Lightnews — Scholar-powered news

LightNews

juweek

@juweek.bsky.social

100 followers 81 following 83 posts

I use code to make diagrams #dataviz #data #creativecode #p5js juweek.studio

Posts Media Videos Starter Packs

juweek @juweek.bsky.social · Feb 21

creating a tool that lets you customize charts quickly

been helpful for me personally, but that's cuz i love a well-designed chart. hoping to share it more widely soon

juweek @juweek.bsky.social · Feb 10

hand drawn charts are always so fun

juweek @juweek.bsky.social · Feb 10

what a great cover

juweek @juweek.bsky.social · Feb 10

building out the foundation of a youtube video analyzer, trying to do some basic NLP (it's mostly useless)

happy with the basics and gonna iterate on the design, but i must admit this project is a great speedrun on many things backend-related

juweek @juweek.bsky.social · Feb 9

now that code is easy to create, some muscles that seem important:

- understanding what people need
- fun interaction design
- backend scaffolding
- speed

juweek @juweek.bsky.social · Feb 8

one thing that's underrated is how fun it is to use your own product, and how annoying it is when it doesn't pass the vibe check

juweek @juweek.bsky.social · Feb 7

my current side project is a food scanner, and thinking of your market advantage is an interesting thought exercise. my current approach is focusing on the visuals

juweek @juweek.bsky.social · Feb 6

here’s what the new dataset looked like, and how the model predicted if they were ‘preferred’ or not (1 is preferred). the results were pretty solid, with a protein-fat threshold somewhere around 5:1

i think ML can be confusing, but understanding what's going on makes it less so

juweek @juweek.bsky.social · Feb 6

similarly, i have a section dedicated to one of python's default classification models. i wrote this back when ML was new to me, but it's fun seeing myself cover fundamentals

I selected a regression model, defined a training set, and trained a script for binary classification

juweek @juweek.bsky.social · Feb 6

the next step was preprocessing the data. I go into more detail in my guide, but here’s a screenshot of what I did. Basically, I made the data as easy to process as possible—standardizing, scaling, and encoding each food item into different categories. data science basics

juweek @juweek.bsky.social · Feb 6

here was what the distribution of the foods looked like. i believe it was per 100g serving

juweek @juweek.bsky.social · Feb 6

the goal was to create a model that could classify foods based on if they met my protein-fat ratio requirements. i got the data from USDA FoodData nutritional database

here’s how the raw data looked . There are fields for Descrip (the actual food item) and a bunch of nutrients

juweek @juweek.bsky.social · Feb 6

in the spirit of education, here's a quick thread on how i built a basic nutrition classifier in python

it helped me identify foods with a high enough protein-to-fat ratio for my diet.

juweek @juweek.bsky.social · Feb 5

my current jam is gamifying experiences, and I've been thinking up quick ideas to code

here's a currency converter for my visual folks out there

juweek @juweek.bsky.social · Feb 3

ever get confused about the different crypto coins? so do I, so I made this chart that compares the market caps and daily transactions of a few ecosystems. this was made with Feb 1 data

it's fun to see how much this looks like a solar system, and how each ecosystem has their own behaviors

juweek @juweek.bsky.social · Jan 31

something like 10 - 15% of earth is used to feed humans right now, and most of that is for livestock. heres a graph

funny thing is if u think about it, there's an inverse relationship between animal welfare and enviro impacts, eg free range/cage free options require more land

god's funny

Sankey diagram titled 'How the World Uses Its Land,' illustrating the distribution of Earth's surface and the allocation of habitable land. The chart shows that 71% of the Earth's surface is covered by the ocean, while 29% is land. Of this land, 22% is classified as habitable, with the remainder being glaciers (3%) and barren land (4%). The habitable land is further divided into different land uses: about half (48 million km²) is used for agriculture, including grazing land (~7%), cropland for food (~2%), cropland for animal feed (~1%), and non-food/biofuel crops (0.4%). Other land uses include forests (~8%), shrublands (~3%), lakes and rivers (~1%), and urban areas (0.2%). The chart highlights that a significant portion of agricultural land is dedicated to livestock rather than direct human food production.

juweek @juweek.bsky.social · Jan 30

visualizing diff social media sites; size of circle equals revenue, number of particles equals MAU (these are estimates from ~2024)

might write on this but my tldr thoughts
- tiktok is growing crazy rn and has hella space to grow revenue
- meta's market reach is underrated

juweek @juweek.bsky.social · Jan 28

i'm mixed on opinions of efficiency, but some trends are easy to see in this map. eg, north west Iran probably has a lot of population centers

juweek @juweek.bsky.social · Jan 28

looking at some maps with Iran, and at least with this study, there is a preference for diving information by geography. kind of like a choropleth, but not quite... this is a map of rural houses (basically museums) by province

juweek @juweek.bsky.social · Jan 25

the solution to avoiding synthetic fertilizers is to:

- eat organic
- reverse osmosis water filter
- buy from farms practicing sustainable farming
- large government action

some are more feasible than others

juweek @juweek.bsky.social · Jan 25

fertilizers have their place in growing societies, but it's worth noting that Europe dropped their usage in the 90s after realizing the environmental consequences. The US kept going, and our usage has remained relatively stable

"Stacked area chart showing the increase in synthetic fertilizer use in the U.S. since 1960, measured in millions of tons. The chart highlights three types of fertilizers: Nitrogen (N) as the largest contributor, followed by Potash (K2O) and Phosphate (P2O5). Fertilizer use peaked around 1980, fluctuated slightly, and stabilized at high levels post-2000. Source: FAO Fertilizers by Nutrient database, July 17, 2024."

juweek @juweek.bsky.social · Jan 25

synthetic fertilizers help crops grow faster, but only ~50% of nitrogen from fertilizers is absorbed by plants. The rest ends up polluting the atmosphere and waterways, and that leads to some of the diseases I mentioned above

"Paragraph explaining nitrogen fertilizer use in agriculture. It notes that plants typically absorb less than half of the applied synthetic nitrogen, with absorption varying by crop type and soil conditions. Excess nitrogen remains in the soil, causing environmental risks such as nitrogen loss."

"Text describing health risks associated with nitrate intake. Higher consumption of water nitrate and meat is linked to increased risks of colon, kidney, and stomach cancers, as well as modest evidence of higher risks for thyroid and ovarian cancers among women."

juweek @juweek.bsky.social · Jan 25

Fertilizer use ballooned mid-20th century during the Green Revolution, when lots of new agricultural tech came about to meet population demands. pesticides and fertilizers were two such techs

Fun fact: some pesticides originated from leftover WWII chemicals

"Excerpt discussing the role of chemical discoveries during World Wars I and II. Highlights include the prevalence of chemical weapons during World War I, their repurposing into insecticides during the interwar years, and the discovery of organochlorines, including DDT, during World War II for use in disease vector control."

juweek @juweek.bsky.social · Jan 25

the US uses more than twice as much fertilizer as the rest of the world

Synthetic fertilizers, especially those with nitrogen, can leach into water bodies, causing nitrate poisoning, which has been link to thyroid diseases and cancers

Why are we still using so much of it?

"Line graph of fertilizer use per capita (1961–2019) in kg/year. The U.S. peaks in the 1980s before declining, Europe follows a similar pattern, China steadily rises, and the world average increases gradually."

juweek @juweek.bsky.social · Jan 20

flow maps are so good; theyre like the map version of a sankey diagram. great way to build visual interest and make reading more participative

an example from 1930s Hungary, showing the supply and distribution of goods along the railway lines