AI-Driven Layouts for Blog Posts: One Article, Four Reading Modes
I rebuilt the āfeelā of my blog by separating content from layout. See how a Payload CMS block tree plus an AI layout engine creates multiple reading modesāwithout AI authoring the post itself.
Post Title: AI-Driven Layouts for Blog Posts: One Article, Four Reading Modes
Post Excerpt: I rebuilt the āfeelā of my blog by separating content from layout. See how a Payload CMS block tree plus an AI layout engine creates multiple reading modesāwithout AI authoring the post itself.
I've been quietly updating ProjectCoops over the last couple of months, and the biggest change is not in the codebaseāit's in the way the blog feels.
If you've visited **TCoops.ca** in the last few weeks, you've already seen the new homepage, design, and structure. What you haven't seen yet is the part I've been working on before the launch but was not ready to share yet! I am excited to share the blog and its AIādriven layout system - the one you are reading right now.
This isn't AIāauthored content. The words are still mine. But the way the post is framed is now handled by an AIādriven layout engine that sits on top of a simple blockābased editor. Each post is defined once as a small tree of meaningful blocksāheadings, paragraphs, quotes, code samples, and imagesāand then reused across four distinct layout modes:
- Editorial, for longāform, scrollable reading
- Immersive, for cinematic, scrollādriven storytelling
- Code Editor, where the post feels like reading a live codeāfile inside a VSāCodeāstyle frame
- AI Remix, an experimental layout built from cards, timelines, and interactive blocks
You can read the same post in four different modes, and notice how the same ideas start to feel like different experiencesānot because the content changed, but because the layout did.
How the system works (in plain terms)
I start by writing the content the same way I always do, then use AI to help turn rough notes into cleaner, more structured text. That's purely a tooling choiceāI still own every sentence and every idea.
The next step is to import that content into a custom block editor based on Payload CMS, where it's converted into a small, reusable block tree. Each block carries its own type and metadata, which keeps the model explicit and easy to reason about.
When the content feels ready, the AI Variant Generator runs over the block tree and produces three layout variants (soon to be four). Each variant is a reāshaped version of the same content, translated into a layout patternālike editorial, immersive, codeāeditor, or an experimental remix mode. The AI doesn't invent new ideas; it interprets the existing blocks and applies layout rules that match the chosen mode.
Once the structure is in place, the system can apply an AI suggestion step that looks for opportunities to group, highlight, or reāinterpret parts of the post. For example, it can suggest turning a series of short paragraphs into a cardāstyle section, or pairing a strong quote with a fullābleed image. Those suggestions are always optional, so I can choose what to keep and what to refine manually.
On the frontāend, those variants are rendered using a single layoutārenderer that maps each block to the right UI component. That means I can change the layout rules or add a new layout mode without touching the original content.
.png&w=2048&q=75)
Why this isn't about AIāauthored content
The core idea here isn't to replace my writing with AI. It's to treat layout as a separate, AIāorchestrated layer on top of explicit, humanāauthored content. The AI never writes the core message. Instead, it helps decide how that message is grouped, emphasized, and framed visually.
Layout is an AI-orchestrated layer on top of explicit, human-authored content. The AI does not write the core message.
For TCoops, this update is a small but intentional step toward treating the site as a lab for AIāassisted design. Every layout mode is a different way of reading the same post, and the AI is only there to help shape the experience, not the story. There are likely bugs and I know there is room for optimization, but as a first concept and deliverable, it operates great! I hope you enjoy it as much as I did building it!
