Ok, now you need to explain how

- R helped you to reverse engineer the state lottery raking in millions^Hthousands for a new Renault

- ....

- PROFIT!

Mortgage math is a form of bond math and takeaway is that the damn *compounding* grows strongly with both _term_ (this $deityforsaken example) and _rates_ (which you may call expensive in historical terms). So... yeah. Be a bank if you can.

I used the first calculator that popped up.

Total interest payment in Mike's example:
30 year $460 (total payments $840k)
50 year $857k (total payments $1237k)

In my next life I want to come back as a bank.

That is common for add-on repositories, also covered in our FAQ. If you know your way around Ubuntu you can use it: I do on my laptop currently running 25.04 (and 25.10 "soon").

Are you compiling from source ie on Unix? If it is Ubuntu, consider #r2u which has all of CRAN (and good chunks of BioConductor) as binaries for #rstats so something like `install.packages("tidyverse")` installs everything in a few seconds -- and all dependencies.

eddelbuettel.github.io/r2u

Tthe argument presented is circular.

That sophisticated sounding "machine learning" talk is added ("oh, look, a quant said so") adds nothing to the fundamental flaw. The method described is standard and calling "boosting" (== combined several weak "learners" to create something better). 2 of 2

I may be behind the eight ball but I still think "it's complicated." These labs may release weights as well as some code tar.gz archives, but we have neither the training data nor parameterisation let alone compute infra so we cannot reproduce the weights, can we? So is that "open sauce"?

Open source != open weights

Open weights is all we have here. Better than nuttin' but also not "open sauce".

Thanks for the note, it is appreciated!

Posting on the forum requires an account which implies tracking and whatnot. As I don't have one that won't work for me, sadly.

Let me introduce you to my new set of friends. They call themselves large language models, and they may have played in that same (rather dirty) sandbox ...

The key is that _inside_ the .deb we ship is the same partial filesystem tree you get from p3m.dev, or would get from r-universe, or from doing R CMD INSTALL --binary locally) *but* very importantly we create the .deb layer around it with *full and complete* dependency resolution. That's the magic.