Make a brand's posts by code
Every post here is drawn by code, not generated by an image model. No GPU, no per-image bill, no queue, and the hundredth variant costs exactly what the first one did. That single fact is what makes a real posting cadence affordable.
Open the studio →The one idea: a post is a stack of layers
Not a picture. A document: a size, a background, and an ordered list of layers painted from the bottom up. Once that clicks, everything else in the studio is obvious.
Three sizes cover the platforms: square at 1080 by 1080, portrait at 1080 by 1350, and story at 1080 by 1920. You can also give exact numbers.
Where images come from
Four sources, and they mix freely inside one post:
The two tabs
The feed, and why it opens half-closed
The grid shows what is still to come on top and what is already on the profile underneath, in that order, because that is where they will actually sit once the queue drains. The point where the two meet is the seam, and it is the only part most people need to look at, so the grid opens showing just that: the next three out and the last three up. The arrow above it opens the rest.
The two panels above the grid
Both read from the machine that actually runs the work, never from a local copy, and both refresh themselves when a run finishes.
Driving it from chat
The studio has an MCP, so your AI can build a post layer by layer, apply your brand, fill a queue and read the panels back to you, without you touching the interface. The editor and the chat are two doors into the same document.
The shortest path to a real cadence
The engineering behind the engine, the bots and the traps they cost us lives in the journal Run a Social Media Autopilot, on the journals page.