Tessera

Ilya Kashnitsky

@ikashnitsky.phd

Demographer / Senior Researcher @dst.dk / Affiliate Member @oxforddemsci.bsky.social‬ / Board @demografi.dk ✨ @datavizartskill.ikashnitsky.phd 📝 https://ikashnitsky.phd 😍 #demography #dataviz #rstats 🙌 #openscience #bibliometrics 🔗 https://bio.link/ikx

1072 Posts3024 Follower 2210 Folgt

Ilya Kashnitsky @ikashnitsky.phd · 3w ago

Fractional counting in #bibliometrics should become the norm, otherwise traditional metrics-based #ResearchEvaluation becomes meaningless too often ⚖ 👀 the example here 👇 blog post: ikashnitsky.phd/2025/ihme-bibl #srats replication using #OpenAlex data: gist.github.com/ikashnitsky/...

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

Here's an exceptionally interesting #dataviz 👀 🔗 www.visualcapitalist.com/cp/countries... 👤 The author, Iswardi Ishak, produces more of these cool carefully crafted global infographics and shares via @voronoiapp.bsky.social 🤩 🔗 www.visualcapitalist.com/creators/iswardi-ishak

Infographic titled "Education Spending as % of GDP," using a color-coded world map and two ranked tables to compare countries by education spending relative to GDP and the percentage of population aged 24 and under.

## Overall Layout

The image is a data infographic with a navy blue background, a legend of seven color-coded spending brackets (from "less than 2%" in dark gray to "greater than 7%" in dark blue-green), and a world map showing countries filled with these colors and labeled with specific percentages, such as Sweden 7.3%, U.S. 5.4%, Russia 4.2%, China 3.9%, and India 2.8%.

## Top and Bottom Tables

Below the map, two tables titled "The Highest and Lowest Among Top 40 Economies" list "10 Highest" and "10 Lowest" countries by education spending as % of GDP, each paired with a flag icon, a colored income-tier tag (High-Income in blue, Upper-Middle in yellow, Lower-Middle in red), and a column showing "% population aged 24 and below." Sweden tops the high list at 7.3% spending with 28.4% young population, while Japan sits lowest at 3.3% spending with 20.8% young population.

## Text and Sourcing

The top-right corner contains a paragraph explaining that education spending reflects national priorities, demographics, and fiscal capacity, and that higher spending can signal investment in future generations. A small source line at the bottom cites "UNESCO via OurWorldInData, Population Aged 24 and below: UN Population Division, 2025," with a copyright notice "© 2026 Iswardi Ishak."

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

The t-shirt from the 20th anniversary celebration of EDSD finally found its way (thx @cosmostrozza.bsky.social) to me (who missed the event) 🎉

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

🚨Oh boy! Here is a masterpiece worth reading carefully, saving as a bookmark in a closeby drawer, and taking out this weapon to use in comments whenever you see a legacy usage of ollama 🫂Friends Don't Let Fronts l Friends Use Ollama 🔗 sleepingrobots.com/dreams/stop-... #foss #llms #ollama #llamacpp

April 18, 2026

+

Friends Don't Let Friends Use Ollama

Ollama gained traction by being the first easy llama.cpp wrapper, then spent years dodging attribution, misleading users, and pivoting to cloud, all while riding VC money earned on someone else's engine. Here's the full history, and why the alternatives are better.

