Conversation
New title and intro, then kept all existing content, but moved most of the case-study specific info to an appendix to instead tell the story of what Claude is doing rather than focusing on the merge example.
petermuehlbacher
left a comment
petermuehlbacher
left a comment
There was a problem hiding this comment.
I looked through the first half or so and the only issues I found were those introduced by me in the first place, so I'll just trust you on this one :)
docs/case_studies/match-clinical-trials-to-papers/notebook.ipynb
Outdated
Show resolved
Hide resolved
| "\n", | ||
| "**1. Specialized orchestration beats general-purpose agent planning for data operations at scale.**\n", | ||
| "EveryRow's `merge()` is purpose-built to decompose a large matching problem into hundreds of parallel agent tasks, with intelligent candidate selection and verification. Claude Code is remarkably clever — it independently invented a TF-IDF + parallel-subagent pipeline — but a general-purpose coding agent can't match a system designed specifically for this class of problem.\n", |
There was a problem hiding this comment.
Actually, this wasn't completely independently—my data scientist tips (a Skill I wrote for my local CC) helped a decent amount here; it should basically explain the delta between this CC performance (not bad!) to the CC performance in the video I sent you (very bad).
Doesn't need to be changed, but I was reminded of this when recording the video with a clean CC instance.
|
Ah, actually you shouldn't delete the setup cell. This makes the tests fail. |
| "source": [ | ||
| "# Setup: install everyrow if needed and configure API key\n", |
There was a problem hiding this comment.
@petermuehlbacher is this the setup you are referring to?
These results are already plotted in the bar chart below
|
@mckenna-futuresearch I pushed some edits which should make the notebook runnable. Let me know if you'd like any more changes |
docs/case_studies/match-clinical-trials-to-papers/notebook.ipynb
Outdated
Show resolved
Hide resolved
3 participants
New title and intro, then kept all existing content, but moved most of the case-study specific info to an appendix to instead tell the story of what Claude is doing rather than focusing on the merge example.