Where dltHub's AI harness ends and Snowflake's CoCo begins
dltHub's AI harness writes the pipeline, Snowflake's CoCo takes over after data landed. Together they ✨
The regexes had a good run
Hand-written patterns decided which Salesforce errors were noise. Now a classifier does it, and it says how sure it is.
Making Sonnet 5 the CoCo default
Cortex Code's model default? Turns out it's just a CREATE AGENT statement away.
Snowflake gave everyone 102 models and didn't tell me
200 users, zero curation, and a bootstrap grant I didn’t ask for.
What if a dltHub pipeline fails without errors?
A watchdog to keep all dlt pipelines in line
70 dlt pipelines, zero comments (and no alert when the schema changed)
Nobody documents dozens of tables with hundreds of columns by hand. So an agent does it, INFORMATION_SCHEMA keeps it honest, and Teams pings when a source API adds a column
Snowflake data movement policies: the copy path gets a guardrail
Masking decides what you can see. Data movement policies decide what you can take.
dltHub notifies Slack and email. But my org runs on Teams 😩
dltHub ships Slack and email alerts out of the box. My org runs on Teams, so I built the third one myself and wired it into all ~70 pipelines with one decorator shadow.
7 shared modules for 27 dlt pipelines
Re-using shared patterns in a dlt workspace
LinkedIn Ads ➡️ Snowflake, with 3-legged OAuth in the middle
LinkedIn's Marketing API needs a browser to authorize. Scheduled pipelines don't have browsers. I store the refresh token in Snowflake.
I migrated a Fivetran pipeline to dltHub for $0.65
One hour, five manual inputs, and $0.65 of tokens... enough to replace a daily Fivetran sync with a tested dltHub job
I needed 2,110 municipal contacts. So I built a crawler that refuses to guess.
A public contact list sounds easy... until 2,110 websites, four languages, changing staff, and source evidence get involved. 😅