# Using AI to Write My Posts.

## Abstract.

For my 100th Anniversary Post, I document the workflow I use to create content with AI assistance. At the heart of the process are three key components:

1.  **Template files** — structured style guides that define the layout, sections, metadata, and conventions for each document type.
    
2.  **A prompts file** — a queue of prompts that the automation script feeds to Hermes Agent, one by one.
    
3.  **An automation script** — the queue processor (`queue_v5.0.1.py`) that orchestrates DwarfStar (the inference server) and Hermes Agent (the AI assistant) to process prompts from the queue file.
    

The result is a repeatable, scalable pipeline for producing consistent, high-quality posts without manual effort.

* * *

## An Introduction.

Writing blog posts and technical content is a core part of sharing knowledge. Over time, I developed a system that moves from a raw idea to a finished, published post through a structured, AI-assisted workflow. The key insight is that AI does not replace the writer — it amplifies the writer's intent when guided by clear templates and precise prompts.

This post itself was created using the same process it describes. The prompt that generated it came from the ideas file (`Ideas_v0.5.12.md`), and the structure was dictated by the template (`Template_Posts_v0.5.1.md`). The automation script (`queue_v5.0.1.py`) fed the prompt to Hermes Agent, which produced the content you are reading now.

* * *

## The Big Picture.

The content creation pipeline has four stages:

### Stage 1: Define the Prompt

I maintain an ideas file, called `/path/to/Ideas_v0.5.12.md`, where I brainstorm the concept for every post I want to create. Each entry evolves into detailed instructions of what to write, which source materials to use, which template to follow, and which tone to adopt. As you can tell by the version number, I constantly archive this file when it gets too large and start a new file with the next version number.

### Stage 2: Queue the Prompts

The prompts file lives at `/path/to/prompt.txt` and is processed by the `queue_v0.5.1.py` Python script. The script reads the prompt file, cuts the top prompt from the file, saves the remainder, and passes the cut prompt to Hermes Agent for processing. Prompts are separated by `<<<<` so the queue script can process them one by one.

### Stage 3: Generate the Content

Hermes Agent receives the prompt and produces the content. Because the prompt references a specific template, the AI knows the exact structure to follow — the sections, the metadata fields, the formatting conventions. The output is a complete document ready for review and publication.

### Stage 4: Review and Publish

The generated document is saved to the Posts directory. I review the content, make any adjustments, and publish. Because the template ensures consistency, the review process is fast — I can focus on substance rather than structure.

* * *

## Prerequisites.

To use this workflow, I need:

*   **Hermes Agent** installed and configured (`hermes setup`),
    
*   **DwarfStar (ds4-server)** running as the inference backend,
    
*   **The queue automation script** (`queue_v5.0.1.py`) deployed and accessible from my PATH,
    
*   **A prompts file** (`Prompts_v0.5.11.md`) with entries for each post,
    
*   **Template files** for the document types I want to create, and
    
*   **A queue file** (`prompt.txt`) containing the prompts I want to process.
    

* * *

## Updating my Base System.

Before generating content, I ensure my system is up to date:

```bash
sudo apt clean && \
sudo apt update && \
sudo apt dist-upgrade -y && \
sudo apt --fix-broken install && \
sudo apt autoclean && \
sudo apt autoremove -y
```

This keeps the system stable and ensures all dependencies for Hermes Agent and DwarfStar are current.

* * *

## Section 1: The Role of Templates.

Templates are the backbone of consistent content creation. Each template is a structured Markdown or AsciiDoc file that defines:

*   **Frontmatter metadata** — title, subtitle, author, version, dates, keywords, copyright
    
*   **Section headings** — the logical structure of the document
    
*   **Content conventions** — formatting, code blocks, quotes, lists, call-outs
    
*   **Closing and attribution** — standard sign-off and document details block
    

### How the Template Is Used as a Guide

When I create a prompt for a new post, the prompt explicitly references the template:

> "You will use the `/path/to/Template_Posts_v0.5.1.md` template as my style guide as I create the post..."

Hermes Agent reads the template and follows its structure. The AI knows:

*   Which sections to include (Abstract, Introduction, Big Picture, Prerequisites, etc.)
    
*   How to format the frontmatter YAML
    
*   Where to place metadata, code blocks, and attributions
    
*   The expected tone and style
    

This means every post follows the same proven structure. Readers know where to find information, and I can focus on the content rather than the layout.

### How the Ideas File Is Used as a Guide

The ideas file (`Ideas_v0.5.12.md`) is a curated list of entries that describe each post I want to create. Each entry is a self-contained prompt that includes:

*   The title and topic of the post
    
*   Which template to use
    
*   Which source materials to reference
    
*   Any specific instructions for the AI
    

The Ideas file serves as my editorial calendar. I can add entries whenever I have an idea, and the queue script processes them in order. This decouples idea generation from content creation — I capture ideas when they come, and the automation handles the execution later.

* * *

## Section 2: Other Templates for Other Document Types.

The same methodology extends beyond blog posts. The `/path/to/Templates/` directory contains templates for several document types:

### Business Reports

The template `/path/to/DCNZ_Report_Template_v0.5.1.ad` is an AsciiDoc file designed for structured business reports. It includes sections for executive summary, findings, analysis, recommendations, and appendices. When I need a formal report, I reference this template in the prompt.

### Cashflow Reports

The template `/path/to/DCNZ_Cashflow_Template_v0.5.1.xls` is an Excel spreadsheet that defines the layout for cashflow tracking. While this is not a text document, the same principle applies — the template defines the structure, and the prompt tells the AI what data to fill in or how to generate the report.

