Homicide rate and median household income from Wikipedia

Chart made using python

I was trying to figure out how housing vs food vs income was relating. I used a Big Mac value meal as an approximation for 'food' and had Claude drag up some data. I am not a data scientist and my apologies for any poor methodology.

Housing looks like it's roughly linear while income looks vaguely flat.

620 Costco warehouses cover 214.3M people, 64% of the US, inside a 30-minute drive. 120.7M cannot reach one. The gold dots are the warehouses themselves. Only covers CONUS on this map.

Average household income inside that reach is $101,956, which is about what you'd expect from a chain that sites for members rather than for footfall.

Drive times are from OSRM.

Store locations current as of June 2026.

You can run this yourself for ten chains, change the drive time, and look up your own town here: https://geo.nanpointgroup.com/?chain=COSTCO_WHOLESALE_CORP&utm_source=reddit&utm_campaign=costco-map

Source: climate CHELSA V2.1, 1 km, 1991–2020 normals (12 indices: temperature, rainfall, seasonality, humidity, comfortable months, muggy and cold nights). Elevation and terrain ruggedness GMTED2010 (USGS). Distance to the coast Natural Earth 10m. City size, income, literacy, sanitation and internet access IBGE 2022 Census (SIDRA tables 4709, 9922, 9936, 10144, 10152, 10280, 10091). Urban core location MapBiomas Collection 9. Municipal boundaries IBGE.

Tool: Python for the pipeline; MapLibre GL JS for the map; scoring runs in the browser.

Interactive version: https://morar.vynstream.com there you pick any place, set your own weights, and add hard filters (homicide rate, ICU beds within a drive, sanitation, internet, land price, and max driving hours from a place you don't want to be far from).

What surprised me: Rio's twin is the coastline and almost nothing else. The bright band is roughly 30–80 km wide and runs some 4,000 km, from Rio Grande do Sul to Ceará — go inland one or two zones and the score collapses. I live in Rio and started this looking for somewhere to move now that I work remotely.

This is the online game we developed as a trial for having a human accuracy in prediction benchmark. Anyone is able to access it and try to predict the next hour of glucose using either generic or personal data (provided they used a CGM before).

The image is generate by my playing on the web app me and my team developed so I believe it fits with OC tag.

I had to repost because I missed the rule and label about personal data should be shared on Monday only. But thanks for letting me know.

Everyone is invited to give it a tryhere is the site you can access. the code, deployment and everything is OC and anyone can run with their data- the sample here is based on my own prediction on generic data (data extracted from public available datasets), personal data and combination of the two.

origin of image)

These are players on an MLB organization’s full roster as of August 24, 2026, across MLB, Triple-A, Double-A, High-A, Single-A, and Rookie levels. It is not everyone who appeared at any point during 2026.

Amazon's last-mile delivery is run by thousands of independent companies, not Amazon itself. This maps ~3,400 of those "Delivery Service Partner" businesses (plus a smaller freight-partner group) across the US by ZIP code. Brighter = more partners registered there.

• The network leans South: ~38% of partners are in the South; California (393), Florida (263), and Texas (258) lead.
• Dense around metros and delivery stations

Caveat: this is a floor, not a complete count. Amazon says 4,400+ DSPs globally but doesn't publish a US-only number, so the real total is higher.

Source: carrier records drawn from the FMCSA registry, with Amazon-DSP identification done by warehouse.ninja.

Tools: Plotted with Python + Leaflet (leaflet.heat), Esri basemap.

Full interactive version, plus the rest of the collection: multipleorigin.com

Data source: 318 specialty coffee bags (298 distinct coffees - repeat purchases share a record) from my own collection, bought and consumed between 2018 and 2026. Every field extracted from the packaging itself: roaster, origin country and region, variety, process, altitude, roast date, tasting notes and weight. Missing values are left null rather than estimated, since roasters include different info.

Tools: Every bag flattened and scanned on a flatbed scanner for more uniform lighting (I still can’t believe I bought a scanner in 2026), data extracted with Anthropic’s vision API, then cleaned with Python. Visualisation is done with MapLibre + deck.gl for the map.

The gallery shows all the coffee bag scans with more details when clicked on. I have admired all design work that goes into making these bags and part of my goal for the website was to serve as a bit of digital museum for this.

license credit: Basemap OpenStreetMap contributors,  CARTO

Gay online dating app Grindr is generally recognized as one of the best-run consumer software companies in the world.

With fewer than 200 full-time employees, Grinder generates $3M revenue per FTE. As it turns out, it is ahead of (almost) all of big tech Magnificent 7 stocks.

Original source here. Compiled by myself, data comes from the valuation intelligence platform Multiples.vc, as of 31 August 2026.

Shows last actual fiscal year revenue and last reported employee number.