iiinova — where your digital self keeps evolving

Xan Gregg @xangregg.bsky.social · 04/10/2026

The 2026 World Jigsaw Puzzle Championship just wrapped up and solvers continue to get faster. Here're trends of recent top finishers with 2 stories: the consistency of former champion León and the sharp 2026 improvement of Peng. #dataviz #jigsaw worldjigsawpuzzlechampionship.com

Smooth line chart titled "Top world jigsaw puzzle championship finishers". The x axis shows time from 2022 to present. The y axis shows the time it takes a solver to finish a 500 piece puzzle. Each smooth line represents the trend of a different competitor. Two are highlighted. Alejandro Clemente León has a relative flat trend line at around 35 minutes. Andrea Peng trend starts flat at around 45 minutes and then falls sharply in 2026 to the low 30s. (She finished 4th in 2026)

040

Xan Gregg @xangregg.bsky.social · 02/10/2026

I made a version with region aggregation to get Europe and Asia on the scale of Canada and Mexico. Also tried a P-spline smoother but with a break for Covid. Preliminary data (Jul, Aug) not shown, so not getting the 2026 summer/World Cup bump. #dataviz

Remake of the quoted chart. A line chart over time, 2000 to 2026 on the X, and monthly international visitor counts on the Y axis. Four lines are shown: Canada, Mexico, Western Europe and Asia. There is a break in the lines around the start of Covid19. Canada shows a prominent decline around the 2025.

040

Xan Gregg @xangregg.bsky.social · 20/09/2026

Is this a hollow bar chart, some kind of a clever compromise for those who want and those who don't want summary bars? From @schreinerdrew.bsky.social in Nature: www.nature.com/articles/s41... Excellent data sharing makes for a quick remake try using just lines at the means. #dataviz

Chart from https://www.nature.com/articles/s41586-026-10510-x, with 3 categorical values on the x axis. Each category has some dots and what look like an unfilled bar chart (just the bar edges).

Remake of previous chart with the hollow bars replaced the horizontal lines when the tops of the bars were. And different jitter.

072

Xan Gregg @xangregg.bsky.social · 16/09/2026

Here's a makeover retry using the same colors. Not entirely happy with the labeling, but interesting to try to mix continuous and categorical nature of the time variable using stacked area and stacked bars. The 80s vs 90s shift is easier to see. #dataviz

Stacked area chart remake of the sequence of pie charts in the parent post. For 4 time points, the chart shows the percent of albums from each decade in the top 500 list.

3162

Xan Gregg @xangregg.bsky.social · 14/09/2026

When do you stop adding pies and use a different chart for a time samples? Pies from en.wikipedia.org/wiki/Rolling.... Data in ALT text for a #dataviz mini makeover challenge.

Four pies charts from the Wikipedia article https://en.wikipedia.org/wiki/Rolling_Stone's_500_Greatest_Albums_of_All_Time. Each pie is a different publication year of the list and each wedge is a decade in album release dates.

CSV data below. "Count" is the count provided in the pie chart legend. "Scaled Count" is the Count value scaled so the total for each edition is 500. The raw 2003 numbers only sum to 482 -- not sure why.

Decade,Count,Scaled Count,Edition
1950s,11,11,2003
1960s,126,131,2003
1970s,183,190,2003
1980s,88,91,2003
1990s,61,63,2003
2000s,13,14,2003
1950s,10,10,2012
1960s,105,105,2012
1970s,186,186,2012
1980s,84,84,2012
1990s,73,73,2012
2000s,40,40,2012
2010s,2,2,2012
1950s,9,9,2020
1960s,74,74,2020
1970s,157,157,2020
1980s,71,71,2020
1990s,103,103,2020
2000s,50,50,2020
2010s,36,36,2020
1950s,9,9,2023
1960s,71,71,2023
1970s,155,155,2023
1980s,71,71,2023
1990s,101,101,2023
2000s,51,51,2023
2010s,36,36,2023
2020s,6,6,2023

240

Xan Gregg @xangregg.bsky.social · 13/09/2026

A new Neil Sloane video on Numberphile, in case you need a dose of math joy. youtu.be/A020pGO5MBs?...

youtu.be

The Immortal Kangaroo Sequence - Numberphile

YouTube video by Numberphile

120

Xan Gregg @xangregg.bsky.social · 09/09/2026

I'll be working the demo floor at JMP Discovery Summit in Baltimore next month. Bring your Graph Builder questions! community.jmp.com/t5/Discovery...

