At Glad Labs, we're building a business with AI, for those who build with AI. As a solo founder operating a "frontier firm" - leveraging LLMs, APIs, and tools to do more with less - I'm intimately familiar with the challenge of consistent content creation. It's easy to fall into the trap of "works on my machine" thinking, and content is no different. You can get a blog post written, but can you get consistent blog posts, aligned with your strategy, without burning out?
We've spent the last few months tackling this problem internally. Not just generating ideas, but surfacing topics with the potential to resonate with our audience of indie developers and tinkerers focused on AI/ML, gaming, and PC hardware. Our goal isn't to replace the editorial process, but to drastically reduce the initial cognitive load - to move past staring at a blank screen, unsure where to start.
The Problem with Traditional Brainstorming
Traditional topic generation relies on manual brainstorming, keyword research, and analyzing industry trends. While valuable, these methods are time-consuming, subjective, and prone to blind spots. As the pace of innovation in AI/ML and related fields accelerates, keeping up with relevant topics and identifying angles that haven't been exhausted becomes increasingly difficult.
Our Approach: LLM-Powered Topic Seeds
We adopted a system using Large Language Models (LLMs) - specifically, Claude Code - to generate topic seeds. These aren't fully-formed blog posts, but rather concise prompts designed to spark ideas. The core principle is to feed the LLM a curated knowledge base (our existing content, industry reports, and relevant documentation) along with specific instructions on the desired topic domain and audience.
Here's a simplified example of a prompt we use:
Generate 5 blog post topic ideas for indie game developers interested in using AI for procedural content generation. Focus on practical, hands-on techniques and tools they can implement today. Prioritize topics related to runtime generation rather than pre-baked assets.
This generates a list like:
- "AI-Powered Dungeon Generation with Python and Pygame"
- "Creating Dynamic Music Scores Using Machine Learning"
- "Generating Realistic Character Dialogue with LLMs"
From Seed to Strategy: Filtering & Prioritization
The raw output from the LLM is rarely publish-ready. We've built a filtering and prioritization layer on top of the topic seed generation. This involves:
- Keyword Analysis: Checking for search volume and competition using tools like Ahrefs.
- Content Gap Analysis: Identifying topics where existing content is lacking or outdated.
- Relevance Scoring: Evaluating the alignment of each topic with our overall content strategy and target audience.
This layer isn't fully automated. A human editor still reviews the prioritized list, refines the topics, and adds their own insights. But the automation significantly reduces the initial workload.
The Tech Under the Hood
Our system leverages a combination of tools and technologies. We use Terraform to manage the infrastructure, and a CI/CD pipeline to automate the process. The LLM interaction is handled via API calls to Claude Code. The entire system is designed to be scalable and repeatable, allowing us to generate a consistent stream of topic ideas. We built this on my NIGHTRIDER workstation -- Ryzen 9 9950X3D, RTX 5090, 64GB of RAM.
Beyond Topics: Outlining with AI
We're now expanding this system to include automated outlining. The LLM can take a topic seed and generate a detailed outline, including suggested headings, subheadings, and key points. This further streamlines the content creation process, making it easier for our writers to produce high-quality, informative blog posts.
The Future of Content at Glad Labs
We believe that AI-powered automation is the key to scaling content creation without sacrificing quality. By combining the power of LLMs with human expertise, we can consistently deliver valuable content to our audience, establish Glad Labs as a thought leader in the AI/ML space, and ultimately, build a sustainable business. The initial results have been promising - we're seeing a significant increase in content output and engagement. And more importantly, it's freeing up time for us to focus on what we do best: building cool stuff with AI.