Tags: LLMs, local, open-source, opinion

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

An early morning run interrupted with a rescue mission. The martin bird is safe with the rescue guy who arrived on a huge evacuator. The 1812 service (Animal Protection Denmark; www.dyrenesbeskyttelse.dk) works fast — arrived in half an hour after the call 🙏

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

This is and impressive and alarming figure. And I'm sure Nature journals, with their normalization of absurd APCs, are to blame here #OpenScience #AcademicSky #ScientificPublishing

infoDOCKET@infodocket.bsky.social · 2mo ago

New From @deltathink.bsky.social: Beyond #APC Tracking – Helping #Libraries Measure Publishing Cost and Impact www.deltathink.com/news-views-b... #scholcomm #publishing

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

Today was a happy milestone — the first international guest coming to DST to give a talk at @skagerrak.bsky.social, and not just any guest but @jschoeley.com himself with a brilliant thought-provoking demographic talk and presentation of his brand new demoscapes.org 😍

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

🌎 Results of all 22 World Cup Finals — at a glance ⚽ #FIFAWorldCup #wc26 #wc2026 #worldcup26

Illustration chart of FIFA World Cup finals showing goals scored in each final from 1930 to 2022. A horizontal timeline runs along the bottom with years labeled (1930, 1934, 1938, 1950, 1954, 1958, 1962, 1966, 1970, 1974, 1978, 1982, 1986, 1990, 1994, 1998, 2002, 2006, 2010, 2014, 2018, 2022). Vertical lines rise from each year to a height indicating total goals, with circular markers showing the two finalist nations using their flags. Background is light beige with faint grid lines. Title at top reads “All 22 FIFA World Cup Finals.” Left side text reads “Goals scored in the final.” Colors are muted pastels; flags provide small bursts of color. Overall layout is wide and evenly spaced.

Ilya Kashnitsky @ikashnitsky.phd · 2mo ago

10 years (😱) since my first proper academic paper. That was a tough yet very educational journey; and it's so much easier to deal with the complex and frustrating academic publishing process when the paper in question is a side-kick project, not *the PhD idea* 🙃 🔗 doi.org/10.1016/j.ci...

Skagerrak Demographic Series @skagerrak.bsky.social · 3mo ago

On 15th June at 11:00 @jschoeley.com will present his shining new project that he was secretly cooking with the relentless help of silicone friends 😍 🔗 events.teams.microsoft.com/event/839793...

Ilya Kashnitsky @ikashnitsky.phd · 3mo ago

📦 {sjrdata} #rstats package updated to include 2025 journal rankings 🔗 github.com/ikashnitsky/...

Column plot of the total number of journal included in Scopus database, 1999-2025. There is a steady growth in the global academic output, yet we see a sharp drop in 2021, which is a result of fraud purging

Xan Gregg @xangregg.bsky.social · 3mo ago

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.

Ilya Kashnitsky @ikashnitsky.phd · 3mo ago

TODAY in less than an hour Laust Hvas Mortensen is delivering a Skagerrak Demographic Series talk @skagerrak.bsky.social @demografi.dk Title: Broken limits to social science? Foundation models for demography. Please register for the talk here: events.teams.microsoft.com/event/8d8c1a...

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

😱 AI will soon take our jobs. Surely you are hearing this claim often enough to dismiss it as fearmongering. The key word is "our". Will it soon grab *my* job? What's really happening in job displacement today? 👇The single most useful AI job revolution explainer sequoiacap.com/article/serv...

.

Diagram divided into four quadrants by two axes — a horizontal axis labeled "JUDGEMENT" on the left and "INTELLIGENCE" on the right, and a vertical axis labeled "OUTSOURCED" at the top and "INSOURCED" at the bottom. Each quadrant lists professional service sectors with estimated market sizes in US dollars.

**Top-left quadrant, labeled "COPILOT TERRITORY"**, contains open-circle bullet points indicating lower automation readiness: Management consulting $300B+, Graphic / UX design $30B+, Executive search $20B+, PR & comms $20B+.

**Top-right quadrant, labeled "AUTOPILOT TERRITORY"**, contains filled-circle bullet points indicating higher automation readiness: Insurance brokerage $140–200B, IT managed services $100B+, Payroll & compliance $50–70B, Claims adjusting $50–80B, Accounting & audit $50–80B, Healthcare rev cycle $50–80B, Mortgage origination $30–50B, KYC/AML $30–50B, Paralegal / LPO $36B, Tax advisory $30–35B, Legal transactional $20–25B, Real estate closing $20–25B, Cost estimation $16B.

