---
title: "Comment by Kartik N V J K on \"Building a self-planning data pipeline agent, and measuring whether it's actually right\" | Hashnode"
description: "Rating everything easy sends it all to the cheap model, and your repair loop already holds the evidence to check that rating. If subtasks marked easy keep comin"
url: "https://hashnode.com/posts/building-a-self-planning-data-pipeline-agent-and-measuring-whether-it-s-actually-right/6aeaaa9dc6c5769a73e6b4f2/comment/6aeb3f9cd32119e270489616"
---

[Discussion](https://hashnode.com/posts/building-a-self-planning-data-pipeline-agent-and-measuring-whether-it-s-actually-right/6aeaaa9dc6c5769a73e6b4f2)

Comment [Building a self-planning data pipeline agent, and measuring whether it's actually right](https://hashnode.com/posts/building-a-self-planning-data-pipeline-agent-and-measuring-whether-it-s-actually-right/6aeaaa9dc6c5769a73e6b4f2)

[Nov 4](https://hashnode.com/posts/building-a-self-planning-data-pipeline-agent-and-measuring-whether-it-s-actually-right/6aeaaa9dc6c5769a73e6b4f2/comment/6aeb3f9cd32119e270489616)

Rating everything easy sends it all to the cheap model, and your repair loop already holds the evidence to check that rating. If subtasks marked easy keep coming back from the sandbox for a second or third attempt, I'd count the rating as wrong and start the next file with the same profile on the stronger model.

[Danda Pranav Krishna](https://hashnode.com/@pranavkrishnadanda)

[Nov 4](https://hashnode.com/posts/building-a-self-planning-data-pipeline-agent-and-measuring-whether-it-s-actually-right/6aeaaa9dc6c5769a73e6b4f2/comment/6aeb3f9cd32119e270489616#reply-6aeb408d29c05f31a4b70c1f)

Good catch, and half of this is already there. A model's "easy" rating can't send work to the cheap model by itself: a deterministic floor raises it to hard when the profiler sees unit issues, mixed date layouts, or grouping plus rescaling. And easy only gets the small model on attempt 1; any retry escalates to the strong model. What it doesn't do is learn across files. An easy subtask that needed escalation on file N starts on the small model again on file N+1. Keying that on a column profile (dtype, the shapes seen, the source-name hints) and starting on the strong model after repeated escalations is a cheap, measurable fix. The eval set can tell us whether it pays off. Adding it to the roadmap, thanks.