If your experience was actually worse than my astonishingly janky mobile experience, that would be pathetic. Pretty bad. It looked cool for a second, and then when I tried to zoom in to where I live, I was instantly too irritated to bother.
Yes, and I insisted quite a bit because I was curious about a coast. I wonder if this is some kind of UI Neuralese, maybe the simulated scroll/touch attempts made sense to only it.
I’d guess it was a) an amalgam of a few good ideas that are pretty stupid together, or b) some hand-rolled techniques that got clobbered by a library imported for something else.
I have a hunch that language models have a hard time controlling their voice depending on context. Like there are very particular voices used in certain locations in a user interface, and a model doesn't necessarily have an intuitive sense of that. You can instruct it to use particular voices, but it doesn't usually know to apply that intuitively, you have to know for it and tell it. To me, that example you quoted sounds like a potentially desirable quality (for the model trainers...) for communicating with the prompter during coding tasks, but I would not ever in a million years transfer that verbatim to a user interface. The way you speak to users is fundamentally different than the way you speak efficiently to prompters.
The reason why I consider it may be potentially desirable for communicating with prompters is that prompters usually need a way to verify the model has done what they asked, but without necessarily needing to review all the code or very long runs of text. They only need a bare minimum to know that their requirement has been met and not a full explanation of everything, so the super terse and efficient way of communicating how the constraints are implemented can be helpful. I understand that a lot of prompters don't need this style, or that some people just hate it unconditionally and need it to be different, that's just my guess for why it might've been reinforced during training.
Experienced prompters also typically know what they're doing, so they don't need the full explanation, only the bare minimum details that apply to this particular case. Again, users are different here, but the model doesn't know the difference in order to write it, so that's another way this can end up happening.
Because the previous iteration said "Rain and _____ as a surface; Height is millimetres of rain per day and encodes _____" and the prompt told it not to encode that into the rain layer... probably
I checked my hometown “lighthouses” which technically exists and technically has a light but only exists as a tourist trap monument built in the 1980s. Well, it’s on there.
I like the idea. But it was a bit jarring that panning with my finger took ages and then pinching to zoom insta-faceplanted me into the Labrador interior.
If anyone wants to mount a rescue, I’m somewhere near a lighthouse.
Funny enough I went into this site wondering if inland lighthouses were a thing; there are a few obvious cases, like the Netherlands using them to demarcate shoals and for navigation of inland waterways, but then there's whatever Germany's doing...
Lovely idea, and a nostalgia trip for me. The sweeping looms brought back some old memories. I worked on merchant ships prior to getting into coding and was the officer of the watch one night crossing the Bay of Biscay going north. It was a clear night with that almost unnaturally still air you get at sea sometimes. I gradually noticed the loom of a lighthouse flashing over the sky just off the starboard bow. I wasn’t expecting landfall so soon and quickly double-checked our position on the chart. I counted the flashes; it was clearly the Phare du Créac'h on the Isle of Ushant. It’s stated as visible at 32 nautical miles, but I was seeing it at almost 90 nautical miles away. The light itself was still below the horizon, but the loom was clearly identifiable (there were no other lights in that area with the same number of flashes and frequency). It must have been some weird temperature inversion effect, but I have never forgotten it. I never saw any other lighthouse from that far away in the rest of my 25-odd years at sea.
Looks nice as already mentioned: pinch to zoom is broken (Firefox on Android here). Also: why am I shown a button labeled 'lighthouses in France' when I'm not actually looking at France?
This seems like it maps way more than just lighthouses. Looking at puget sound, at least, it seems to mark all the navigation buoys as lighthouses as well.
Sounding like desktop/laptop with a mouse is the only way to interact with this site. It's a great first try though. I spent a few minutes poking around for sure.
Here is the Icelandic list of lights (isl. Vitaskrá). It includes every light in Iceland. Sailors are supposed to know how to read this and identify a light from their patterns. It is very rear that sailors sail by lighthouses these days though, but an extremely handy skill to have in case your GPS fails for some reason.
