Resources

'Vibe coding' with AI agents in scientific research

The article provides a critical perspective on 'vibe coding' with AI agents in the context of scientific research. Vibe coding involves creating software using large language models (e.g. ChatGPT, Claude) through prompting and minimal human-written code. It highlights the risks of relying on AI for coding tasks that require precision, such as data wrangling and model formulation in scientific research. The author shares experiences with agentic AI systems and discusses the potential for producing logically flawed results due to over-engineering and a lack of manual logic checks. Guidance on safely using AI agents for scientific coding is offered.

Go to Resource

A Claude Skill for _brand.yml, and sharing with Quarto 1.9

Stephen Turner created a Claude Skill for generating '_brand.yml' files, which define brand styling for Quarto documents. This skill streamlines the creation of consistent visual themes across various document formats like websites and PDFs. A _brand.yml file is a single YAML file that sets colors, fonts, and typography. Turner tested the skill by creating a brand.yml for the UVA School of Data Science, demonstrating ease of use and consistency in branded outputs. The skill leverages Quarto 1.9's feature to share brand.yml via GitHub. Users can create brand.yml files based on existing websites, institutional color schemes, or custom descriptions.

Go to Resource

A Few Claude Skills for R Users

This post highlights various Skills for R users developed by the community to work with coding AI tools, specifically tailored for Claude, an AI assistant by Anthropic. These Skills offer guidance on using the latest features in R coding, promoting best practices, and improving productivity. The author mentions resources for finding these Skills and provides definitions for terms like Claude Models and Claude Code. The post also differentiates between CLAUDE.md files and Claude Skills based on their scope and context usage, suggesting the latter for task-specific commands. Community contributions, such as "Claude R Tidyverse Expert" and "Claude Code R Skills," are exemplified.

Go to Resource

Add last rendered or modified time to Quarto

Garrick Aden-Buie's blog post introduces 'now,' a Quarto extension that allows the automatic update of time information in Quarto documents. This extension saves time by eliminating the need for manual updates of dates in documentation footers. By adding the extension using 'quarto add gadenbuie/quarto-now,' Quarto users can employ shortcodes like '{{< now >}}' and '{{< modified >}}' to display the current or last modified time. The extension supports customization of time output formats and may significantly streamline Quarto project maintenance by ensuring date accuracy without manual intervention.

Go to Resource

Adding a logo to images with {magick} and {purrr}

Jadey Ryan shared her experience with automating the process of adding logos to images using the R packages {magick} and {purrr}. She also highlighted new merchandise such as tote bags and embroidered hats on her Etsy shop. Additionally, she announced a free webinar series on soil health where she will demo the {soils} R package. Alongside, she informed about the Parameterized Quarto workshops she is giving, focusing on efficient report generation using R and Quarto, with the next one scheduled with R-Ladies Abuja.

Go to Resource

Adding alt text to figures in quarto with Claude Code

The blog post describes a new workflow for adding alternative (alt) text to figures in Quarto documents, leveraging Claude Code, to make the process less tedious and more manageable. It emphasizes the importance of thoughtful alt text creation, details how Quarto's features can be used to extract figure details, and introduces a custom slash command for Claude Code to assist in writing alt text. This command uses Amy Cesal's formula, combining chart type, data description, and key insight into the alt text, utilizing source code and prose context for accurate and informative descriptions.

Go to Resource

AI capabilities built into Positron

Recorded at posit::conf(2025), this demo showcases practical ways to use the Positron Assistant for tasks like debugging code, generating visualizations, and even creating entire Quarto reports. It also introduces Databot, an experimental agent for exploratory data analysis, demonstrating how these tools can accelerate your workflow while keeping you in the driver’s seat.information, advertising opportunities, and terms of service. It also includes a nod to privacy and safety principles and the operational workings of YouTube. Additionally, it promotes YouTube's experimentation with new features and the service NFL Sunday Ticket. The 2025 copyright date suggests forward-looking engagement with these matters. The content outlines multiple facets of YouTube's online presence and user interface.

Go to Resource

AI-Assisted Clinical Dashboard Development

This article showcases the development of a Clinical Dashboard Prototype using AI-assisted tools within a constrained 48-hour period. The development team faces the challenge of creating a functional PK/PD simulation platform for a stakeholder meeting without enough time for traditional development cycles. By leveraging AI tools like Claude Code, V0.dev, and others, they demonstrate the acceleration of dashboard development. The focus is on the interplay of AI tools to achieve functionality and user experience, reflecting on the strengths and limitations of each tool in the process.

Go to Resource

AI-Powered Data Science in Positron

Positron is a new, free AI-powered code editor by Posit for R and Python, featuring agentic AI capabilities. It was demonstrated by Ryan Johnson, a Data Science Advisor at Posit PBC, who emphasized the tool's ability to simplify data science workflows while keeping the experts in control. Positron is designed to automate tedious tasks, and this demonstration on October 29, 2025 will showcase how AI can enhance productivity without taking over the expert’s role.

Go to Resource

Automate subset plots with ggplot2 and purrr

Cedric Scherer's blog post, 'Efficiency and Consistency: Automate Subset Graphics with ggplot2 and purrr,' guides readers through the process of automating the creation of subset graphics in R using the ggplot2 and purrr packages. It explains how to eliminate redundant work when generating explorative or explanatory charts for various data subsets by iterating over a vector of groups with a custom function. The post provides a practical tutorial on improving efficiency and consistency when visualizing relationships for different numeric variables, with a focus on polished charts and including examples and shortcuts for data exploration.

Go to Resource

Creating a data pipeline with Github Actions & the {googledrive} package for the Canadian Premier League soccer data initiative!

Go to Resource

Creating post summary with AI from Hugging Face

Dr. Mowinckel discusses the use of Hugging Face's AI to automate the creation of SEO-friendly summaries for blog posts, which is both time-efficient and enhances discoverability. The tutorial encompasses acquiring the Hugging Face API key, structuring requests, and handling responses with the R package httr2. It also highlights the importance of concise summaries for both SEO and repository metadata, and details the workflow from content preparation to writing summaries to a file. Hugging Face’s community efforts and accessible APIs are commended for their ease of use and functionalities.

Go to Resource