Kyle Walker

Enjoyed the podcast by @DKThomp and @conorsen on American migration and housing trends.

They observed that Americans are increasingly choosing mid-size Southern metros over larger ones like Atlanta, so I took at look at the latest numbers (metros 300k+).

Checks out!

More Turf.js / #rstats / Shiny experiments:

Using a client-side buffer and client-side difference operation to smoothly erase areas from another feature.

Keep an eye out here for more examples and a release announcement!

it's working pretty well so far!

I'm designing it so everything happens in the map, so it's not a competitor to sf or anything like that. It'll be most useful in Shiny when you want to do spatial ops in-app and avoid an expensive round-trip to R/sf to do it

Though you can use it outside Shiny too

Client-side #GIS via TurfJS is coming to the #rstats mapgl package...

Shown here: Voronoi polygons around a circle layer created *in the map* - no external GIS software or R functions required

Tons of Shiny implications here that I can't wait to show you

Ha, that’s not in Mapbox GL JS yet!

For #rstats users: this uses an experimental branch of the mapgl package that wraps Turf.js to generate the circle on mousemove.

When this is ready, you'll be able to do client-side GIS in R - which opens up a whole world of possibilities for fast spatial analytics in Shiny!

In web mapping, understanding projections and distortion is critical - but far too many maps *still* use Mercator for global views.

Watch the video which shows how a 500km circle gets distorted around the globe using different map projections in @mapbox.com GL JS.

All data are in 2023 inflation-adjusted dollars from the 2019-2023 ACS, via the #rstats tidycensus package.

Learn how to analyze and map this data yourselves:

walker-data.com/tidy...

walker-data.com/mapgl

💰 #5: Census tract 301, Western Connecticut Planning Region (Darien)

Mean household income: $703,799

💰 #4: Census tract 47, Westchester County NY (Bronxville)

Mean household income: $712,542

💰 #3: Census tract 4112, Harris County Texas (Houston / River Oaks)

Mean household income: $739,421

💰 #2: Census tract 99, Westchester County NY (Scarsdale)

Mean household income: $817,621

💰 #1: Census tract 197, Dallas County Texas (Highland Park)

Mean household income: $854,857

America's wealthiest Census tracts by mean household income

While mean incomes can be misleading due to skewed distributions, they offer insights not available from median household income in the ACS, which is top-coded at $250,001.

The top 5 (minimum 500 households):

Will models get better at this? Probably

But I still have a hard time seeing a future where humans are not "in the loop" way more than the hypesters suggest

I say this all the time - your expertise still matters!

I get the excitement about AI subagents that can work together on projects

But the further you get away from the work being done, the more you are trusting the LLM to make "correct" decisions

And LLMs still make incorrect decisions where the answer would be obvious to you

These Are America’s 50 Wealthiest Suburbs in 2025 Six new suburbs have emerged in our 2025 rankings.

Link to the original article: www.gobankingrates.com/money/wealth...

Means can be misleading, but so too can median household incomes in the ACS which are top-coded at $250,001.

Example: the mean income of the top 5% of households in Scarsdale (ACS var B19081_006) is $2.16 million!

Worth thinking about if you are analyzing income with the ACS.

tidycensus::get_acs(
geography = "place",
variables = c("B19025_001", "B19001_001"),
state = "NY",
geometry = TRUE,
output = "wide"
) |>
dplyr::mutate(mean_income = round(B19025_001E / B19001_001E)) |>
mapgl::maplibre_view(column = "mean_income")

I'm intrigued by this study of America's wealthiest suburbs by @GOBankingRates.

Scarsdale, NY tops the list at over $600k mean household income; Fort Worth-area Southlake ranks 7th.

If you want to reproduce the analysis, here's some tidycensus #rstats code you can use:

From the 2024 #30DayMapChallenge: a mobile-first #rstats Shiny app that geolocates you then shows you how far you can walk in 10 minutes.

Try it out: walkerke.shinyapps.io/day-30-final...

Code: github.com/walkerke/map...

To build it:

-- Scaffold in XCode (Claude Desktop helped me set it up)
-- Design with Claude Code in Zed
-- Build and test in XCode / on my phone

If the concept is straightforward enough, you can go a long way without knowing Swift or mobile development. This only took me a few hours total

Then, when you've finished your drawing;

-- Download it as a GeoJSON file with `download_button = TRUE` (a unique feature to mapgl!);

-- Use it in your Shiny apps (try it for lasso selection!) with `get_drawn_features()`.

Learn more: walker-data.com/mapg...

The draw control in #rstats's {mapgl} package has three polygon modes:

-- Use the default to draw a polygon vertex by vertex;

-- Use `freehand = TRUE` to draw lasso-like polygons by clicking and dragging;

-- Use `simplify_freehand = TRUE` to get coarser freehand output.

I’m already using it a ton!

If I get enough interest I’ll publish to the App Store for $0.99.

If this interests you, like the post or comment “QR” and I’ll send you info when it’s ready!

A use-case for vibe coding: building iOS apps for personal use.

It’s often a pain when I’m sent a document with QR codes on my phone- I can’t point my camera at it!

So I built an app that bulk extracts QR codes from images and PDFs then saves links / allows for PDF export.

The Zed editor @zeddotdev is a masterclass in space usage for an IDE

When I’m developing apps & software I need a clean, uncluttered UI and Zed is the best I’ve found

It’s why I’ve struggled to use VSCode and its derivatives (Cursor, Positron) for dev work; too much going on

Building an AI-powered location explorer with Shiny and Claude – WALKER DATA GIS, demographics, and data science consulting

On the blog: building an AI-powered location explorer with Shiny and Claude

Intrigued by apps you’ve seen that bridge GIS and LLMs? You can build them yourselves in #rstats!

Read the post: walker-data.com/posts/ai-loc...

A dynamic, interactive heatmap over a 3D globe?

It’s all possible with @mapbox.com and #rstats.

Learn how: walker-data.com/mapgl/articl...

How Anthropic teams use Claude Code Discover how Anthropic's internal teams leverage Claude Code for development workflows, from debugging to code assistance.

These are the most relevant points to my work, but there are TONS of other insights that will be useful to you.

Read the guide: www.anthropic.com/ne...

How are you using these tools?