Claude Code 18: When the Reclassification Is Massive But the Trends Don’t Change, Something Interesting Is Happening
Digging into why a $11 LLM replication of a PNAS paper works at all
This is Part 18 of an ongoing series on using Claude Code for research. But it’s also my third post on using Claude Code to replicate a paper that used natural language processing methods (specifically, a Roberta model with human annotators). The first part (below) set it up and included a video recording of it.
Claude Code Part 14: I Asked Claude to Replicate a PNAS Paper Using OpenAI's Batch API. Here's What Happened (Part 1)
I’ve been experimenting with Claude Code for months now, using it for everything from writing lecture slides to debugging R scripts to managing my chaotic academic life. But most of those tasks, if I’m being honest, are things I could do myself. They’re just faster with Claude.
Claude Code 15: The Results Are In: Can LLMs Replicate a PNAS Paper? (Part 2)
TL;DR I used gpt-4o-mini to replicate the text classification from Card et al.’s PNAS paper on 140 years of immigration rhetoric. Here’s what happened:
In Part 2 (linked above), I showed you th…
