The R Core team receives a prestigious award in the world of Statistics, a month-long recognition of the rOpenSci package maintainers, and a collection of helpful tips for obtaining robust R code from AI agents. Episode Links This week's curator: Eric Nantz: @rpodcast@podcastindex.social (Mastodon) & @rpodcast.bsky.social (BlueSky) & @theRcast (X/Twitter) 2026 Rousseeuw Prize for Statistics Awarded to R Core Team for Transforming Statistics Computing Worldwide Celebrating Our Maintainers during Maintainers Month 10 tips for getting better R code from your AI coding agent Entire issue available at rweekly.org/2026-W26 Supplement Resources The Rousseeuw Prize for Statistics 2026 https://www.rousseeuwprize.org/2026 {btw} A complete toolkit for connecting R and LLMs https://posit-dev.github.io/btw/ Supporting the show Use the contact page at https://serve.podhome.fm/custompage/r-weekly-highlights/contact to send us your feedback R-Weekly Highlights on the Podcastindex.org - You can send a boost into the show directly in the Podcast Index. First, top-up with Alby, and then head over to the R-Weekly Highlights podcast entry on the index. A new way to think about value: https://value4value.info Get in touch with us on social media Eric Nantz: @rpodcast@podcastindex.social (Mastodon), @rpodcast.bsky.social (BlueSky) and @theRcast (X/Twitter) Mike Thomas: @mike_thomas@fosstodon.org (Mastodon), @mike-thomas.bsky.social (BlueSky), and @mike_ketchbrook (X/Twitter) Music credits powered by OCRemix Super Buck II - Super Mario Brothers 2 - Estradasphere - https://ocremix.org/remix/OCR00577 Escape Route - Cave Story - Corran - https://ocremix.org/remix/OCR01513
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Hello friends, we are back of episode 232 of the R Weekly Highlights Podcast and if you're new to the show, this is the weekly podcast where we talk about the awesome highlights that are shared and this is this week's issue of R Weekly that you can find at rweekly.org. My name is Eric Nance, hardly June is almost over, there's a lot going on for many of us, I know I'll be on vacation next week so that's a little PSA right now, we probably will not have an episode next week because yours truly is not going to be near a microphone
but you know I digress. In any event, this time I am not alone, back from his recent, trip adventures is my awesome co host Mike Thomas. Mike, how are you doing today? I'm doing well, Eric. Yeah, it's summertime, so schedules are tricky, but we will try to do a doubly good job in the highlights today to make up for the fact that we'll probably skip next week. Yes, unless you wanna take the reins on your own, you know, I'll never turn you down on that, but Your tech setup is probably beyond what I can handle. Yeah. You just give it a month, you could get to it, but maybe for later. But let's, let's use this tech setup to our advantage here. Those of you who are aware, I have, revamped my setup and I've solved yet more bugs and hopefully that will be reflected in the recording of this. But what are we talking about today? Well, we're talking about the latest issue
for week twenty six of this year, and who curated this one? Somewhat unexpectedly. It was yours truly, but it was it was a good pitch in when needed, and we got the issue out slightly later than we usually do, but it's here and we got some awesome stuff to talk about. But as I always say, well for myself and the rest of the curators, we always get tremendous help from all of you in the community with your poll requests and other wonderful suggestions. So let's dive right into it and we're kicking off the highlights with a very
fun and rewarding highlight and that's almost literally here because the very reason we are here on this podcast, especially for me, is the R language itself, that is obviously the focus of R weekly, right, and I am very pleased to announce that this year the R project itself has been awarded the 2026 Russo Prize for statistics.
