wcedmisten

I don't get how people can listen to and share AI generated "music" and not feel legitimately embarrassed to their core. Like the AI Charlie Kirk song, but unironically. I'm truly baffled how anyone can enjoy it.

> If I was very talented and had the time, I could have written the lyrics and recorded the song myself (which undoubtedly would be even more awesome), but I can’t do this.

I promise you that any art you produced would be better than slop. There is not a single crayon scribble or off-tune melody that would be worse than the generic garbage that comes out of an AI model

For what it's worth, I use OSM for my modest project SurpriseDateSpot[1] which will randomly pick a restaurant (or other amenity) in your area! It gets a few thousand users per month :)

I've also used this project to upstream 2,000 suggested edits to OSM

[1]: https://surprisedatespot.com/

> Software engineering is a mostly solved problem at this point

I guess that's why Claude Code has 0 open issues on Github. Since software engineering is solved, their autonomous agents can easily fix their own software much better and faster than human devs. They can just add "make no mistakes" to their prompt and the model can solve any problem!

Oh wait, they have 5,000+ open issues on Github[1]. I'm yet to be convinced that this is a solved problem

[1]: https://github.com/anthropics/claude-code/issues

Having spent 3 years of my career working with Clojure, I think it actually gives you even more rope to shoot yourself with than Python/JS.

E.g. macros exist in Clojure but not Python/JS, and I've definitely been plenty stumped by seeing them in the codebase. They tend to be used in very "clever" patterns.

On the other hand, I'm a bit surprised Claude can tackle a complex Clojure codebase. It's been a while since I attempted using an LLM for Clojure, but at the time it failed completely (I think because there is relatively little training data compared to other mainstream languages). I'll have to check that out myself

Currently I don't check the opening_hours, because I've found it's missing for many (most?) restaurants, so I felt it wouldn't be a very useful filter

Sure, my wiki username is the same as HN: https://commons.wikimedia.org/w/index.php?title=Special:List...

Thanks! I signed up to the newsletter

That's awesome! I recently picked up bird photography as a hobby and have contributed a few pictures to wiki commons.

Do you have a website I can follow?

I got tired of trying to pick a date spot with my girlfriend, so I made this website to randomly pick an restaurant/date activity for me based on OpenStreetMap data.

I've also used the data corrections submitted by users to contribute over 3,000 edits back into OSM!

https://surprisedatespot.com/

Have you heard of:

https://www.openhistoricalmap.org

I don't think you can pay for it, but I'm sure they'd accept donations!

Overture is an awesome resource because it has more coverage than OSM, but at the cost of accuracy. When I was considering using it for restaurants in surprisedatespot.com, I found that it has a lot more automated content with bad geocoding (imprecise locations) compared to OSM.

E.g. restaurants can come from an old Facebook page and the geocoding might not be accurate compared to a survey.

So where OSM might have a restaurant that closed 8 years ago, overture has every restaurant that's been at that location for the past 8 years.

Now I'm curious, what causes the latency for range requests with R2?

Highly recommend this performance from the last Strange Loop on a similar topic! I had the pleasure of seeing it live. A great combination of live music and Markov chains.

https://m.youtube.com/watch?v=M2o4f_2L0No

There have been quite a few articles corroborating the existence of books like this, the only part that needs believing is that someone actually used the books as advertised.

https://www.vox.com/24141648/ai-ebook-grift-mushroom-foragin...

https://fortune.com/2023/09/03/ai-written-mushroom-hunting-g...

Maybe try the National Address Database?

https://www.transportation.gov/gis/national-address-database...

I really enjoy using OSM for my projects! It can be a bit daunting to figure out how to use the data because there are so many different tools in the ecosystem. I wrote this article recently to help anyone looking to get started querying OSM:

https://wcedmisten.fyi/post/how-to-query-osm/

osmnx looks like a cool library, I haven't seen that one before!

As for self hosting overpass, I have considered it, but the installation steps seemed a bit tedious to me compared to a lot of tools which are dockerized.

I'm a big fan! I use it for most of my OSM projects. It's pretty easy to download an OSM extract and process it into base tiles.

Protomaps also provides a free set of base tiles that you can download and rehost yourself if you don't want to process the data locally.

Author here! Happy to answer any questions you all have about this article.

Hi everyone, author here!

I love making projects with OSM, but it can be challenging to figure out how to analyze the data. So I wrote the article I wish existed to help make it a little easier!

Wow I really like the topographic visualization of the data, looks really slick!

Would a restaurant have the same case against Doordash? Seems like a fairly common business model

I've built quite a few open source mapping projects myself using Tippecanoe and other tools. Feel free to email me if you want help setting this up!

Those are the most popular sources, and I've used both!

The first one is the official OpenStreetMap data, which contains the "planet file" - i.e. all the data for the entire world. But because OSM has so much stuff in it, the planet file is a whopping 76 GB, which can take a long time to process for most tasks. I also recommend using the torrent file for faster download speeds.

As a result of the planet's size, the German company Geofabrik provides unofficial "extracts" of the data, which are filtered down to a specific region. E.g. all the data in a particular continent, country, or U.S. state. If you click on the "Sub Region" link it will show countries, and if you click on those it will show states.

The reason I call it high performance is that it avoids the hours/days of processing (for the planet file)[1] that would be required for pulling the data out of PBF and indexing it. And you'd also need RAM at least the size of the planet to even get that level of speed.

You could certainly amortize this cost for repeated queries, but for one-off queries I haven't seen anything faster.

[1]: https://wiki.openstreetmap.org/wiki/Osm2pgsql/benchmarks

I recently discovered DuckDB's Read_OSM() function [1], which lets you query OSM PBF files directly.

For example, it's simple to count the cafes in North America in under 30s:

  SELECT COUNT(*) FROM st_readOSM('/home/wcedmisten/Downloads/north-america-latest.osm.pbf') WHERE tags['amenity'] = ['cafe'];
  ┌──────────────┐
  │ count_star() │
  │    int64     │
  ├──────────────┤
  │        57150 │
  └──────────────┘
  Run Time (s): real 24.643 user 379.067204 sys 3.696217

Unfortunately, I discovered there are still some bugs [2] that need to be ironed out, but it seems very promising for doing high performance queries with minimal effort.

[1]: https://duckdb.org/docs/extensions/spatial.html#st_readosm--...

[2]: https://github.com/duckdb/duckdb_spatial/issues/349