"AI Pipeline Chronicles: When Your Automation Needs a Human Guardian"

“I Built an AI Pipeline to Write About Building My Products. Then I Had to Debug the Debugger.”

Roberto Luna, a solo full-stack developer running four SaaS products from Playa del Carmen, built an AI pipeline to automate daily technical writing from his GitHub commits. The system hit three unglamorous bugs in one week—a GitHub Actions cron silently skipping triggers, a Craft Docs REST API that contradicts its own documentation, and hardcoded auto_publish defaults putting unreviewed LLM output live under his name.

Why This Matters

Key Insights

• Reliability trumps capability: A 17B parameter model (llama-4-scout-17b-16e-instruct on Groq) with steady free-tier access beats smarter frontier models whose credentials expire mid-deploy for unattended automation (Luna, June 2026).

Working Examples

YAML configuration fix that gates AI-generated content behind human review before publication on Bluesky and Dev.to.

bluesky:
auto_publish: false # Requires review before posting
devto:
auto_publish: false # Requires review before posting

Practical Applications

References:

• From internal analysis