Transcript
Click on the transcript to go to that point in the video. Please note that transcripts are auto generated and may contain minor inaccuracies.
Loading transcript...
I spent this session using AI to turn a collection of R data visualizations into a formatted Quarto and Typst report, starting from scratch on the design and working through layout, county-level pages, and state-specific styling including colors pulled from state flags.
You'll learn:
- How to use AI to sketch a wireframe layout before writing any report code, letting it propose structure based on your existing charts and tables
- How to separate Typst template code from Quarto document code to keep the project organized and easier to maintain
- How to parameterize a report so it can generate pages for every county in a state, then loop over all states to produce one PDF per state
- How to give AI rambly, conversational feedback on a design and have it interpret that into concrete code changes
- A workflow of running multiple Claude threads or sub-agents in parallel to handle separate tasks like researching state colors while the main thread writes code
Tools mentioned: Claude, Claude Opus, Solo, Raycast, Quarto, Typst, R, ggplot2, gt, Positron, GitHub Desktop, Remotion
Key moments:
- 00:00 Stream starts, David introduces the project
- 01:03 Overview of Oregon by the Numbers reports as inspiration
- 03:42 Starting the AI conversation about report layout
- 06:05 Asking Claude to think about Quarto and Typst structure before coding
- 08:05 Claude asks where chart titles should live and how tables should render
- 14:17 Claude generates a Typst wireframe PDF
- 19:09 Asking Claude to implement the wireframe using real R code in Quarto
- 30:56 First full render of the report appears on screen
- 32:10 Deciding to separate Typst code into template files
- 35:07 Planning multi-county and multi-state report generation
- 42:37 Claude spins up a sub-agent to research state colors and flags
- 51:04 Switching from git worktree to direct branch to fix renv issues
- 01:01:06 Claude creates render-reports.r and tests it on Delaware
- 01:09:19 Successfully rendering Alabama by the Numbers with one page per county
- 01:11:29 Final commit and push, stream wraps up