in your register. The input is your writing, so the output has something real to imitate.
- Credits are reserved before generation
- The cap check and the increment happen in one atomic statement, so concurrent requests cannot both slip under the same limit. If the provider fails, the reservation is refunded.
- You review everything
- Nothing a refine produces is scheduled on your behalf. Variants arrive in the composer for you to accept, edit, or discard per channel.
The narrow job
There is a version of "AI for social media" that writes your posts. It produces text that is fluent, on-topic, and completely forgettable, because it has no information about your work that you did not already have.
The job worth automating is the boring one. You wrote 1,400 characters that read well on LinkedIn. Getting that onto X means restructuring around a much tighter budget; getting it onto Bluesky means 300 graphemes and a different hashtag convention; getting it onto Instagram means a caption that assumes the image is doing the work.
That is transformation, not creation, and the input is genuinely yours.
What it costs
| Refine | Credits |
|---|---|
| Single platform | 2 |
| Multi-platform | number of channels + 1 |
Multi-platform is priced as one call plus one per channel because that is what it is — a single model call producing a variant per channel, rather than N independent calls. Refining for four channels costs 5, not 8.
Your plan sets the monthly allowance. Free workspaces have none.
Why the metering detail matters
Credits are reserved before generation, in a single statement that checks the cap and increments the counter together.
The naive version reads your usage, compares it to the limit, then writes the new total. Two requests arriving at once both read the same number, both decide there is room, and both proceed — so a hard cap is not actually hard. Doing the check and the increment as one conditional update removes that race.
The other half is the refund. If the model provider errors after a reservation, the reservation is released. A provider outage should not quietly eat a day of your allowance.
This is the same reason nothing else in Postonus gates spend with a read-then-write check. A limit that can be beaten by two simultaneous requests is not a limit.
Voice profile
A refine reads your existing posts to stay in your register, which is what keeps the output from drifting into the flat, upbeat, faintly American cadence that makes AI text obvious.
It is not magic and it is not a personality transplant. If your draft is dull, the variants will be dull in your voice. The model is matching how you write, not deciding what is worth saying.
What it will not do
- It will not schedule anything. Variants land in the composer.
- It will not invent statistics or claims to fill space.
- It will not silently drop content that does not fit a shorter platform — restructuring for a tight budget is visible in the variant you review.
For turning a post that already performed into content for a channel you are trying to grow, see auto-repurposing, which starts from your published winners rather than from a fresh draft.
Found this useful? Pass it on.
Questions
Frequently asked questions Everything the first evaluation usually needs.
- Does the AI write my posts for me?
No. You write the post. AI Refine re-fits your draft into each platform’s native shape — character limits, tone, hashtag norms — and you review every variant before it publishes.
- How are AI credits counted?
A refine on a single platform costs 2 credits. A multi-platform refine costs the number of channels plus one, because it is one model call producing a tailored variant for each channel you target.
- What happens if I run out of credits?
The refine is refused with a message naming your plan’s allowance. Credits are checked and reserved before generation starts, so you are never charged for a call that did not happen.
- What if the model provider fails mid-generation?
The reservation is refunded. A failed generation does not cost you credits.
- Are AI credits included on the free workspace?
No — a free workspace has no AI credits. It can still write, plan, and hold dated drafts; refining needs a paid plan.
- Can I edit what it produces?
Yes, per channel. A variant is a starting point in the composer, and editing one channel’s version does not touch the others.
Keep reading
- Auto-repurposePoint a Growth Autopilot at a platform you already win on and a channel you want to grow. Postonus mines your top posts and drips reshaped drafts in for review.Read
- Multi-platform publishingWrite once and publish to 16 platforms, with per-channel overrides where the platforms differ. Limits are enforced as you type, from the rules workers use.Read
- Queue & reliabilityA queue that flags a post at risk before its publish time, refuses to send something an hour late, and turns a crashed worker into a retryable failure.Read