**Bottom-left quadrant, labeled "WATCH"**, contains open-circle bullet points: Recruitment $200B+, Advertising $100B+, Freight brokerage $100B+, Admin assistants $80B+, Clinical trials / CRO $80B+, SEO / SEM $50B+, ERP implementation $50B+, Corporate training $50B+, Market research $45B, Cybersecurity $30B+, Architecture $25B+, Patent / IP $15–20B, Travel mgmt $15B+.

**Bottom-right quadrant, labeled "NEXT WAVE"**, contains filled-circle bullet points: Supply chain & procurement $200B+, Pharmacy back-office $30B+, Wealth mgmt ops $30B+, Medical admin $20B+, Fund administration $15–20B.

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

Jonas is yet again producing insanely cool interactive demographic #dataviz (jschoeley.shinyapps.io/hmdexp is still the best #shiny app I know of) 😍 Explore demoscapes.org and watch this space, stay tuned for more cool updates 🐟

Jonas Schöley@jschoeley.com · 4mo ago

What is the main reason for premature male mortality in Mexico? @ikashnitsky.phd and Aburto answered this question in 2019 for every Mexican state, every year since 1990 and all ages from 15 to 50. Their rich results are available on demoscapes. demoscapes.org/collections/...

Jonas Schöley @jschoeley.com · 4mo ago

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

😆 After the last Hard Fork episode (@kevinroose.com & @caseynewton.bsky.social) I couldn't resist trying out Talkie-1930, a chat based on an LLM with information cut-off in 1930. Just imagine you are a time-traveler 🤯 🔗 talkie-lm.com/chat

User
can human survive on a motorbike at 250 km/h?
talkie-1930
No; it is impossible to survive at such a speed for any length of time. The human organism cannot keep pace with a speed greater than 150 km/h, and to maintain even this pace for any considerable time would severely tax the heart and lungs. At 250 km/h the air resistance opposing the forward motion of the rider would be equal to about 44 lb, and the power necessary to overcome this would be 4½ H.P. The task of providing this power would soon become prohibitive, and the rider would be compelled to walk. On the other hand, if the speed were reduced to 100 km/h the air resistance would be only 22 lb, and the power necessary to overcome it 2½ H.P. In other words, a man can ride a motorbike at 100 km/h with comparative comfort, but cannot survive at a greater speed.

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

Damn, do I ever learn that simple fact that (at least in coding #rstats) everything's taking more time than I initially think 🙈😅🙃

Peter Dalgaard @pdalgd.bsky.social · 4mo ago

R version 4.6.0 "Because it was There" (source version) has been released. It should be on CRAN by now. #rstats The choice of codename is in remembrance of R Core member Tomàš Kalibera (1978-2026). Among many other things, he was a keen mountain climber.

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

This is quite unbelievable 😱 China has been unrolling 10k km of high-speed railway every 4 years since 2008 🚀

Yasar Sharfudeen@newbie2k25.bsky.social · 4mo ago

Day 24 #30DayChartChallenge — South China Morning Post Theme Day China’s high-speed rail network expanded from its first 350 km/h line in 2008 to 50,000 km by 2025. Source: China State Railway Group / official government data Tools: R + ggplot2 #RStats #China #HighSpeedRail

A line chart titled “China’s High-Speed Rail Boom” showing operating length increasing rapidly from 2008 to 2025, with milestones at 10,000 km in 2013, 20,000 km in 2016, 40,000 km in 2021, and 50,000 km in 2025.

troy_phd @troyhernandez.bsky.social · 4mo ago

After I messaged my agent from the train this morning to push a forgotten commit, I figured it was time to submit to CRAN and publish more than a Bluesky post. R is great for building AI agents: #rstats #CodingAgent cornball.ai/posts/r-as-a...

