The best clear example I've seen of LLMs making money is a company that now generates custom email text instead of using standard email templates. They increased engagement by some meaningful metric like +15% which translates into hundreds of millions of dollars in revenue.

Great example. Do you know what sorts of input they're using to drive this custom messaging?

Not really.

I know the original email was something like "Alert: you have a new thing: X Thing"

And the new emails are a prompt something like "we know all of this about the user and all of this about the X thing, write an email alerting them to the new thing with these particular goals".

I really don't know much about it so I'm being pretty vague and generic.

I know lots of junior developers using tools like Cursor. When talking to them about it, they say they're not really learning how to program and don't know what they're doing most of the time. I do question how effective they actually are.

I recall that before AI, junior developers weren't very productive their first year on the job, but they became at least 10x as productive as they ramped up.

I'm left wondering if the "AI boost" that junior devs are getting now is leaving them less productive than if they had the ramp up that we had in the past. Maybe AI is making them 2-3x as productive but they're staying stuck there. Whereas without AI they might learn more and reach higher productivity.

The experienced devs I know use AI as a collaborative tool on the side. Like asking Claude or ChatGPT targeted questions. That's what I'm doing as well. I know I can code much faster than the junior devs using Cursor that I interact with.

I've used WebAssembly for complex cross platform SDKs. I write the core SDK once in WebAssembly and then wrap it with an API layer for each SDK. It sure beats updating and testing complex logic for 30 different SDKs.

How is MapBox going for this free tool? Is it costing you money?

It definitely will if it blows up more. I'm willing to eat it for now because I think it's art that more people should see!

Is MapLibre GL a cheaper (free?) open source alternative?

Cool stuff btw. I’m trying to visualize weather model data myself (millions of points) at https://futureradar.net and have been researching client-side techniques like yours.

It is very cool art!

I was at Microsoft in 2007. We were told to say "Bing it" but everyone was using Google to look stuff up for work.

Our company at the time forced us to use Bing as the default search engine, so people started searching “google” on Bing to just to get back to Google. We were told “they’re both search engines, just use this one”, just like a person who bought an iPhone from a dollar store.

If you want to be extremely resilient with your infrastructure, you should be multi-cloud anyway. So you can rely on US infra like AWS and also something like Hetzner. Then you're still good if the US stuff goes crazy.

I'll check it out!

Fwiw, after spending a few minutes on your website, I see nothing that indicates your software works on Google Drive or Photos. It just looks like a Backblaze alternative. There might be some room to showcase your offering a bit more.

I am a Backblaze customer so I will definitely consider your software as an alternative.

I haven't seen entrepreneurship generally be looked upon favorably in the corporate world. It's often a red flag that you'll get bored in a niche role or be difficult to manage. At least for technical ICs.

As I understand it, even a low 6-figure exit is a huge plus for you in the entrepreneurial world. It should be a lift for you in the eyes of people like founders and investors.

For context, currently a faang IC.

I am trying to get that "plus" to founder and investors indeed.

Does this not require a massive database of tiles?

I ask because I've been looking to self host some sort of map tile server and they seem to have database in the hundreds of GB.

Good question. Corviont is region-focused (you package one region), not "host the whole planet". Hundreds of GB is usually the full planet at high zoom / lots of layers.

Also it’s not one giant tile DB - there are 3 datasets:

  - map tiles (PMTiles)
  - routing tiles (Valhalla tiles)
  - geocoder index (SQLite)

For Monaco all three are tiny - you can see the exact files here: https://github.com/corviont/monaco-demo/tree/main/data

For small countries like Austria/Slovakia, each is typically hundreds of MB.

The size of the tiles will depend on how much detail you want.

Maptiler has a bunch of datasets, anywhere from 385MB to 527GB but the OSM dataset is only 70GB. (MBTiles format)

Things lining up (aka luck) is important but people who work hard create more opportunities for that to happen.