Photo of an event space for the JMP Discovery Summit

000

Xan Gregg @xangregg.bsky.social · 30/08/2026

World's strangest broken axis? It looks to be a negative break, with the 60-120 distance being greater than the 0-60 distance. Or does the ⊣⊢ signify something else? #dataviz

Line chart from a research paper (https://www.science.org/doi/10.1126/sciadv.aec0481), with x axis having ticks at 0, 15, 45, 60, 120. Above the 60 on the axis is a break indication, but the 120 is spaced farther from 60 than 0 is.

210

Xan Gregg @xangregg.bsky.social · 29/08/2026

Cool to see my Steepest webapp in @visualisingdata.com's newsletter. Unfortunately, Carto just started requiring API keys, so there's a watermark until I get that resolved. xangregg.github.io/steepest

Screenshot of Steepest Roads app showing a map with roads colored by steepness and an API watermark,

000

Xan Gregg @xangregg.bsky.social · 14/08/2026

My Arya list for #dataviz: radar charts, log-scale bars, smiling beeswarms, gratuitous quadratics.

bar chart with 4 bars and a log 10 scale y axis. Bars start at 1.

Dot plot with classic beeswarm jitter tendrils.

Scatter plot with overlaid fitted curves. One is quadratic, mostly dominated by one outlying point.

010

Xan Gregg @xangregg.bsky.social · 13/08/2026

Nice detailed heatmap but I can't help think of all the dots that are hidden under the state borders. #dataviz

Close-up screenshot of the quoted post's drought map near Delaware/Maryland when border lines are almost as big as the state interiors.

060

Reposted by Xan Gregg

Brad Jones @bradjones.bsky.social · 10/08/2026

Did you know that Coloradans have wagered more than $1B on table tennis since 2020? I sure didn't. Sports betting is a little bit of a foreign topic for me, but it was eye-opening digging into the numbers. www.co-political-landscape.org/1-billion-on...

A graphic showing the monthly wagers on table tennis. In 2020, Coloradans were betting an average of $10M each month on table tennis games. By 2025, the average monthly wager had increased to $30M.

Notes: Each point shows the total online wagers for table tennis in that month. The grey line shows the smoothed average. Source: Colorado Department of Revenue Sports Betting Monthly Reports (2020-2026).

2228

Xan Gregg @xangregg.bsky.social · 09/08/2026

Chess position recall by age and rating. The paper found age insignificant because it only looked at its linear effect, but recall drops after age 50 regardless of rating. #dataviz

Contour plots showing chess position recall for random chess positions and real chess positions against age and Elo rating. Using data from paper "Recall of Briefly Presented Chess Positions and Its Relation to Chess Skill" https://pmc.ncbi.nlm.nih.gov/articles/PMC4361603/

032

Xan Gregg @xangregg.bsky.social · 29/07/2026

I added a single-color mode to my steepest-roads app to put everything on an equal scale instead of highlighting the top ranked roads in a different color. xangregg.github.io/steepest/

screenshot of a map with roads colored by steepness and a color:Dual|Single button.

030

Xan Gregg @xangregg.bsky.social · 26/07/2026

Before I had an e-bike, I planned my bike trips to avoid steep hills and dreamed of an app to help. Now with AI, I can build my own. Red for steepest, indigo for others, yellow underlay for long inclines. Varying road width for elevation is my #dataviz flourish. xangregg.github.io/steepest/

A street map snippet with some road segments colored in gradients of red or indigo to indicate steepness and with varying width to indication direction.

080

Xan Gregg @xangregg.bsky.social · 23/07/2026

The green lines are standard deviations above the model in red. Interesting way to see residuals while keeping the data in its original context. #dataviz

070

Xan Gregg @xangregg.bsky.social · 11/07/2026

That far left data point has an out-sized amount of leverage on the quadratic fits. For comparison, here's a smoother on the data behind the middle image.

Reproduction of section B in the quoted chart using p-spline smoothers instead of polynomial regression.

030

Xan Gregg @xangregg.bsky.social · 05/07/2026

This fit didn't look like p=0.020 to me, so I extracted the data and checked. It looks like the Line+CI is for all the data, but the stats exclude the low outlier points. #dataviz jov.arvojournals.org/article.aspx...

