"𝗜 𝗵𝗮𝘁𝗲 𝘀𝗰𝗵𝗲𝗱𝘂𝗹𝗲𝘀 𝗬𝗔𝗠𝗟."

That's what a user told us in 2023 about our declarative YAML schedules. Honestly? It was fair. The UX was clunky and glitchy.

But the user base kept growing anyway. Every time we tried to kill YAML schedules, the data wouldn't let us - market pull beat the loud early feedback. Then AI agents showed up and settled the argument entirely. Newer competitors in the market shipped something similar. #dbt Cloud™ talked about it and but never did.

𝗪𝗶𝘁𝗵 𝗬𝗔𝗠𝗟 𝘄𝗲 𝗴𝗼𝘁 𝗮 𝗹𝗼𝘁 𝗿𝗶𝗴𝗵𝘁: git-tracked, auditable, transparent, and incredibly agent-friendly. The UX did not: one giant file for every schedule, painful edits, no easy way to verify anything.

So we rebuilt it. 𝗦𝗵𝗶𝗽𝗽𝗶𝗻𝗴 𝘁𝗵𝗶𝘀 𝘄𝗲𝗲𝗸 𝗶𝗻 𝘃𝟲:

𝗡𝗼 𝗺𝗼𝗿𝗲 𝗼𝗻𝗲-𝗳𝗶𝗹𝗲 𝗹𝗶𝗺𝗶𝘁. Create schedules as individual files anywhere in a .𝗯𝗼𝗹𝘁 folder, nested as deep as you want. We assemble them for you. If you're running hundreds of schedules, this one's for you.

𝗔 𝗹𝗶𝘁𝗲 𝗲𝗱𝗶𝘁𝗼𝗿 𝗨𝗜. Edit and commit schedule changes straight from the UI — same git tracking, no IDE required.

𝗖𝘂𝘀𝘁𝗼𝗺 𝗯𝗿𝗮𝗻𝗰𝗵𝗲𝘀. Pick which git branch your YAML schedules read from.

𝗪𝗵𝘆?

Data engineers are writing schedules with AI agents on #paradime, and they want pipelines that self-heal. Agents read differently than we do - granular files mean fewer tokens and surgical fixes instead of guesswork. Meanwhile, plenty of people writing code today don't have a full dev setup, so a lite UI gives them room to make quick changes. And when agents write pipelines, those pipelines need to be tested away from production - hence custom branches.

This is what deep thinking about AI-native data engineering looks like. Not a demo but mission-critical infrastructure for enterprises.

If you want to learn more on how you can build declarative pipelines for ingestion, dbt™, and AI agents, sign up for a free trial at https://app.paradime.io or send me a DM on LinkedIn.