This is a prize that admittedly I did not know about until this highlight, but, the award for this is $1,000,000 spread through the five, you know, core team members that I'll list here shortly, and to the rest of the core team members. So who is getting the the prize? Well, I can tell you right now it is the fantastic
quintet of Doctor. Brian Ripley, Martin Moxkler, Kurt Hornick, Peter Delgaard, and Luke Tierney. They get a split of that first half of the 500,000 and then the other half goes to the rest of the core team members, and this has been awarded for the R project because of its massive impact on statistics, the open nature of R itself, and it is long
overdue recognition. We know from being in this community and seeing the growth of ours since we started using it, in fact, I did not start using it when it was first out, contrary to what many of you might think, I did wait a few years until I got my hands on it, but the explosion of growth, the explosion of adoption is thanks to many reasons, right? It is based on the very solid underpinnings of BASAR itself, which those of you may know was inspired by the s language
with Rasahaka and Robert Gentleman were inspired by s to try and make a more flexible and open version of that. That's just one part of the story, right? The other part is that unlike, know, some other languages that have three letters in it, this is truly open, released under the GPL license and it's not just that the language itself is on its own, which is already great, it is obviously the extension system, the packaging system that we have however many thousands of packages now, I've lost count now since the last time I looked up that metric.
That is a massive reason why r is where it is today, but it can't be where it is without the foundation that the R Core team has laid out. Just like any project that has its ups and downs, right, and we've been very transparent about some of the things that happen in the community and with the project itself, but even with that it lives on, it's not going away anytime soon, there has been renewed effort that we covered in previous highlights on getting new, new members in statistics and data science to help contribute
to the project itself. I know it's a huge focus for many people and I would not be where I am today without it. So it is a huge congratulations on behalf of our Humble Little Podcast here to share you know the very you know big congratulations to the r corps team for being awarded the twenty twenty six russo prize for statistics I am very excited for them and you may be wondering what else is in store for this award they will have an award ceremony
later this year on November 4 in Lovun, Belgium, as a formal ceremony. So I'll be watching for that, I don't know if they'll record it or not, but one way or another it'll be great to see the R Core team at that event. So yes, huge congrats to the R team, it is well deserved. So this prize is supposedly the Nobel Prize for statistics and comes with this incredible million dollar prize as well and, you know, how well deserved this is for the the team that's put this together. And, Eric, our little language, ours is growing up.
Incredible. It's out of the teenage years, Mike. Who would have thought? I can't believe it. It just warms my heart, on this this nice Friday. Great way to go into the weekend for us. And, you know, some of the the publication that that comes out of this announcement is is interesting in sort of what's highlighted. You know, the fact that R is now considered trusted infrastructure for, you know, very high stakes things like the work that's done at the United States Food and Drug Administration to decide sort of what what food and and drugs go on the market and it's safe for the public consumption.
Major pharmaceutical companies, obviously, the R Pharma Group that you've had quite a hand in, Eric, has driven a lot of that and global central banks like the Bank of England are are referenced as, you know, heavy R users in their critical day to day work and I can attest to that, you know, working in the the banking and finance space particularly in the niche that we work in, the the farm credit system which is a network of banks in The United States that lends money specifically to farmers and agricultural businesses.
We work with a lot of, you know, data scientists and data analysts at those these farm credit banks, and they're about eighty twenty r to Python in favor of r, which is pretty incredible. Yeah. Yeah. And, you know, one of the the other interesting things in the award recognition that they they call out is ours sort of massive volume of of published transcripts that's now being used by AI agents to learn everyday data tasks. So,
yeah, they're keeping up with the times in this award as well. I know obviously the positive team is is doing a lot of work to enable R to be as useful as possible with the new sort of AI agentic coding world that we live in. But this is just a a fantastic recognition for the decades of of hard work that have gone into this R project. And, again, it's easy for us sometimes to take for granted, especially when we're leveraging open source software, all the hard work that has gone into it and and really just the underpinnings
of, what we have. And I'm so glad that it's recognized in this this great, great way. It is, you know, like you said, very easy to just go to your day to day and just rely on this being there and not even thinking too much more of it. And you know, as just an open source purist, it is just, you know, very gratifying that when I was doing some of my first work in grad school with R for the first time learning it, that I didn't realize at the time the impact it was going to have. It was right now just a way to do some coursework and in the
future after that my dissertation, and that's when the light bulb really went off that, hey, this is capable of things that proprietary software is not because it lets the researchers of a statistical methodology write the package with their manuscript that saved my dissertation. That is the stuff that you cannot articulate enough, at least I cannot, of how game changing that was and what would eventually be. We're, I'm using R today and being able to spread that
enthusiasm in different ways, think is, yeah, it's infectious to me and hopefully nothing will happen in terms of the future that would be a detriment to the project. We all know that open source does evolve, but it's not a direct one to one, but I feel like R has the long staying power of the Linux kernel itself with how entrenched it is in data science and statistics. Yes, other languages are going to catch up eventually on some of the things that R is in the front of, but in the end, it's never an either or, right? We often go through multiple frameworks,
but in this day and age, good luck finding anything that has the breadth of what R has on the statistics side, which is what this award is very much all about. So yeah, congrats again to the RCORE team and here's to many more years of success for for them as well as the community at large. And as I alluded to, we are in a mood for celebration in this episode of R Review Heights because we just celebrated the maintainers
of the R project itself and getting their great recognition with that award. Well, there's also the community of packages that, you know, have many authors and contributors around the world, and one organization in particular has been leading that charge of best in class in reproducibility, novel statistical workflows, and technical achievements with the R language, and that is of course the R Open Sci project.