Chart with regression line and 18 data points. The confidence interval looks wider than it should for the given p-value of 0.020.

Chart with regression line and 18 data points. Same as fit as before but showing a p-value of 0.0677

Chart with regression line and 18 data points, with one outlier point marked with an X and excluded from the fit which has a p-value of 0.020.

182

Xan Gregg @xangregg.bsky.social · 05/07/2026

The caption says "Each box represents the median and upper/lower quartiles with the whiskers showing the 5th and 95th percentiles, and any points beyond this range shown individually." Why not use a standard box plot? Is this a common/named variation? #dataviz journals.plos.org/plosone/arti...

Figure 3 from the cited paper. Showing what appear to be six box plots.

021

Xan Gregg @xangregg.bsky.social · 03/07/2026

This chart from www.researchgate.net/publication/... is akin to the ones in my paired charts study. Interesting how the two groups are the same by any statistical measure but the box plots and violins could pass as different. Original and my repros #dataviz

Chart with correlation on the Y and two categorical groups on the X. Each group is shown as overlaid dots, box plot and violin plot.

Chart with correlation on the Y and two categorical groups on the X. Each group is shown as overlaid dots, mean line, and confidence interval.

000

Xan Gregg @xangregg.bsky.social · 26/06/2026

Mildly annoyed that the caption says the dots are at 10 week increments, but there are 17 dots on the left side and 15 dots on the right side (12.7 week incr). Surely just a caption error not affecting the analysis, but still... #dataviz Dementia vs shingles vax paper www.nature.com/articles/s41...

Figure 3a from the cited paper. Figure title is "The effect of the zoster vaccine on new diagnoses of dementia". Chart shows a center 0 vertical line with separate dots and trend lines on the left and right to show dementia rates vs age.

140

Xan Gregg @xangregg.bsky.social · 13/06/2026

Just realized that with a different color scheme, my baguette prices dot plot starts to look like baguettes. #dataviz

Dot plot of baguette prices with main cities assigned a unique color. There are tall stacks of dots at common prices that resemble baguette when using yellow and orange shades for the dots.

281

Xan Gregg @xangregg.bsky.social · 10/06/2026

How would you remake this #dataviz from Nature? My wrong first impression was that ChatGPT usage is declining after being steady at 1%. Scraped data values are in the ALT text in case anyone wants to try it. Original source is pubsonline.informs.org/doi/10.1287/....

Chart from 2021 to 2026 with 4 lines which sum to 1 at each month. The orange line (0-15% AI usage) starts near 1 and starts declining in 2023 while the other lines increase at roughly the same rate. By 2026 they are all around 25%. Count data below, scraped from the original paper cited in the post.

month,n_0_15,n_15_30,n_30_70,n_over_70
2021-01,92,5,1,0
2021-02,81,2,0,0
2021-03,79,5,1,0
2021-04,79,4,0,0
2021-05,91,4,1,0
2021-06,78,2,2,0
2021-07,84,6,0,0
2021-08,83,2,0,0
2021-09,83,6,0,0
2021-10,88,2,0,0
2021-11,81,2,0,0
2021-12,117,9,1,0
2022-01,92,4,1,0
2022-02,87,2,0,0
2022-03,81,3,1,0
2022-04,65,6,0,0
2022-05,77,2,0,0
2022-06,77,4,0,0
2022-07,72,8,1,0
2022-08,72,3,1,0
2022-09,76,6,0,0
2022-10,94,7,1,0
2022-11,86,6,1,0
2022-12,72,8,1,0
2023-01,73,4,0,0
2023-02,87,3,2,2
2023-03,102,3,1,0
2023-04,76,6,0,1
2023-05,73,6,9,0
2023-06,81,3,6,1
2023-07,81,5,12,2
2023-08,80,7,15,5
2023-09,92,10,13,0
2023-10,78,13,15,1
2023-11,75,12,7,1
2023-12,54,8,9,1
2024-01,92,12,14,4
2024-02,74,13,9,5
2024-03,67,11,9,10
2024-04,78,17,19,3
2024-05,80,16,24,3
2024-06,73,16,11,3
2024-07,73,14,14,3
2024-08,60,18,18,8
2024-09,56,10,19,12
2024-10,64,22,22,6
2024-11,52,22,23,10
2024-12,62,22,16,10
2025-01,60,25,25,11
2025-02,59,31,22,12
2025-03,66,27,25,17
2025-04,52,31,33,24
2025-05,47,21,26,17
2025-06,53,30,17,19
2025-07,51,32,31,20
2025-08,50,31,29,17
2025-09,48,24,25,30
2025-10,60,27,27,29
2025-11,50,36,25,29
2025-12,38,31,26,22
2026-01,53,33,31,43
2026-02,46,28,33,31

