Really inspiring read, Bronson! I love how you point out that the specification-first mindset unlocked everything. The idea of watching the agents do the build while you step back into reviewer mode feels like such a powerful shift for analysts who’ve never felt comfortable writing traditional code.
Curious how you’re thinking about versioning the agents or handling edge cases where data quality is really messy. Would love to hear more about your approach with Production data!
Thank you for the comments! Regarding data quality, that's exactly what the `data-cleaner` agent handles. I have a very messy sample data file that I used to help test it. There are a few common things it will automatically do (trim white space) but for other larger impact items (duplicate records, missing data, outliers, etc) is asks the user how to handle them, I wanted the user to have final judgement on those types of actions.
Really inspiring read, Bronson! I love how you point out that the specification-first mindset unlocked everything. The idea of watching the agents do the build while you step back into reviewer mode feels like such a powerful shift for analysts who’ve never felt comfortable writing traditional code.
Curious how you’re thinking about versioning the agents or handling edge cases where data quality is really messy. Would love to hear more about your approach with Production data!
Thank you for the comments! Regarding data quality, that's exactly what the `data-cleaner` agent handles. I have a very messy sample data file that I used to help test it. There are a few common things it will automatically do (trim white space) but for other larger impact items (duplicate records, missing data, outliers, etc) is asks the user how to handle them, I wanted the user to have final judgement on those types of actions.