And we are happy to say that throughout the month of June, it has been the celebration of the many maintainers that have been part of the rOpenSci umbrella with their innovative packages. And in particular, our next highlight is a blog post coming from the rOpenSci blog by Nenina Bellini Salbenin, I never get that right, but I'm trying my best, and she blogs here about the different aspects that have been performed by ropensci in this month of recognition.
One of which is, if you do follow ropensci on the social medias out there, you will have seen that throughout the month, they have actually spotlighted a maintainer, you know, very specifically throughout the days. And in fact, this blog post has a great little hex graphic here of all the maintainers and their pics here, and you can see exactly who they are, but it is a diverse group, just like R itself, the community in large. We have 37 maintainers in total from 15 countries represented around the world, maintaining
more than 50 packages that definitely serve hundreds, thousands of researchers, statisticians, data scientists, you name it, they are all out there. So they've had a a wealth of a campaign this month to have about 39 posts across the across LinkedIn and Mastodon, and that's automatically a bridge of blue sky. So all the three of the the big ones that we see in the social, circles.
And you get a snippet of each of the each of the maintainers here in the post. I know in particular, I worked very closely with the esteemed Will Landau, author of the Targets package. He is one of those listed here, but there's a who's who of ours here that if you've been using those packages and maybe didn't know the the face behind the package, this is a great chance to get in touch with them, maybe send them some kudos and thank yous for all the tireless work that they do to maintain these packages.
It's not easy, folks. It can even test to me being a an author internally of packages, but also very few externally, my most recent one being Shiny State, and I got a lot of work to do to make that up to date in the coming weeks. But it is sometimes a thankless effort, right? Sometimes we don't hear about all the success stories and a maintainer might just hear when things break. So it's a great time this month as June is wrapping up to give a little positive vibes, positive shout to these wonderful maintainers under our open size umbrella,
and I'm, and again that should be a year long effort, but it's great that at least in the month of June we have the R Open Sigh effort really doing a great job to recognize this esteemed group that's highlighted here. Yeah and I feel like just something about the language that there's just really this community aspect around it that our open side, you know, is one of the organizations that helps to foster and, you know, obviously, they do a whole lot more than just sort of community development and the work that they do is fantastic but this this spotlight, as you mentioned, Eric, is important to, you know, kind of like the the prize that we just discussed to step back a little bit and you know, just recognize what we have and be grateful for what we have and the people that make it possible and it it certainly takes the community to obviously,
you know, develop this rich package ecosystem but also maintain it and let maintainers know when there's an issue with a package through filing detailed issues hopefully and being able to contribute in this this way that obviously makes open source what it is and allows us to all kind of work together on the language and the tooling in a way that makes it as as robust as possible. So, this is a great shout out. I know you shouted out Will. You know, a couple other notable mentions. I think here in this blog post, Yaron Ooms for
his work on foundational tools like magic and PDF tools and Julia Silge, for all of her work. I think one highlight here was on Qualtrics and you know, really sort of this this message is, you know, the acknowledgement of a lot of work that often goes unacknowledged and if you can, I I say this quite often but if you have the the means or your organization has the means to contribute to you know, through either your time or your your money to efforts like our open side? I think Numb Focus
is sort of the group that lives behind that from a nonprofit standpoint, but, you know, a great, great opportunity to continue to keep these communities going and these organizations going that do so much for us. Absolutely, and again, that's a great, you know, concept that I've been learning even in my podcasting adventures, giving value back can take many different forms. Yes, there is the financial side of it. There's also the time that you can take to even just report issues or share success stories
in whatever means that you can, and also contributing back if you can put that fix in there. That positive reinforcement can take many forms. There's a phrase that's been coined in the people I listen to called value for value. If you're getting value from a open source project or a podcast or any other endeavor that the creator or the author is putting their time and effort into it, you have many ways to give that kind of value back. So I, yeah, don't hesitate to find the way that works for you and again, it's great to
recognize these folks from time to time, but yeah, if you have the means, I definitely echo what you say, Mike, and we should be front and center on making that a regular thing going forward. And rounding out our highlights today, you know, on top of leveraging R in our daily work, we are, of course, leveraging new ways of authoring our code, authoring our programming, and that is often through the help of an agent, an assistant in the world of AI.
