Where I Am on the Agentic Engineering Ladder
Bassim Eledath wrote a framework for levels of agentic engineering. I mapped my own work against it.
13 posts
Bassim Eledath wrote a framework for levels of agentic engineering. I mapped my own work against it.
Qumio's memory layer went through three architectures in two days. The final version is just the agent talking to Obsidian directly.
Qumio strips PII from email data using a local model before formatting it. One agent handles fetch, redaction, and categorization on-device.
Small language models stop after producing output that looks complete. If you need a file write, do it before the final response.
At 3 AM during the Qumio build, Claude told me to stop for the night. I didn't expect project management from an AI.
I typed 'lets add more blogs' into a command router and watched four layers of AI tooling figure out what I meant.
The generate-blog-post skill is a markdown file that tells Claude where to find stories, how to avoid sounding like AI, and what to do when a draft gets rejected.
Qumio v1.0 went from empty repo to seven working skills in 98 commits over 15 hours. Here's what that actually looked like.
Qumio's entire capability system is text files. No code, no deploys, no builds. Just prompts that hot-reload in under a second.
If the main agent can do something itself, it will skip the delegation step. Every time.
Every first draft from the AI had em dashes, 'key insights', and rule-of-three lists. I had to build rules to stop it.
I gave Claude discretion on five design decisions at once. Here's how that went.
How a morning routine problem led to a Telegram AI assistant running in Docker.