Test AI annotations on student problem sets
A chemistry professor ran 11 students' problem sets through Claude to see if its annotations could help manual grading; the project was accidentally left public, exposing student information.
Done withClaude
- What they did
- Chemistry professor Paul Robustelli made a project called "PS3 Grading" in Dartmouth's Claude enterprise portal. He uploaded 11 students' problem sets from his CHEM 76 class and ran them through Claude. He wanted to see whether its annotations could later serve as a reference for his manual grading. He says Claude had no role in any grade.
- How it went
- The project was visible to everyone in the Dartmouth Claude group, with students' names and work exposed. It came down May 2 after The Dartmouth asked for comment. The source reports no results on annotation quality.
- Worth knowing
- New projects default to private. Sharing happens only if you pick the organization-wide visibility option, and Robustelli says he thought the project was private, so check the setting before uploading student work.
Try it yourself with Claude Code
Take the student problem sets in [folder] after I remove names and IDs from them. For each one, annotate where the work goes wrong against [solution key or rubric] and write a short comment I could use while grading by hand. Save the annotations in a local folder [output path] and stop when every file is done. Keep everything private, don't upload it anywhere, and pause and ask me before any step that shares or publishes files.
Source: thedartmouth.com · May 1, 2026
Five of these in your inbox every morning
The best things people got an AI agent to do, each with the prompt to try it.
More like this
- GitHub Copilot
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Run browser agents for small admin tasks
An educator tried agentic browsers that made a meeting poll, looked up and summarized colleagues' teaching schedules, and reformatted a slide deck, though clunky and slow.
Reconcile bank statements with invoices
Cowork cleaned vendor names, categorized expenses, matched invoices to transactions, flagged missing ones, built an Excel file and renamed invoice files.
Turn Notion PRDs into scoped Linear tickets
Devin's docs describe reading a Notion PRD to create Linear tickets, producing implementation plans for a backlog, and posting customer health summaries to Slack.
- Custom agent
Contact job candidates for initial HR screening
A job seeker reports an AI agent calling and emailing them for a first HR screening.
Learn personal scripts from a tools folder
Claude Cowork read a folder of personal scripts for email, CRM, research and drafting, added them to memory, and carried the capabilities to a new interface.