And I know for me, it felt very intimidating to get into this side of it before I really started practicing with it, but I fully acknowledge there are certain aspects of this I still don't quite understand and or I just don't have enough practice enough. So I always appreciate, and in fact, I've heard this feedback from community members that attended a recent event that we had in our pharma called GenAI Day. Sometimes there is a lot of assumptions done by our presenters or those giving like a virtual workshop that people have kind of been through
experiences with leveraging AI for developing code, and some just may not have made that leap yet, they just want a, you know, kind of like a curated set of tips to help them along the way. Well, our last highlight definitely does just that. We have the esteemed Sharon Machlis giving us her 10 tips for getting better R code from your AI coding agent of choice. So Mike and I will kind of highlight some of the tips that resonate with us here. There are some that we have definitely talked about in the show before, but it's great to see again this list put together in an accessible and digestible fashion.
I think one thing that we have covered a lot is the use of skills, which amount to what are basically markdown files or some specific front matter at the top to instruct your coding agent, your AI agent, to leverage that when it needs to tackle a problem that deals with what that skill is addressing. So I've been very much a consumer of skills. I still haven't actually authored my own skill yet, but that is on my list this year. But it's one thing you just consume them once,
but one thing that's important that Sharon highlights here in one of her tips, you can make these evolve over time, whether they're what you created or whatever you've gotten from the community. Maybe there is something you can take from the community, you can morph it a bit to what you want to do, and that's what Sharon has done in her setup here. She's written her own Customize R skill. It gets inspiration from skills from like the POSIT team and others, but she curates that, or she like maintains that
and adds to it from time to time as she learns things. It's just like anything, right? If you learn a new workflow or learn a new technique, update these skills so that your agent can benefit from it. It's not just a set it and forget it. There is an evolving life cycle to these skills that I greatly appreciate. I think that's a wonderful thing that she highlights here. Another one that is easy to miss, I think, when you start with this, because some agents kind of do a good job of it on its own and others just kind of go full bore right away of your request,
you want to let it plan things first. That is what often is called plan mode, and when you ask it to explicitly come up with a plan before it actually updates files, most of the agents, like Clog Code and the like, will literally show you a markdown type of format of the plan and you will be able to look at that in great detail, and you can tell it before it makes any edits if you like what you see or if there are things you want to tweak. Sometimes just that extra step actually helps the agent on its own to keep on task
instead of going full crazy with the edits right away without that plan in place. So I think that is something that I typically see when I ask it a pretty ambitious kind of ask, but sometimes I'm just, you know, very careful and say, even if it looks like a trivial thing, come up with a plan first, especially if I'm building a Shiny app and I'm kind of opinionated about, don't stray away from like the Goldman framework or don't stray away from like BSLIB for the UI or things like that.
Want to see that as a plan in front of me before I let that go, you know, crazy with things. The other thing that I have not been as good on, at least learning on, but I've seen other framers kind of address it, is how these agents learn from past experience. There are certain frameworks like Hermes and the like that will literally have a memory file that it will update in time as it's carrying out like a task, maybe
a task on a repeated fashion or whatnot, but I'm not quite sure how to best approach this from a day to day perspective. There are some agents like Claude Coat do come with some built in auto memory that I think happens behind the scenes, but it may not always be guaranteed. But that's where making sure you keep up to date with a Claude MD you might put in your project and or agents MD, depending on which framework you're in, I think can be quite helpful as you find yourself repeating certain things.