R as CLI Agent Harness

One line: 1corteza::chat() That starts an AI agent in your R console. It can read files, run shell commands, query git, search the web (with a free Tavily API key). And because it’s running in your R ...

cornball.ai

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

💡 Changes life expectancy follow changes in total death counts @araksha.bsky.social ✨ 📝 doi.org/10.31219/osf... 🔗 #rstats code: github.com/ikashnitsky/... 🧙‍♂️ no ai jumpstarter this time, I worked off @jschoeley.com's code, all here github.com/ikashnitsky/... DAY 16 -- causation 💫 #30DayChartChallenge

Scatter plot with a fitted regression line illustrating the near-perfect negative linear relationship between year-on-year relative changes in life expectancy and year-on-year relative changes in total death counts for Italian males, covering the period 1950 to 2023.

The horizontal axis is labeled "Δ log D" (change in log total deaths), ranging from approximately −0.15 to beyond 0.15. The vertical axis is labeled "Δ log e₀" (change in log life expectancy at birth), ranging from below 0.00 to above 0.02. Reference lines cross at the origin (0.0, 0.00).

Dark teal circular data points are tightly clustered along a descending pink regression line, confirming a strong inverse correlation: years with rising death counts correspond to falling life expectancy, and vice versa. The majority of points fall near the origin, reflecting small year-on-year fluctuations. Two notable outliers appear in the upper-left quadrant (large life expectancy gain, large death count drop). One prominent outlier is labeled "2020" in pink, located far to the lower right, representing a large spike in deaths and a sharp drop in life expectancy during the COVID-19 pandemic. A second unlabeled point sits near it, slightly above and to the left.

The title reads: "A near-perfect linear link between changes in life expectancy and total death counts." The subtitle reads: "Year-on-year relative changes in life expectancy VS year-on-year relative changes in total death counts, Italy, males, 1950–2023." The data source is credited to UN World Population Prospects 2024 (wpp2024 R package), with attribution to Ilya Kashnitsky @ikashnitsky.phd, part of the #30DayChartChallenge 2026, Day 16, theme: causation.

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

🇩🇰 Yesterday we had a wonderful yearly meeting of @demografi.dk — Danish Demography Day with one session of brilliant speakers and plenty of coffee break time to socialize 🫂 (👇 SDU is ready for the yearly science fest for kids 😍)

Andrew Heiss @andrew.heiss.phd · 4mo ago

THE MONKEYS SUCCEEDED! For a fun time, run this in #rstats See the full #rstats code here: gist.github.com/andrewheiss/... There are 13 seeds between 1 and 200,000,000 that do this!

$withr::with_seed(37035397, { paste0(LETTERS[sample(26, 5, replace = TRUE)], collapse = "") }) #> [1] "HEISS"$

Andrew Heiss@andrew.heiss.phd · 4mo ago

instead of finishing writing an exam I'm unleashing 200 million virtual monkeys on keyboards to see if they can type my last name in order

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

DAY 12 -- Flowing Data 🌊 #30DayChartChallenge Explorations of the US names are always fun. Here we look at the most popular names by sex and distinguish them by timing of their peak popularity 🗻 🔗 #rstats code: github.com/ikashnitsky/... 🧙‍♂️ pplx chat: www.perplexity.ai/search/day-1...

Streamgraph titled "Name Waves: the Ebb and Flow of American Baby Names," displaying the popularity of the top 10 baby names per sex in the USA from 1950 to 2022, sourced from the US Social Security Administration. Stream width reflects total births. Color encodes peak era: warm tones (yellow, orange, red) for early-peak names and cool tones (pink, purple, blue, teal, green) for recent-peak names.

