About this Event
Claude Code in Practice
Abstract: Claude Code operates from the command line, with access to the files, software, and commands of the machine it runs on. Given a task described in plain language, it examines what it needs, does the work, reads the result, and revises where that result falls short. Natural language combined with direct access to a working system is what distinguishes it from an assistant in a browser, and what makes it consequential: the work happens on real systems, under the user's own account, at whatever level of independence the user allows. Applied with method, it removes a great deal of routine labor from computational research. That method is the subject of this workshop.
The workshop covers what Claude Code is capable of and where it is best applied; how it is configured, including the choice of model and the context it works from; how it is instructed, followed, and interrupted while work is under way; how it is extended to observe the conventions of a project or group; and what it costs to use, which is rarely where people expect. Participants then work on Midway3, using Claude Code on the cluster itself: asking it in plain language about storage, installed software, and what has run recently, carrying out a short analysis whose correct answers are known in advance, and submitting the finished analysis to the cluster as a task.
Learning Objectives:
- Work fluently inside Claude Code itself, including slash commands, skills, MCP connections, and the configuration that governs how it behaves.
- Give Claude Code the context it needs about your project, so its work reflects your conventions and your environment rather than generic defaults.
- Match the autonomy you grant Claude Code to the task at hand and to the fact that the hardware is shared.
- Review and verify work that Claude Code produced, using the commands it ran and a known result as the basis for judgment.
- Direct a multi-step analysis from an opening question through to a result, refining the instruction as the work develops.
Presenter: Youzhi Yu
Duration: 2 hours
Level: Introductory. Concerned with using Claude Code rather than with building software; no programming experience assumed.
Prerequisites: An RCC account, and access to Claude Code through a University of Chicago or personal subscription. No Python, high-performance computing, or machine learning experience required. Participants should bring a laptop.
Event venue & nearby stays
John Crerar Library - Kathleen A. Zar Room, 5730 South Ellis Avenue, Chicago, United States
Concerts, fests, parties, meetups — all the happenings, one place.