The combination of a memory and the skills, I think, can be pretty fruitful. But that is an area I want to explore more fully, but it was great that Sharon called that out here. But, Mike, you've seen this post here. What are some of the tips that resonated with you? Yeah. I'll highlight a couple more. One that I didn't use enough when I started working with ClaudeCode but now has become, you know, absolutely necessary is that root level Claude. Md or agents. Md file as you mentioned. Or I guess if you are, on the Google train, you can do a Gemini. Md knowledge file as well. And
the way that you can think about of it is it's sort of like a read me. It gives a background for your project, but it may also give some information on, like, the project structure that Claude can take a look at, you know, what sort of files live in which directories. You know, we do a lot of things like, hey. If I'm developing this package, I wanna make sure that we're sticking to these particular packages for data wrangling. You know, say I wanna stay in the tidyverse or on the Python side, we'll say, hey, I wanna use polars. I don't wanna use pandas.
You know, if you need more context on something, you can stick a hyperlink in there that Claude will be able to hit. And, they're just super helpful. You can kind of think of them as like a read me for your coding agent that every time you initialize a new, you know, agentic coding session, this file will kinda be the first thing that it looks at to get context and provide some of the harnessing to rein in, your coding assistant to leverage the practices that you want it to leverage.
Another thing that I will highlight here kind of in that that same vein is make sure that your agent is is writing tests and doing code reviews is is one recommendation from Sharon here. I would also say that, you know, make sure that it writes tests early and that any changes that it does to your code base after those changes are made, it it re executes those unit tests to ensure that, you know, any of these coding changes, if you're kind of letting your agent, you know, do most of the work and you're being a little hands off, you wanna make sure that the code that it's writing isn't breaking
anything else that you've already reviewed and, you know, has passed muster in your opinion. And I think that, you know, leveraging agents to to write unit tests, it not only only can be a great practice, like, defensively, but there's so much opportunity, I think, for when we had to manually write unit tests in the past. It's kind of a friction based process. Like, it's it's not frictionless. Whatever the antonym for that is. Yes. And it takes time, and it it actually takes quite a bit of thinking as well. And, yeah, again,
allowing your agent to write tests or at least start writing some tests for you to get past that blinking cursor or guide it to say, hey, I wanna test this thing specifically. You know, we've been leveraging a lot more, you know, what I call, like, test where you're actually trying to match, you know, the the data output specifically and not just saying, hey, is this function going to throw an error if I pass this type of argument and is it going to be silent if I pass this other type of argument which are the easy tests to write, right? The harder ones are where you're actually, you know, trying to verify real input and output from
your function and you know, agentic coding tools can just really help you get over that hump a lot quicker and there's also no excuse, you know, as Erin as Sharon points out in this particular bullet point here in her blog post that, you know, why not have your agent do code review? I think reviewing its own code can, you know, potentially lead to sort of false positives if you will or Mhmm. False negatives probably where it says, oh, no. This code all looks good, and that's because I wrote it. I know that there's a lot of folks out there that are using one, you know, agent, like, codex to write the code and then Cloud Code to review it or vice versa. I think actually the opposite is what we're seeing more often in the wild. But,
if you have a repository that you've worked on in the past or something that you're working on now without agentic coding, you know, why not let your your coding agent of choice take a look at it and review it? And I think you'll be quite surprised at at what it might find in terms of, you know, gaps in your code that it can just tighten and harden things up for you a lot of the time. So that's, another great tip. And I guess the last one that I will talk about here in her blog post is the use of the BTW
R package. I don't know if we say by the way or BTW. I think that's the shorthand for what the kids say nowadays. But it is a fantastic little package that sort of solves, you know, this issue where these models are trained on some R code out there in the wild and maybe historical information that's not necessarily up to date, may not sort of match the versions of the packages that you have installed on your system.
