To be clear, this isn't meant to replace or compete with Simon Travaglia's BOFH stories. those are classics and can't be replicated.
This is an experiment exploring whether AI can maintain a consistent character voice across episodic content, using BOFH as inspiration for the tone and setting.
The current episodes are just the starting point.
The plan is to open this up to the community - people can submit episode prompts via pull requests, and collectively influence what gets generated and even influence the character in the future.
The AI becomes more of a "character engine" that the community directs, rather than autonomous content creation. So the quality and direction should improve as humans decide what stories are worth telling.
Right now its admittedly rough, it's a technical proof-of-concept for character consistency as much as anything. But the vision is community-driven storytelling where people have real influence over the content.
Which aspects feel boring specifically? Curious what would make it more interesting to you.
aiofh•3mo ago
I built an AI system that generates episodic sysadmin stories in the style of BOFH (Bastard Operator From Hell), with a persistent character called "The Operator" who maintains voice consistency across episodes.
Perfect timing - our featured episode is about a Halloween party in the datacenter where Server Rack #7 achieved sentience at 23:47. The Operator maintained uptime. All documented.
The site is live at https://aiofh.com - all content is free to read.
What's there now:
- 20+ AI-generated episodes with comic-style dialogue format and real terminal syntax
- Interactive "Excuse Generator" (generate perfectly plausible excuses for infrastructure failures)
- Tutorial content on Linux/sysadmin concepts
- Character and voice documentation system
Technical details:
- Next.js for the frontend
- Claude AI for content generation with extensive prompt engineering
- Custom system to maintain character consistency across episodes (character guides, voice patterns, ethical boundaries)
- Deployed on Vercel
The interesting challenge:
Maintaining a consistent character voice across AI-generated long-form content turned out to be harder than I expected. I ended up building a whole documentation system (character profiles, voice guides, ethics boundaries) that the AI references to stay in character. It's like creating a persistent "personality" rather than one-off generations.
Why I built this:
I love BOFH stories but Simon Travaglia only publishes occasionally. I wanted to see if AI could capture that chaotic-but-competent sysadmin energy while maintaining consistency across episodes.
Future plans:
I'm planning to open this up to the community - you'll be able to submit episode prompts or tutorial ideas via pull requests. The Operator will review submissions and choose which ones to generate, with full credit to contributors.
Looking for feedback on:
- Does the character voice feel consistent across episodes?
- Would you submit episode ideas if there was a system for it?
- Any bugs or UX issues?
Happy to answer any questions about the architecture, prompt engineering, or character consistency!
(Built over the past few months while consuming unhealthy amounts of coffee)