Predicting how Heathrow is using it's runways in the browser

One coder’s ‘AI’ project meets pilot eye-rolls and sleuths

TLDR: An engineer built a browser tool that uses machine learning and free flight data to show Heathrow’s runway setup. Commenters called it overkill, arguing official airport info (ATIS) or a quick look at flight trackers already tells you, igniting the classic clash: fun DIY build versus use-the-simple-tool

A window seat near Heathrow became a side project: a browser page that uses machine learning to guess which runways are landing and which are taking off, based on plane dots from OpenSky and a grid of the sky. Cute? Yes. Controversial? Absolutely. The comments went full soap opera.

First punch: skeptics said you don’t need “AI” at all. One user bragged they opened Flightradar24 and spotted “Westerly ops in one second.” Another waved the pilot handbook: this info already lives in ATIS (the airport’s weather/operations broadcast) and D‑ATIS (the text version) — “just parse the words,” they snapped. The classic meme dropped fast: when all you’ve got is AI, every problem looks like a model.

Then came the subplot: an armchair sleuth spent 15 minutes triangulating the header image, matching a taxiway nub and — yes — the author’s LinkedIn. CSI: Runway Edition. Meanwhile a moderator popped in to hint the whole thing might be off‑topic, fueling more eye‑rolls.

Still, a quieter crowd appreciated the hack: collecting data on a tiny budget, dodging pricey APIs, and making a one‑glance site for commuters beats fiddling with apps. Verdict from the stands: fun project, spicy premise, and a textbook fight between build it because it’s cool and just use the thing that already works.

Key Points

  • The author built a browser-based tool, “lhrNet,” to display Heathrow’s current runway operation state.
  • The method models a grid over Heathrow, marking cells with aircraft as 1 (present) or 0 (absent), treated as a 1-bit image.
  • A machine learning image-classification approach categorizes the airport into one of seven operational states.
  • Live aircraft data is collected from OpenSky Network’s anonymous tier; labels come from Heathrow Runways posts.
  • Initial JSON storage of 2D arrays became large; a custom binary encoding was considered to compress data.

Hottest takes

“Not easily definable”? Based on what?
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.