And this BTW package really lets your coding agent access all the information about those packages that are installed in your environment. So it can write code that is sort of specific to the environment that you have set up for your project. And, you know, BTW, you know, lets these LLMs access variables in your current R environment via a an MCP server, which is really cool as well. And I think enables, you know, Cloud Code to provide you with the most,
relevant output for the project at hand, your code style, the packages that you're specifically working with in those particular versions, to make sure you're not using deprecated functions and things like that. And I think just be a much stronger coding assistant with this MCP capability. So if you haven't tried out the BTW package in R and you're using agentic coding, there's really no reason not to. Yeah. I need to well, I'm one of those people who needs to take heed of what you said. I've heard of it. I have experimented a bit with MCB tools, but I haven't put the two together yet. So I definitely want try that out because I see in my sessions,
Claude will just spin up either trying to grab the documentation from like where the package was installed on its own and may not find it, and or try to go online to find it, but giving BTW as the bridge between what is actually there in a way that native R can understand and giving Claude or whatever your agent of choice is to grab that information. It just makes things a lot easier. So yeah, that is an area I definitely need to get better with and it's, I'm glad to hear that it sounds like you've been having some great success with that workflow. I'm sure it makes things a lot easier for you. Definitely. And hopefully, the rest of the issue will make things easier for you to keep up with what's happening in the R community. That's what Art Weekly is all about, right? So this issue, it was a tough choice to figure out the top three for the highlights here. We had a lot of great additional posts here.
I'll just shout out a couple real quick here. Sometimes just looking at what users struggle with, and that's where Albert Rapp's got a nice little fit to the point post on what he's seen reported to him on what's kind of the building, the learning curve that some of the newer R users might have and some other additional things based on what you just shouted out, Mike, snapshot testing testing by Jacob Sobelowski. He's got another wonderful comprehensive post that goes well beyond just getting that fancy screenshot. It is much more than that. And he does a great job of highlighting those
details. So yes, go to rweekly.org to find all those wonderful additional links, new packages, updated packages, and much, much more. And where can you keep up or how can you contribute to the project? Like I said, there's a small handful of us maintaining this effort, keeping the lights on and trying to progress, but yeah, we rely on all of you for contributions of the packages you're writing or you know about, or great use cases of RN Data Science.
We are just a poll request away. You go to rw.org, find that little Octocat thing in the upper right corner, click that, and you will get a link to our poll request template. You can update the issue draft right there and give us that awesome link that you found. And we love hearing from you in the community, much like how OpenSci has done their social media spreading out of the maintainer, kudos, we are also on those same services. I am on BlueSky, where I am at rpodcastbksky.social.
I'm also on Mastodon, or I'm rpodcastspodcastindex.social and I'm on LinkedIn, search my name and you'll find me there and I hope by the time you're listening to this or shortly thereafter, I will have also shared the links to the recent Gen AI Day and R Pharma recordings that I had stayed up wearing late last night editing, but I got it done. So, Mike, where can the listeners find you? You can find me on blue sky at mike thomas. Bsky.social or on LinkedIn if you search Ketchbrooke Analytics, k e t c h b r o o k, you can see what I'm up to.
Awesome stuff. And like I said at the outset, there is a very, large chance that we will be not we will not be back next week because yours truly is gonna be away from a mic. And unless my cohost, Mike, is very adventurous, I doubt he'll be doing that, but we'll find out. Either way, we hope we thank you so much for, listening to this week's episode of our Rookie Highlights. And yeah, when we are back we will be happy to have you back for a new episode which will likely be in two weeks. Thanks again for listening.
Our takes on the increasing removal of friction in good and not-so-good ways thanks to AI, a visual tour of hot air balloon accident data, and a case study of novel image recognition with the rplatipus package. Episode...
The flow of both data and analytical jobs just got smoother with mizu, opening the door for multiple agent harnesses in Positron, and an optimal statistical model tailored to complex experimental designs with a friendly...
What happens when that online data set isn't so easy to download anymore? We learn of two packages to ease the pain of obtaining these files and then some. Plus a visit to the visualization corner with a handy way to...
The celebration of rOpenSci's 15-year anniversary continues with the community sharing their perspectives, and a fun Shiny app that could be quite useful during celebrations on a weekend or the holidays. If you ever...
We learn about the variety of ways you can contribute back to the core of R, a roundup of the excellent choices available for package and environment reproducibility, and an amazing collection of Shiny applications and...