15. Resources¶
← Practical tips · Contents · Next: Contributing →
Start with the two references below. The rest is there for questions that arise as you work; it is not a prerequisite reading list.
Our main references¶
Ten Simple Rules for AI-Assisted Coding in Science, by Bridgeford and colleagues, is the starting point for this book. Its ten rules cover preparation, interaction, testing, review and focused improvement. Read it early, then return to it after a few sessions with an agent. The paper links to companion worked examples. It was published in PLOS Computational Biology in July 2026.
Russ Poldrack, Coding with AI, chapter 5 of Better Code, Better Science, develops the practical workflow: context files, test-first development, recognising failed approaches and using version control. Read the sections relevant to your current task. The chapter is being revised, so specific tool details may change.
For this lab book, the progression is deliberate: first make a small analysis work, then learn to check and maintain increasingly complex analyses. You do not need all the infrastructure in these references on day one.
Learning the programming basics¶
The rest of Better Code, Better Science covers software testing, project structure, workflows and scientific validation.
The Good Research Code Handbook is Patrick Mineault's guide to organising research code. His Claude Code for Scientists also informed the original lab book's workflow and treatment of generated test data. These are supplementary reading alongside our main references.
Tool documentation¶
- Claude Code installation: current setup instructions.
- Claude Code best practices: context, planning and review workflows.
- Claude Code permissions: what access settings control.
- Claude Code skills: writing and loading reusable procedures.
- uv projects: Python environments and dependency records.
- Snakemake tutorial: optional reading once you need to coordinate several analysis steps.
For evidence about context length, see Chroma's Context Rot report. Its experiments are useful background, not a universal token limit for your sessions.
Lab resources and optional tools¶
Lab members can access ai-skills through the lab's GitHub organisation. Follow its README for current installation instructions; page 10 introduces the idea.
GenomeDK documentation is the reference for cluster work. Pages 11 and 12 cover lab examples and known issues.
RainCloudPlots is the lab's plotting project. Crawfurd's academic research skills contains optional review procedures. Inspect any procedure before adding it to your workflow.