Claude Code at Scale

Days 3–5 · scientific programming interest and ability

The second half of Week 1 moves into Claude Code — using coding agents for real scientific programming, and scaling from a single session to many.

Getting started

• The basics of using Claude Code for scientific programming.
• Driving it effectively: clear tasks, good context, and tight feedback loops.
• Validating the results — using agents for development without trusting output blindly.

Scaling up

• Running parallel instances to work on multiple workstreams at once.
• Managing and orchestrating those workstreams so they stay coherent.
• Connecting to tools and data with remote MCP, so agents can work against real resources.

Converting codebases across languages

A key highlight of the week: using Claude Code's built-in skill to convert codebases across languages — for example, migrating legacy scientific code to a modern language and cloud architecture, then validating the result.

Outcomes

By the end of Week 1 you should be able to use Claude Code for scientific programming, orchestrate multiple parallel instances effectively, work against data and tools through remote MCP, and migrate legacy code across languages — all while validating what the agent produces.