I checked the area near my house and there are about 10x more lighthouses shown than really exist. Some are buoys but most don’t correspond to anything.
AI generated with all the attendant quality problems - worst of all the data. There are a lot of non-lighthouses listed on the map, and I can easily come up with a couple of real lighthouses that aren't present.
I really hate all these vibecoded map projects, because the (debatably) slick looking UI implies that the real work of getting the data has been done. But it hasn't, whoever made this just made a bare minimum effort to slap in some plausible looking (if you don't look too close) data and then walked away. The fact that a lot of it is wrong doesn't matter if it looks good, right?
Although this one look a lot better. cool! like how the beams look like they're ray traced.
I feel like this is a cool application of using ai - before something like this would be unfeasible to make due to the amount of lighthouses. Parsing through the amount of data to create a description of every lighthouse would take months+.
Fun fact, there is a .gov effort to sell (or give away) a lot of old lighthouses in exchange for agreed maintenance. It's sort of a terrible deal, since you can't live in them or use them for commercial purposes, but you do still have to maintain it to a certain standard. But I'm sure someone is out there with the desire.
probably just to keep history alive. A lot of (very) wealthy people buy non-sense because they don't know what to do from their money. Then, why not buying a lighthouse :-)
This map has very annoying and weird spelling errors, to the point where I suspect this is just AI generated.
Zooming in to the lights I am intimately familiar with in Suðurnes Iceland (includes the objectively best lighthouse in the World Reykjanesviti) I see “Garðskagaviti” correctly spelled, but the main lighthouse is spelled “Gardjskagi” that letter ð is also misspelled as dh in nearby Sandgerdhi.
My favourite bit about people who use OpenStreetMap data for lighthouses is that they end up including <https://www.google.com/maps/@52.2647013,-2.1151705,51a,15y,2...> lighthouse in the exceptionally landlocked Droitwich (shows up as Hanbury Wharf Lighthouse)
They also often include "pairs" of lighthouses (high / low, front / rear) where newer ones have replaced older ones, but the older structure is still there (and often is now a museum), but isn't the active lighthouse.
Although this one has done a reasonable job at culling most of those out, at least on the south coast of England.
Lovely idea. As a sailor, I still rely on lighthouses and their individual signature. In times of GPS jamming, we shouldn't give up longstanding technology.
It's cool data and I wish the person who prompted Claude to build this would have at least made an attempt to make it optimized and readable. Like, what does "Fl • every 5 s" mean? It's frustrating because this could have been amazing!
It jumped around when I tried to use it on a mobile phone. I was surprised how few Greenland has.
I'm fairly sure that there are more in outlying islands of New Zealand. The Chatham Islands seem well marked but there are none marked on Stewart Island or the Snares.
My trusty M1 Max went full bonkers immediately. GPU temps spiked at 101°C and both CPU and CPU showed 100% workload. Also probably the first time this year the fans went on.
Another lighthouse map [1] was discussed on HN in 2022 [2]. It also has some data-quality/completeness issues due to being based on OpenStreetMap data, but avoids the AI slop aesthetic of this one.
Sometimes they change the font, but the UI also jumps out as a second clue. There's always too much info density.
But I am not against it if done properly. Here I had my local AI go at the UCB Logo docs to turn it into a more pedagogically-useful tut: jansavagelogo.surge.sh
The site is far from complete, and now I have to read all of it, to make sure nothing is wrong.
They are not in a busy part of the ocean, but I suppose it is to warn people away. That archipelago is not fully charted, so it is probably a good idea to keep away from much of its coast. Probably dozens of uncharted reefs.
Uncharted archipelagoes in 2026? Couldnt satellite imaging help with this? And I don't think it's near any shipping lanes theyre incredibly remote, theyre more likely there so the research vessels that enter the Port-aux-francais harbor can see the shore
"Rain as a surface; Height is millimetres of rain per day. Nothing else is encoded."
Why does AI generated ridiculous sentences like these?