The chart is split into two sections stacked vertically. The upper section, labeled "Girls" in bold teal, shows a wide stream that peaks broadly around the 1980s–1990s before narrowing toward 2022. Names labeled within the streams include Linda, Mary, Susan, Lisa (warm tones, prominent in the 1950s–1960s), Patricia and Jennifer (mid-era, orange to pink), and Sarah, Ashley, Jessica, Elizabeth (cooler tones, prominent from the 1980s onward). The lower section, labeled "Boys," shows a similarly shaped stream widest in the 1950s–1960s and tapering toward 2022. Names labeled include James, John, Robert, William, David, Michael (warm yellow-green tones, dominant in the 1950s–1970s), and Joseph, Matthew, Christopher, Daniel (cooler teal and purple tones, peaking in the 1980s–1990s). Credit text reads: "Data: US Social Security Administration via {babynames} · #30DayChartChallenge 2026 · Day 12 · FlowingData · Ilya Kashnitsky @ikashnitsky.phd."

Ilya Kashnitsky @ikashnitsky.phd · 4mo ago

For no specific reason, I want to share one of my all time favorite memes 😂

Two-panel cartoon illustration comparing how UI designers and users perceive a baby mobile. The left panel, labelled "UI" in bold black text at the top, shows two smiling cartoon adults — a dark-haired person in yellow and a brown-haired person in blue — leaning happily over a cot railing. A speech bubble reads "I love it!" and another near the cot reads "Me too!" Small cute animal toys — a blue bear, pink pig, and orange cat — hang from the mobile above a contented baby lying face-up in light blue clothing. Musical notes float in the air. The right panel, labelled "Users" in bold black text, shows the baby's point-of-view perspective of the same mobile from below: a striped orange tiger, a blue elephant, a tan rabbit, and a large pink pig are seen upside-down and distorted, spinning chaotically with motion lines and musical notes scattered around. The watermark "VKCOM/PITERSKIS_PUNK_WALL" appears in small text in the upper right of the right panel.

Ilya Kashnitsky @ikashnitsky.phd · 5mo ago

🎨 {linuxcolors} a small #rstats package with the identity colors of the most popular #Linux distros 🐧 💎 #ggplot2 ready with scale_{color/fill}_linux() functions 🔗: github.com/ikashnitsky/... 📦 DAY 13 -- ecosystems 🌍 #30DayChartChallenge ✨ #FOSS world is a unique human #ecosystem

$Illustration-style painting on a white background featuring a central hexagonal frame surrounded by colorful paint splashes and drips, displaying numerous Linux distribution logos arranged across the composition. At the center of the hexagon, a group of five painted penguins — three adults and two smaller ones — represent the Linux mascot (Tux), rendered in a realistic watercolor style in black, white, and yellow. Below the penguins, bold black text reads **LinuxColors**. Surrounding the central frame, the following Linux distribution logos are visible, each rendered in a painterly, dripping style: - **Fedora** — blue circular logo, upper center-left - **Arch Linux** — cyan/teal upward triangle with "tm" mark, upper center - **EndeavourOS** — purple and red triangle with text "ENDEAVOUROS", upper center - **openSUSE** — green chameleon logo with text "openSUSE", upper right - **Void Linux** — dark green circular logo with text "VOID", center-left - **Pop!\_OS** — yellow circular logo with "P!" symbol, lower center-left - **Linux Mint** — green "lm" stylized logo, lower center - **Ubuntu** — orange circular logo with three dots, lower right - **Kali Linux** — grey dragon/kite shape, left side - **Zorin OS** — blue circular "Z" logo, lower left - **Artix Linux** — blue circular "A" logo, bottom left - **Nix / NixOS** — blue snowflake-style logo, upper left Additional unidentified logos appear in dark navy (an "X" triangle, lower center) and teal/green shapes on the right side. Paint splatters in red, yellow, blue, green, and purple are scattered throughout.$

Silvio C. Patricio @scpatricio.bsky.social · 5mo ago

One of my favorite PhD thesis chapters is now published in @pnas.org Using cohort mortality data from 12 countries, I find no evidence that the rate of aging has slowed down. Longevity gains seem more consistent with a later onset of aging. www.pnas.org/doi/10.1073/...

