START HERE
Know the job of each tool.
| Tool | Use in camp | Required? |
|---|---|---|
| R | Run code and carry out statistical analysis. | Yes |
| RStudio Desktop | Edit scripts and inspect files, objects, and plots. | Recommended |
| Quarto | Render the optional field notebook and slides. | Yes |
| Codex | Use project files when helping with code. | Yes |
| Claude Code | Use as an alternative coding agent. | Optional |
OPTIONAL PRIMER · 49 MINUTES
Watch a gentle introduction to R.
This recorded walkthrough introduces the R environment and the basic workflow we will use in Math Camp. It is designed for students opening R for the first time and also works as a concise refresher before the coding lessons.
INSTALL ORDER
Install the tools before you open the project.
- 01Install R from CRAN. On Windows, accept the default installer options. On macOS, choose the package that matches your processor. RStudio should find R automatically.
- 02Install RStudio Desktop.
- 03Install Quarto. Restart RStudio after the installation so the terminal can find Quarto.
-
04Download the complete Math Camp repository.
Use GitHub’s Code → Download ZIP and unzip the
download, or run
git clone https://github.com/harvard-api209/math-camp.git. -
05Open
math-camp-2026.Rprojin RStudio. Opening the project file keeps the paths used in every exercise available. A script opened by itself from Downloads loses that project context. -
06Sign in to Codex with the Harvard-supported account.
Open the same project folder in Codex and confirm that you can
see
START-HERE.md.
Keep scripts, data references, notes, and generated output inside the project. This keeps the workflow independent of Desktop files and temporary console history.
READINESS GATE
Start the analysis only after every check is green.
Open the Terminal tab in RStudio. Run the same command from the project root on macOS or Windows:
Rscript code/check_setup.R --codex-confirmedThe checker verifies the project, R version, required packages, Quarto, frozen data, and your manual confirmation that Codex opens with the project. A red result gives you a repair instruction. A yellow result means that only the Codex confirmation remains.
If a package is missing, install exactly the list reported by the checker in the R console. Then run the gate again. A typical repair is:
install.packages(c(
"dplyr", "ggplot2", "jsonlite",
"readr", "tidyr"
))-
□The final line says
RESULT: GREEN. -
□RStudio shows
math-camp-2026as the open project. -
□Codex can read
START-HERE.mdin the same folder. - □You can explain what each check established.
CODING AGENTS
Give the agent useful project context and protect private data.
Codex is the common baseline. You may use Claude Code as an alternative. Both agents can read project files, propose code, explain errors, and run checks. The student decides whether the evidence supports a policy claim.
Read this project without changing files. Identify the data directory, lesson pages, and any setup instructions. Tell me what you know from the files, what you infer, and what you still need me to decide.
Use only public, synthetic, or course-approved data. Never paste personal records, credentials, API secrets, confidential research material, or identifiable student information into an AI tool.
IF SOMETHING FAILS
Bring the exact evidence from the failure.
Save the exact error, the command that produced it, your operating
system, and the complete output of
Rscript code/check_setup.R. Save the full text and
use a screenshot as a supplement, because a screenshot may hide
useful context.
On Windows, confirm that you extracted the ZIP before opening the project. On macOS, if you see a security warning, reopen the official installer through System Settings. Continue only with an official download. Students with a green result may help diagnose problems in class. Each student must also run the gate on their own laptop.