We like Langfuse for observability via OpenTelemetry. Prompt management is too basic for our needs.

Let me know what features are missing in prompt management

I've had the same issues until I got top of the line Roborock. It never loses map and almost never gets stuck, washes itself with water and drain connection. I'd rather not buy a Chinese brand but here we are...

Then countries with even more greedy elites would have even lower rates, which is not the case.

A few examples of fertility rates (higher is better, 2023 data):

  Sweden: 1.45
  US: 1.62
  Kenya: 3.21

You’re not making the point you think you’re making

Everything is relative. I'm an immigrant from a post-USSR country and the US is still orders of magnitude more democratic and free

Doubt. What are post-USSR countries (except for Ukraine) where government detains lots of people who hasn't committed any crimes? How many people get wrongfully killed by cops in post-USSR countries? In U.S. that's like a sport for cops to find an excuse to unalive someone.

And what is democratic about the fact that majority of people votes for candidate A, yet candidate B becomes the president because... because it's people don't actually vote for president, they vote for someone who counts pro-some party and it's THEY who vote for president in the end. What's democratic about corruption being completely legal (lobbying)? Do you know a single post-USSR country where lobbying is legal? (Hell, how can it be legal at all? there's no distinction between lobbying and corruption, that's the same thing!)

How does it compare to docling?

Docling primarily uses AI models to extract PDF content, this project looks like it uses a custom parser written in Java, built atop veraPDF.

Correct me if I am wrong, but Docling can do both. It has also, among other strategies, a non-AI pipeline to determine the layout (based on qpdf I believe). So these projects are not that different.

While it has a PDF parser, my understanding is that it is mainly used to break a PDF document into chunks, which are then handed off to various specialized models. From its docs: "The main purpose of Docling is to run local models which are not sharing any user data with remote services."

Who did he took over SpaceX from?

Glad that you had to dig that deep to find something wrong. But: SpaceX was taken over from NASA/DoD (~ TRW). People generally don't say that because the people actually involved feel that it was NASA that abandoned them/their ambitions rather than the other way around.

But you know the same story emerges as elsewhere. Mr. Musk's was a financier, nothing else.

This is the most nonsense of nonsense claims.

Electrician or plumber

I’ll put you to work as an electrician if you get a journeyman license in Minnesota, my job selling and running electrical work won’t be automated any time soon.

thinking models produce a lot of internal output tokens making them more expensive than non-reasoning models for similar prompt and visible output lengths

It can be relatively safe with proper handling. There are independent laboratories you can send this "product" to for analysis. Often people organize group buys with some random samples tested.

US has Eli Lilly with a competing product (Tirzepatide)