Social content and tool workflow
Turning an AI tool takeaway into something worth sharing.
An exploration of social content and AI tooling with visible progress, revision history, and human approval at two stages.
Chose content direction, visuals, and human approval steps; collaborated with AI coding agents on workflow exploration.
Python, content dashboard, workflow state records, PNG carousel output
Instagram content and review prototype with eleven states and revision history
2026-10-10
The starting point: a takeaway still needs shaping
After using an AI tool, I often have an observation I want to share. Turning it into Instagram content requires choosing an angle, organizing the points, editing the words, and arranging a readable sequence of images. This work began with that content need and an exploration of tools to support it.
I also wanted to understand where a piece stood, which version had been approved, and what changed along the way. That context makes a draft easier to resume and gives AI-generated material a clear place for review.
The decisions: approve the framework and the finished content
I chose the visual direction and retained two human approval steps: one for the content framework and one for the finished result. The first checks the intended message. The second brings words, visuals, and reading order together before a publishing decision.
The workflow uses eleven states and one authoritative state file. Rework and Undo preserve history, and some transitions remain manual. Being able to inspect and revise a result is part of the prototype’s design.
The implementation: keep the output available for review
IG-Post-Assistant includes dashboard, metrics, and publishing modules. It also contains seven PNG slides and a posting record for a July 6, 2026 self-introduction. Those artifacts provide a concrete content example to inspect.
I explored the tooling with AI coding agents while keeping content and visual choices under human review. Scheduling has not passed acceptance, and the metrics provider has not been tested with a real token. The modules therefore do not establish fully automated publishing or performance tracking.
The evidence: produced content with validation still ahead
The evidence supports finished content, a posting record, and a defined review workflow. The public account offers an entry point to the shared material. I have no measured efficiency gain or reach growth to attribute to the tooling.
The next step is to validate the connections between production, approval, and publishing while preserving the reasons for revisions. The goal is a process I can follow and content whose quality I can personally check.
Have a thought? Let’s talk.
Find me on Instagram to talk about this topic.
Talk with Kevin on Instagram →