The Library
Data Engineering & Analytics
Everything Nullhaus has published in this pillar. Free, open, no login. Back to the library
-
Instrument the domain, not the interface
Analytics built on button clicks and page views can tell you what people touched. It cannot tell you whether the product worked for them, and it…
-
Indexing a chain that keeps changing its mind
A blockchain is not an append-only log, it is a consensus process that occasionally retracts what it told you. Almost every first-generation indexer is built on…
-
Your pipeline has a dashboard, not tests
Software fails loudly. Data fails politely: the job succeeds, the dashboard renders, the numbers are wrong, and the first person to notice is someone making a…
-
Idempotent pipelines: why re-running should be boring
If a job runs twice on the same input, do you get the same result, or a mess? That question separates a pipeline you operate calmly…