140

Xan Gregg @xangregg.bsky.social · 08/06/2026

Is it a treemap? Is it a bar chart? It's a packed bar chart! Bonus chart from my look at the data behind Le Baguette Index: number of participating bakeries in each city. Almost 1/5 are in Paris. #dataviz

Packed bar chart showing a breakdown of 1640 bakeries across 150 cities. The top 5 cities are shown as horizontal bars aligned with the left edge, with Paris occupying most of the top row. Other cities are shown as bars stacked above the primary ones.

030

Xan Gregg @xangregg.bsky.social · 04/06/2026

I'm puzzled by this paper's lead chart. Connecting socioeconomic status levels emphasizes that difference, which is not the point of the paper. I reordered the data to connect across time points which helps see the gap widening over time and a different trend shape for the P and HP schools. #dataviz

Paneled line chart of z-scores across three time periods, 4 school categories, 2 genders, and 2 socioeconomic levels (above or below median).
Original https://www.nature.com/articles/s41586-025-09126-4/figures/1

091

Xan Gregg @xangregg.bsky.social · 04/06/2026

The baguette prices at lebaguetteindex.fr present an interesting #dataviz challenge: 1500+ bakeries with very few unique prices. I'm not exactly sure of their jitter method, but I tried a few myself at rawdatastudies.com/2026/06/04/j....

Dot plot of 1640 baguette prices, with generous jitter to avoid tall stacks of dots. https://lebaguetteindex.fr

Dot plot of 1640 baguette prices, with uniform jitter so that each price is represented by a seven-dot-side tower of dots and there is a gap between each tower. https://rawdatastudies.com/2026/06/04/jittering-baguette-prices/

000

Reposted by Xan Gregg

Michael Friendly @datavisfriendly.bsky.social · 03/06/2026

#stats #rstats If you use or teach EDA, you might be interested to learn a bit of its #history, from a _Nightingale_ article a few years ago. Sort of an ethnographic study of those who actually developed EDA nightingaledvs.com/remembrances...

nightingaledvs.com

Remembrances of Things EDA, Nightingale

Published for Tukey Day, June 16, the 107th anniversary of John Tukey's birth This article recounts the origins and influences of the movement in...

0183

Xan Gregg @xangregg.bsky.social · 31/05/2026

Weekly Bluesky posts tagged as #dataviz or #datavis. Highlighting #30DayChartChallenge and #30DayMapChallenge. Are the 30day challenges really so prominent, or so they weigh more heavily in the searchPosts api results?

Bar chart of Weekly Bluesky posts tagged as dataviz/datavis. Subsets that mention 30DayChartChallenge or 30DayMapChallenge are colored blue and orange and peek in April and November.

071

Xan Gregg @xangregg.bsky.social · 30/05/2026

A chess position recall study was lacking any charts, so I made one. The way the curves start to separate for advanced players is potentially interesting, but it's driven by just three participants and the Elo scores are very rough estimates. #dataviz www.tandfonline.com/doi/full/10....

Chart with smooth trend lines and 20 data points each. X axis: "Elo estimated from a 10-position quiz". Y axis: "Avg pieces correct". a red curve is for recalling 20 real chess positions and the blue curve is for recalling 20 random positions. The blue curve has a steady increase vs Elo rating. The red curve is higher and follows a similar rate of increase but then accelerates around an Elo of 1600.

110

Reposted by Xan Gregg

Xan Gregg @xangregg.bsky.social · 24/05/2026

Round 2 of my paired univariate chart survey is open for public testing. Now with a binary response (same or different) and fewer trials (72). See if you can beat my score. #dataviz pairstudy121.pages.dev?group=b3

A sample trial from the survey, showing two vertical box plots with two buttons below them to choose Same source or Different sources.

screenshot of last page of the survey with a reported alignment score of 80%

041

Xan Gregg @xangregg.bsky.social · 25/05/2026

Beeswarm makeover as a slightly smoothed Wilkinson dot plot. Original from www.nature.com/articles/s41.... Another difference: original violins have equal widths; mine have equal areas. #dataviz

Figure 1d from https://www.nature.com/articles/s41586-025-10076-0, showing 700 observations split into four conditions and shown as violin plots overlaid with dots using classic beeswarm jitter.

