Claude Code for Scientists

Companion code for Claude Code for Scientists: Turn Your Data, Papers and Lab Workflows into Reproducible Pipelines with AI Coding Agents, by Suryaprakash Tripathy.

18 builds, one folder each, in one repository. Clone it once and you have everything in the book. Each build stands alone: open its folder, read its README, run it, then run its gate.

Start here. Chapter 3 sets up the clone, the environment and the requirements, and ends with a script that tells you whether you are ready: science.nexagenlabs.com/setup

Which build belongs to which chapter

Chapter	Build	Printed address
0	00-reading-code	/ch00
2	02-what-an-agent-does	/ch02
4	04-first-session	/ch04
6	06-prompting-like-a-reviewer	/ch06
7	07-verification-gates	/ch07
8	08-git-and-environments	/ch08
9	09-literature-triage	/ch09
10	10-statistics-assumption-checks	/ch10
11	11-survival-and-dose-response	/ch11
12	12-publication-figures	/ch12
13	13-sequence-and-omics	/ch13
14	14-docking-prep	/ch14
15	15-image-and-plate-reader	/ch15
16	16-lab-data-management	/ch16
17	17-manuscript-assembly	/ch17
18	18-deposition	/ch18
19	19-grant-and-thesis-drafting	/ch19
21	21-error-hunt	/ch21

Chapters 1, 3, 5, 20, 22 ship no build, so there is no address for them. If you typed one, you are in the right place already.

The builds

• 00-reading-code — Reading Code You Did Not Write
  Prints one small table five ways, then lets an assertion fail. science.nexagenlabs.com/ch00
• 02-what-an-agent-does — What an AI Coding Agent Actually Does, and What It Cannot
  Two readings of the same two files, one gated by an assertion. science.nexagenlabs.com/ch02
• 04-first-session — Your First Session: A Messy Export to a Versioned Dataset
  Turns a messy plate-reader export into a tidy, versioned dataset. science.nexagenlabs.com/ch04
• 06-prompting-like-a-reviewer — Prompting Like a Reviewer: Inputs, Outputs, Units, Failure
  Normalises one plate to its controls two different ways. science.nexagenlabs.com/ch06
• 07-verification-gates — Verification Gates: Never Trust a Number You Did Not Reproduce
  Four families of verification gate, and a fault injector to test them. science.nexagenlabs.com/ch07
• 08-git-and-environments — Git, Environments and Notebooks Without Becoming an Engineer
  Two things that make a result hard to reproduce. science.nexagenlabs.com/ch08
• 09-literature-triage — Literature Triage: Screening That Survives Peer Review
  Screens a corpus at two strictnesses, then measures both screens. science.nexagenlabs.com/ch09
• 10-statistics-assumption-checks — Statistics with Assumption Checks and Honest Effect Sizes
  Four demonstrations of counting the wrong thing. science.nexagenlabs.com/ch10
• 11-survival-and-dose-response — Survival, Dose-Response and Growth: Fitting What You Measured
  Fits a four-parameter logistic, and a Kaplan-Meier by hand. science.nexagenlabs.com/ch11
• 12-publication-figures — Publication Figures That Regenerate on Demand
  Regenerates a figure from its data rather than editing it by hand. science.nexagenlabs.com/ch12
• 13-sequence-and-omics — Sequence and Omics: FASTQ to Differential Expression
  A count matrix with a known answer and two seeded design problems. science.nexagenlabs.com/ch13
• 14-docking-prep — Structure and Docking Prep: PDB Cleaning and Batch Runs
  Builds a receptor file with the defects real PDB entries carry. science.nexagenlabs.com/ch14
• 15-image-and-plate-reader — Image and Plate-Reader Data: Counts, MTT and qPCR
  Three kinds of plate and image data, each with a known answer. science.nexagenlabs.com/ch15
• 16-lab-data-management — Lab Data Management: Sample Tracking and Audit Trails
  Three mechanisms for knowing that data has not quietly changed. science.nexagenlabs.com/ch16
• 17-manuscript-assembly — Manuscript Assembly with Verified Citations
  Checks a manuscript against its own evidence. science.nexagenlabs.com/ch17
• 18-deposition — Supplementary Data, Availability Statements and Deposition
  What a deposit needs before a stranger can use it. science.nexagenlabs.com/ch18
• 19-grant-and-thesis-drafting — Grant and Thesis Drafting: Structure First, Facts Checked
  Two checks on a draft: are claims evidenced, is the language hedged. science.nexagenlabs.com/ch19
• 21-error-hunt — Catching AI Errors: An Evaluator's Checklist
  Summarises the defect log, and gates it against the chapter's claims. science.nexagenlabs.com/ch21

Gates

Every build ships the gate that checks its output. A gate exits 0 when it passes and 1 when it fails, which is the convention at a command line. A hook only blocks on exit 2, so a gate wired into one goes through tools/as_hook.py. Chapter 20 explains why, and the gates page points at the list that gives the exact path and both commands for every one.

Tooling

Version numbers, command names and configuration shapes change faster than books do, so no chapter depends on one. TOOLING.md records what each chapter was verified against, with the date. Where the book and that file disagree, the file is right.

References

Appendix D as machine-readable data, with a DOI on every entry that has one and the verification status of each: science.nexagenlabs.com/references

Errata

Corrections to the printed book are listed on the errata page. If you have found an error in the book or the code, please report it: the page says how.