Tristan, it sounds like we’re dealing with two competing realities:
- AI handling ingestion end-to-end,
- versus paying a SaaS vendor just in case.
We love automated maintenance which is why we already did it without AI. And while claude 3.5 was pretty good at building dlt pipelines, the latest Claude can even use our incremental logic, parallelism, etc to build efficient pipelines.
We even generate dbt scaffolds programmatically and evolve schemas, so really any maintenance would happen in dbt directly. https://hub.getdbt.com/dlt-hub/
So given building and maintenance are already near-free, what do you think will be value to let a vendor run the code AI writes for you?
I saw MCP mentioned on the Changelog podcast recently…interesting to think about it as the framework to drive aggregation/integration
Also, kind of love the term “medium code”. Low-code tools just aren’t cutting it, and it’s so frustrating not to be able to have AI generate something I can copy+paste into them
Tristan, it sounds like we’re dealing with two competing realities:
- AI handling ingestion end-to-end,
- versus paying a SaaS vendor just in case.
We love automated maintenance which is why we already did it without AI. And while claude 3.5 was pretty good at building dlt pipelines, the latest Claude can even use our incremental logic, parallelism, etc to build efficient pipelines.
We even generate dbt scaffolds programmatically and evolve schemas, so really any maintenance would happen in dbt directly. https://hub.getdbt.com/dlt-hub/
So given building and maintenance are already near-free, what do you think will be value to let a vendor run the code AI writes for you?
I saw MCP mentioned on the Changelog podcast recently…interesting to think about it as the framework to drive aggregation/integration
Also, kind of love the term “medium code”. Low-code tools just aren’t cutting it, and it’s so frustrating not to be able to have AI generate something I can copy+paste into them