Remake of the first with smoother dot jitter. (Figure 1d from https://www.nature.com/articles/s41586-025-10076-0, showing 700 observations split into four conditions and shown as violin plots overlaid with dots using classic beeswarm jitter.)

050

Xan Gregg @xangregg.bsky.social · 25/05/2026

My #dataviz of the "change" in US sleep duration seen in the NHANES survey when they changed the wording of the question from "how many hours" to "sleep and wake times". From www.normalcurves.com/sleep-and-ex... after a paper reported the finding without noting the wording change.

Chart of mean sleep times for 8 survey periods (and one placeholder for the incomplete covid era survey). There is a jump (from 7 hours to ~7:45 hours) after the first 5 periods and a corresponding annotation saying "change in question wording". Standard deviations are shown as shaded regions around the mean, spanning about 1.5 hours each.

051

Xan Gregg @xangregg.bsky.social · 24/05/2026

041

Xan Gregg @xangregg.bsky.social · 17/05/2026

Monthly bus ridership for US "college towns" defined as municipal transit systems with >1.5 riders in school months vs summer months. Sorry for the crowded labels and random colors. Using a cycle-constrained p-spline fit so the starts and ends are aligned. #dataviz data.transportation.gov

Smooth trend curves for monthly bus ridership at "college towns". Noticeably dip in the summer and smaller dip at year end.

050

Xan Gregg @xangregg.bsky.social · 17/05/2026

Using US monthly bus ridership data, I ranked "college towns" based on municipal transit ridership ratios between school months and summer months. One of these doesn't belong. #dataviz data.transportation.gov/Public-Trans...

Bar chart of bus ridership ratios for 26 "college towns" with the highest school-to-summer ratios (greater than 1.5). The towns/schools in descending ratio order are:
Harrisonburg, VA (James Madison)
Blacksburg, VA (Virginia Tech)
State College, PA (Penn State)
Ames, IA (Iowa State)
Lubbock, TX (Texas Tech)
Denton, TX (North Texas)
Champaign-Urbana, IL (Illinois U-C)
Lynchburg, VA (Liberty)
West Lafayette, IN (Purdue)
Bloomington, IN (Indiana)
Kenosha, WI (K-12)
East Lansing, MI (Michigan State)
Gainesville, FL (Florida)
Ithaca, NY (Cornell)
Athens, GA (Georgia)
Flagstaff, AZ (Northern Arizona)
Fort Collins, CO (Colorado State)
Santa Cruz, CA (UC Santa Cruz)
Bellingham, WA (Western Washington)
Lawrence, KS (Kansas)
Madison, WI (Wisconsin–Madison)
Tallahassee, FL (Florida State)
Northampton, MA (UMass & others)
Iowa City, IA (Iowa)
St. Cloud, MN (St. Cloud State)
Chapel Hill, NC (UNC Chapel Hill)

100

Xan Gregg @xangregg.bsky.social · 12/05/2026

How dare that sentence length study approximate some novel publication dates with midlife of author. Let's see what it looks like with accurate dates. Much effort later: hmm, no difference in the trend curve. #dataviz (see the original at dragonfly.hypotheses.org/1152)

Chart of 700 novels and their publication dates on X (1820 to 1940) and average sentence lengths on Y (0 to 70). There are two versions of the data, one using low precision dates in blue and the other in red using high precision dates. The trend lines are nearly identical and descending.

020

Reposted by Xan Gregg

Thomas Lin Pedersen @thomasp85.com · 07/05/2026

Small multiples of radar plots are the non-whimsical version of chernoff faces and somewhat hooks into our pattern recognition visual system... That being said, I've never been compelled to make a radar plot...

251

Xan Gregg @xangregg.bsky.social · 02/05/2026

I've posted the raw data from the pilot runs of my paired-chart survey (volunteers and paid Prolific workers). Not enough data to say much about different chart types, but at least some participants have good treatment-response correlation. github.com/xangregg/pai...

Chart with four overlaid trend curves all increasing. x =Wasserstein earth-mover distance; y = survey responses on a four point scale of surprisingness: not, slightly, quite and very. The four curves are for the four injected differences in the pairs: bimodal, spread, skew, and location.

