Sheet W-00 / Writing

Writing

Notes from the drawing board: what breaks in production, why, the workaround that did not hold, and the fix.

Sec. 01 / Register of sheets
2 sheets
No.TitlePracticeDateWhere
W-06What is an agent harness, and why does your agent need one?The model is a function. Everything that makes it an agent lives in the harness: the loop, tools, context, memory, guardrails and verification. Why demos break without it, and what to build.AI agentsThis site
W-05Author Fabric data pipelines in PythonThe portal and a JSON file are the only two editors. Write the pipeline in Python, check the graph, and export the folder Fabric syncs from Git.Data platformsThis site
Sec. 02 / By practice

Four practices, one building.

Each piece belongs to one of the four practices on the cover sheet: the foundation, the columns, the door and the figures. Follow a practice to its sheet, or read the pieces filed under it.

Drawn by
BastiaanFounder, datalyft · Weesp, NL

Bastiaan is the founder of datalyft and a data and AI (platform) engineer. He has built on AWS, Snowflake, and Microsoft Fabric, writes about the Modern Data Workflow on Medium, and now designs, builds and runs AI agents for business workflows.

LinkedInMediumGitHub
Sec. 03 / Start

Reading is the cheap part. Where would you start?

A 30-minute conversation about your situation: where you are, what is possible, and what a sensible first step looks like. No obligation.