### AsciiDoc Files

The template `/path/to/Template_AsciiDoc_v0.5.1.ad` provides the structure for documents that need AsciiDoc formatting. AsciiDoc is used for technical documentation, PDF generation, and content that benefits from advanced formatting features like conditional blocks, cross-document references, and custom roles.

### PDF Files (from AsciiDoc Files)

PDF files are generated from AsciiDoc source files using the AsciiDoc template. The workflow is:

1.  Generate the AsciiDoc content using the `Template_AsciiDoc_v0.5.1.ad` template
    
2.  Convert to PDF using `asciidoctor-pdf` or a similar tool
    
3.  Review the PDF output
    

This gives me full control over the PDF layout while still benefiting from AI-assisted content generation.

### Generic Markdown Files

The template `/path/to/Template_Markdown_v0.5.1.md` is a lightweight template for general-purpose Markdown documents that do not need the full blog post structure. It includes basic frontmatter, section headings, and formatting conventions. This is useful for quick notes, internal documentation, and content that does not need the post template's full section hierarchy.

* * *

## Section 3: The Automation Pipeline.

The full pipeline from prompt to published post works like this:

**I add a prompt** to the prompts `prompt.txt` file, referencing the appropriate template and source materials, that I had stitched together in the Ideas file.

**I use the /queue command** that Hermes Agent supports, and define the prompt delimiters (<<<<) that separate the different prompts in the `prompt.txt` file (which may already contain other queued prompts).

**Hermes Agent runs the /queue command** until all of the prompts in the `prompt.txt` file have been processed:

1.  I launch Hermes Agent (which uses DwarfStar for inference)
    
2.  I write the /queue command into the Hermes input field
    

1.  I follow this command with the location of the `prompt.txt` file
    
2.  I follow the location with an explanation of the delimiters (<<<<) that separate the prompts
    
3.  **I remind Hermes** that each delimited prompt needs to be cut from `prompt.txt` file and processed separately, in order, from top to bottom
    
4.  **I specify the directory** where the generated result(s) are saved
    
5.  **I hit the ENTER key** once all the details are added to the Hermes input field
    
6.  I check the output(s) and fine-tune the results
    

* * *

## The Results.

This workflow has produced tangible improvements in my content creation process:

*   **Consistency**: Every post follows the same proven structure. Readers get a familiar, scannable layout.
    
*   **Speed**: I can generate a complete post in minutes rather than hours. The AI handles the structure and initial draft; I handle the review and refinement.
    
*   **Scalability**: I can queue multiple prompts and process them in batch overnight. The queue automation handles the full cycle — server management, prompt delivery, and shutdown.
    
*   **Quality**: Because the prompts reference real templates and source materials, the AI has rich context. The output is grounded in my actual files, not generic AI filler.
    
*   **Flexibility**: The same pipeline works for blog posts, business reports, AsciiDoc files, PDFs, and Markdown notes — just by changing which template the prompt references.
    

* * *

## An Analysis

The key insight is that AI writing is not about handing a vague idea to a chatbot and hoping for the best. It is about:

1.  **Defining the structure upfront** with a template file that the AI can follow precisely.
    
2.  **Providing rich context** through the prompt, including source materials, file paths, and specific instructions.
    
3.  **Automating the execution** with a queue script that handles the infrastructure so I do not have to.
    

The templates encode years of writing conventions and structural decisions. By passing them to the AI, I transfer that knowledge into every generated document without manual effort. The prompts file acts as both a queue and an editorial calendar — ideas are captured when they arise, and the automation processes them on schedule.

The queue automation script adds reliability: it manages server startup, detects conflicts, restarts between prompts, and shuts down cleanly. I can start the queue and walk away, knowing the script will process every prompt and leave both services in a clean state.

* * *

## In Conclusion

Using AI to write my posts is not about replacing the writer — it is about amplifying the writer's intent through structured templates, precise prompts, and reliable automation. The template ensures consistency, the prompts file captures ideas, and the queue script handles the execution.

This post itself was created through this exact workflow:

*   The prompt came from `Prompts_v0.5.11.md`, starting at line 61.
    
*   The structure came from `Template_Posts_v0.5.1.md`.
    
*   The generation was executed by the queue automation script feeding Hermes Agent.
    
*   The result is the document you are reading now.
    

The same approach works for business reports, cashflow reports, AsciiDoc files, PDFs, and generic Markdown files — each has its own template, and the same pipeline drives them all.

Until next time: Be safe, be kind, be awesome, kia kaha!!

* * *

## Document Details: Using AI to Write My Posts

**Document Subtitle:** Swimming in a Cesspool of AI Slop.

**Document Version:** v0.5.1

**Document Author(s):** Brian King

**Document Attributions:** Brian King (author), Hermes Agent (AI assistant)

**Hash Tags:** #AI #Writing #Hermes-Agent #Templates #Automation #Prompts #Content-Creation #Workflow #AI-Slop

**SEO Title:** Using AI to Write Blog Posts with Hermes Agent and Templates

**SEO Description:** How I use Hermes Agent, structured templates, and queue automation to produce blog posts, reports, AsciiDoc files, Markdown files, and PDFs.

**Document Creation Date:** Monday 24th August 2026

**Last Update:** Monday 24th August 2026

© Copyright 2020-2026 DigitalCoreNZ. All rights reserved.