110

Xan Gregg @xangregg.bsky.social · 20/04/2026

Another look at test-mode pairs-study results, comparing friend volunteers (you all) and Prolific workers. Each line is a different chart type—not a lot of difference there. I'm using Kolmogorov-Smirnov distance as a measure of surprise on the X. #dataviz blog: rawdatastudies.com/2026/04/19/v...

Panel of smooth trend lines for two groups of participants: friends and Prolific users. Each line is a different chart type (unlabeled). y = survey response on a 1-4 scale. x = statistical difference measure (Kolmogorov-Smirnov distance)

110

Xan Gregg @xangregg.bsky.social · 13/04/2026

I decided to spring for a round of Prolific workers for my paired-chart study. Not surprisingly a drop in quality from the socials volunteers. I still need to do attention checks. Here's how responses aligned with a stat measure of diff. Still collecting data: xangregg.github.io/pairstudy121...

Paired dot plot with means lines and shaded confidence intervals. One group is volunteer testers and the other is Prolific workers with alignment score on the y axis. The volunteer testers have generally higher scores.

020

Xan Gregg @xangregg.bsky.social · 13/04/2026

I added CSVW csvw.org support for metadata export in my RDS→CSV webapp. The out-of-the-box spec seems to be missing a lot of useful column properties like factor levels and ordering, but you have to start somewhere. xangregg.github.io/commacomma/

Screenshot of comma, comma converter web app with buttons for Download Data and Download Metadata.

010

Xan Gregg @xangregg.bsky.social · 12/04/2026

Found a recent JavaScript parser for R data files at github.com/jackemcphers...; works well enough that I made a wrapper app for CSV conversion xangregg.github.io/commacomma/. Now I no longer have to dig up my R environment to make the conversion when I find an RData file in the wild.

Screenshot of a web app named "Comma,Comma" with just a single drop zone for a file upload, listing a few input file types.

010

Xan Gregg @xangregg.bsky.social · 30/03/2026

Each run has a different mix of chart variations (but same 4 basic types). In case that makes anyone want to try the survey multiple times, go for it!

010

Xan Gregg @xangregg.bsky.social · 29/03/2026

I've updated my pairs study from initial feedback (thanks!): new question framing, rebalanced effects, phone-friendlier and now showing your results at the end. Testing round 2 now open: give it a try at xangregg.github.io/pairstudy121... #dataviz

Screenshot of one trial of the study, showing a pair of box plots with the question "How surprising would it be if samples A and B were from the same source?" 4 answer buttons are: not/slightly/quite/very surprising.

Screenshot of last page of a completed study with a summary of the participants responses and how they aligned with expectations.

0132

Xan Gregg @xangregg.bsky.social · 23/03/2026

I've been vibe coding a #dataviz study aimed at detecting "Is this anything?" between two distribution views. I could keep tinkering forever, but it's ready for test feedback. Try it out (10 min) and send me your comments about the study or about the app. xangregg.github.io/pairstudy121...

Four pairs of charts showing distributions as a teaser for the study content. Two box plots, two bands plot, two dot plots and two violin plots.

Example page from study showing two box plots with overlaid dots with the question: How much evidence do these charts provide that A and B are genuinely different?
Four choices: No/Weak/Moderate/Strong Evidence

340

Reposted by Xan Gregg

Niklas Elmqvist @nelmqvist.bsky.social · 16/03/2026

New post: "When Everyone Is Super" on what coding agents mean for CS research. Everyone got the same superpower. The hard parts of research (questions, methods, rigor) just became relatively more important. The easy part (code) just became relatively less so. medium.com/@niklaselmqv...

medium.com

When Everyone Is Super

What does coding assistants mean for computer science research?

063

Xan Gregg @xangregg.bsky.social · 08/03/2026

#dataviz exercise: trying a smoothed dot plot version of the beeswarm.

Re-creation of jittered dot plot from quoted post. Swiss election results by rurality and predominant language of each community.

1150

Xan Gregg @xangregg.bsky.social · 08/03/2026

Mosaic (marimekko) chart of all the responses to "Are you a large language model?", severely cleaned, by source of participants. #dataviz

Mosaic (marimekko) chart with 4 sources on the x axis (Cloud Research, Prolific, Qualified MTurk, and MTurk) and four responses on the y axis to the question "Are you a large language model?": Yes, <LLM response>, English, No. The MTurk sources have many more non-No responses.

010