How I Built an Automated Content Factory with an AI Assistant
You don't need to be a coder to build things anymore.
By Daniel Zirpoli (aka AIficionado)
The Problem: The Manual Content Grind
The process of creating a high-quality blog post from a simple idea is incredibly time-consuming. It's a journey filled with manual, repetitive steps.
- Finding a relevant, interesting topic.
- Drafting an initial take on the subject.
- Structuring that draft into a coherent blog post.
- Polishing it meticulously for publication.
My Goal was simple: Could I automate this entire workflow, from discovery to final draft?
The Solution: A Team of AI Specialists
Instead of one giant, monolithic program, I created a team of three specialists that work together. This is known as a "microservices" architecture, where each program does one job really well. This design made it much easier for me to make changes to one part without breaking another.
The Core Workflow
kurator
(The Researcher)
laruta
(The Project Manager)
blogplus
(The Writer)
Meet the Team
Job: To read the news and find the most interesting articles for our blog.
It searches for news using specific keywords I gave it. An AI model then reads each article and scores it based on criteria I defined in a simple prompt file, such as relevance and engagement potential.
Key takeaway: I didn't have to code a complex scoring algorithm. I just had to *describe* what makes a good article.
Job: To take the article `kurator` found and turn it into a unique, polished blog post.
This isn't a single step; it's a multi-stage writing process, just like a human writer would use:
- **Essay Draft:** First, it writes a thoughtful essay to understand the core concepts.
- **Outline:** Next, it creates a logical structure for the blog post.
- **Elaboration:** It then writes the full post, focusing on adding depth and detail to meet a target word count of 800-1200 words.
- **Final Polish:** Finally, a "Hemingway" specialist refines the text, making it clear, concise, and easy to read, and adds the final tags and disclaimers.
Key takeaway: The quality comes from breaking the task into smaller, specialized steps.
Job: To manage the entire workflow and make sure the specialists work together smoothly.
It runs `kurator` first. It then takes the winning article, scrapes the content, and hands it off to `blogplus` in the correct format. It even handles all the technical setup, like ensuring all the necessary code is downloaded and ready to run.
Key takeaway: This orchestrator is what makes the system a true, hands-free pipeline.
The "Magic" Isn't Code, It's Instructions
This is the core message. How did I, a "non-coder," build this? I didn't write thousands of lines of code. I wrote **instructions in plain English.** The "source code" for the AI's behavior lives in simple text files.
My primary job was to be a good **manager and communicator**, clearly defining the goals for my AI assistant.
Example from `prompts/write_blog.txt`:
# GOALS
- Transform the provided materials into a compelling blog post...
- Write in a clear and concise style...
- Perfectly adopt the specified "Tone of Voice."
- Naturally weave the specified "Keywords" into the text for SEO.
- Tailor the language and examples to the "Target Audience."The Result: A Human-AI Partnership
The final output is more than just an automated article; it's the product of a collaboration. The system *does* automate the creation process, but the highest quality result comes from a partnership.
The AI handles the heavy lifting of drafting and structuring, but a handcrafted final touch from a human editor is what ensures the quality. This system doesn't replace the creator; it empowers them to be a director and editor, focusing on what matters most: the ideas, the message, and the quality that automation alone can't achieve.
This Is The Final Blog Post Title
Published on August 1, 2025
Full Automation: The Publishing Pipeline
This isn't just a set of scripts I run manually. The entire system is deployed as a hands-free, automated pipeline that publishes content without any intervention.
GCP Cloud Scheduler
Every two days, a scheduled job automatically triggers the `laruta` orchestrator.
GCS Bucket
The finished blog post is saved to a Google Cloud Storage bucket.
GCP Watcher Job
Another cloud job constantly monitors the bucket for new files.
GitHub Push
When a new post appears, it's automatically pushed to a dedicated GitHub repository.
Netlify Publication
Netlify detects the change in the GitHub repo and automatically deploys the new post to the live blog.
See It Live: blog.zirpo.deWhat's Next? The Road Ahead
1. Fully Automated Image Generation
Currently, the system generates a prompt for a cover image, but I create the image manually on my local machine. The next step is to make this hands-free.
- **The Plan:** Use a PUB/SUB model. The cloud job will publish the image prompt to a topic.
- My home server will subscribe to this topic, generate the image using the prompt, and upload it.
- This also involves refining the image prompts, styles, and LoRAs to ensure consistent, high-quality visuals.
2. Enhance the Writing Agent
The goal is continuous improvement of the writing quality.
- The focus is on enhancing the AI's ability to produce even more insightful and engaging articles, while ensuring the readability (Hemingway rating) remains excellent or even improves.
Conclusion: The Future is Augmentation, Not Obsolescence
This project proves that AI's true power isn't in replacing humans, but in augmenting our abilities. It's a tool that changes the nature of our work for the better.
This doesn't make writers or coders obsolete. It makes them architects and quality editors. A purely automated result won't be as good as one that's handcrafted. The key skill is shifting from pure execution to strategic direction and refinement.
The future of creative and technical work is a partnership. We provide the vision and the critical final touch; AI provides the tireless execution to help us achieve our goals faster and at a greater scale than ever before.