Writing
Short notes from X and longer essays on software, data systems, and engineering practice.
Short
We built software free as in open and got free as in beer thrown in. Grow the bazaar and get cathedrals too.
I've been building data pipelines and workflows for years in different industries. I finally got around to making the OSS package I wanted to improve this type of work. It's fun, fast, and feature packed.
There's a divide how firms are thinking about making existing software "agent native":
- Approach A: Add an MCP server, maybe add a CLI, add auth tokens. 'Use our Claude integration'
- Approach B: The application == the agent. Requires totally new approach.
Buy vs Vibe is the eng leadership discussion of 2026.
More notes or follow @chetmancini on X.
Long
- An Easier Way to Consider the CAP Theorem
CAP is usually taught as "pick two of three." A more useful framing is a single question — when the network partitions, does your system keep consistency or availability?
- Three Learnings from Real World Large-Scale Machine Learning
Notes from tuning production segmentation and quality-score models at Intent Media after moving logistic regression to Spark MLlib on EMR.
- The Art of Abstraction for Engineers
A visit to the Cooper Hewitt and the Free Universal Construction Kit, and what connector pieces between toy systems teach about designing abstractions between software systems.
- Data Visualization: How to Make a Really Bad Graph
A tongue-in-cheek guide to the common traps in chart design — multiple y-axes, unnatural intercepts, mismatched scales, ratios, and the obligatory linear trendline.