From the field.
Practical thinking on AI strategy, engineering, and operations. No fluff, just what works.
AI for the foreman, not the C-suite
Most AI adoption in manufacturing and logistics is designed for people who don't touch the product. The AI that actually works on a shop floor looks nothing like the AI that works in a boardroom.
Your agent doesn't need a bigger context window
Every AI tool has amnesia. Teams keep reaching for more context when the real fix is memory organization. When to use markdown vs a real database, why structured retrieval beats grepping prose, how Hermes and Open Claw do memory, and why a kanban board is the coordination layer that holds up.
What persistent memory actually changes
Most agents are stateless. Every session starts from zero. Hermes breaks that assumption at the architecture level, and the difference compounds.
The model is not your moat
A team held the model constant and only changed what surrounded it. Score went from 52.8% to 66.5%. Ranking jumped from outside the top 30 to top 5. The model didn't change.
Why your agent project isn't in production
88% of AI agent projects never ship. Here's what's actually breaking them.
Prompting software agents with test-driven development
The fastest way to make software agents reliable is to stop treating prompts like prose and start treating them like executable specifications. Use tests, fixtures, and feedback loops to shape what the agent does before you ask it to do more.
Using Claude Code as an Agent Swarm Scheduled with Cron Jobs
How to orchestrate multiple Claude Code agents running autonomously on schedules — turning a single AI coding assistant into a fleet of specialized workers that maintain, test, and improve your codebase around the clock.