The rhythm of aging: Stability and drift in the individual rate of senescence | PNAS

Human aging is marked by a steady rise in the risk of dying with age–a process demographers call senescence. Over the past century, life expectancy...

pnas.org

Ilya Kashnitsky @ikashnitsky.phd · 5mo ago

DAY 7 -- multiscale #30DayChartChallenge ⚖ Demographic Transition around the world 🌐 👩‍🎓 read more about Demographic Transition on @ourworldindata.org: ourworldindata.org/demographic-... 🔗 #rstats code: github.com/ikashnitsky/... 🧙‍♂️ pplx chat: www.perplexity.ai/search/day-7...

Bivariate choropleth world map titled "Humanity in transition, Demographic Transition," subtitled "Life expectancy (y) × Total fertility rate (x), 2023." Countries are colored using a 3×3 biscale grid encoding two variables simultaneously: life expectancy at birth (y-axis) and total fertility rate (x-axis).

Key regional patterns: sub-Saharan Africa is deep orange (low life expectancy, high fertility, 26.1% of countries); North America, Europe, and Australia are deep purple (high life expectancy, low fertility, 21.4%); central categories of mid life expectancy and mid fertility account for 15.0%. The Middle East and North Africa show mixed orange-brown tones; Russia and Central Asia display medium purple.

Footer reads: "Data: World Population Prospects via {wpp2024} · #30DayChartChallenge 2026 · Day 7 · Multiscale · Ilya Kashnitsky @ikashnitsky.phd." [help.siteimprove](https://help.siteimprove.com/support/solutions/articles/80000863904-accessibility-image-alt-text-best-practices)

Ilya Kashnitsky @ikashnitsky.phd · 5mo ago

DAY 10 -- pop culture 📸 #30DayChartChallenge I'm revisiting #TidyTuesday dataset from 2021 The Billboard Hot 100 to explore how old were the shooting stars of the musical industry at their first peak 🌋 🔗 #rstats code: github.com/ikashnitsky/... 🧙‍♂️ pplx chat: www.perplexity.ai/search/day-1...

Horizontal ridgeline (density) chart combining dot strip plots showing the age at first Billboard Hot 100 #1 hit, by music genre, spanning 1958 to 2021. Five genres are displayed as stacked rows: Country, Hip-Hop, Rock, Pop, and R&B.

Each genre has a filled density curve above a horizontal dot strip, where each dot represents one artist. A central filled circle marks the median age for each genre. A horizontal line extends through each density plot indicating the interquartile range.

**Country** (salmon/terracotta): Density skews right, with a wide spread from roughly age 22 to 45. Median sits near 37. Notable labeled artists: Carrie Underwood (approx. age 22) and Kenny Rogers (approx. age 43).

**Hip-Hop** (golden yellow): Density peaks sharply around ages 22–26, with a right tail to ~32. Median near 26. Labeled artists: Lil Nas X (approx. age 20) and Drake (approx. age 27).

**Rock** (light green): Broad, flat distribution from ~25 to beyond 50, with a long right tail. Median near 28. Labeled artists: The Monkees (approx. age 22) and Santana (approx. age 52).

**Pop** (sky blue): Concentrated bell shape peaking around ages 20–28. Median near 25. Labeled artists: Michael Jackson (approx. age 17) and Lionel Richie (approx. age 33).

**R&B** (muted purple-grey): Moderate spread from roughly 17 to 37. Median near 26. Labeled artists: Chris Brown (approx. age 17) and Earth, Wind & Fire (approx. age 32).

The x-axis runs from age 10 to 50. Text at bottom right reads: "Data: Billboard Hot 100 via TidyTuesday 2021 · Birth years: Wikipedia / For groups, lead/founding member's birth year used / #30DayChartChallenge 2026 · Day 10 · Pop Culture · Ilya Kashnitsky @ikashnitsky.phd".