One question about OIDN: is the speed increase due to the fact that we can get away with smaller sample sizes, or should I also be seeing speed increases when using the same sample sizes as before!? 🤔 I am very happy with the noise reduction, but can't really say I noticed shorter rendering times...

I tried migrating from Rstudio, but was really disappointed by the debugging experience, see e.g. github.com/posit-dev/po.... I think the docs and migration guides should warn people that the debug experience is not up to par yet. For analysis, it's great, for package development, not so much...

I hope to wrap this up for Christmas. After, I am determined to write up the experience, share some code, and to properly thank some of the tool builders - you first among them! 😊 I have spent way too many hours on this project, but my head is already full of ideas for future development... 🤣

I did unsuccessfully experiment with transparent_background = TRUE before posting, but sometimes it is hard to keep track of the the parameters and their intended interactions between plot_3d, render_highquality and render_scene.

Thanks Tyler, much appreciated! 👍 It is too bad there is no way of removing the noisy background AND of keeping the shadows (meaning you can layer them on an arbitrary non-noisy background color), but I can see why this is not an easy ask.

Naive cropping using fuzzy color selection is unlikely to work reliably for my set of input images. Just to give you an idea, here's a render of a 3x3 km tile of my home city of Münster:

Using higher values for "samples" or denoising via OIDN (can't get it to work on Windows, see github.com/tylermorganw...) would help, but not really solve the fundamental issue. Gist here: gist.github.com/RichardMeyer.... Any ideas greatly appreciated!

I tried to use the left image as a clipping mask for the right picture. This fails, among other things, because the parameter 'shadowwidth' to rayshader::plot_3d is only respected when shadow = TRUE, meaning dimensions and positioning will always diverge between the two renders.

Great read, thanks for sharing! 👌

This is the kind of content I'm here for, well done! 👌 Focusing on what truly matters! 🤣

Maybe check this article "R Packages for Data Quality Assessments and Data Monitoring": www.mdpi.com/2076-3417/12...

A good overview of some of the relevant literature is in Lindner, Fabian: Does Saving Increase the Supply of Credit? A Critique of Loanable Funds Theory. In: World Economic Review 4, 4. (2015), S. 1-26. Particularly love to see him quote Wolfgang Stützel, who really should be more widely read.

I know Bluesky is not a place for nuance, but I am always surprised to see Triffin brought up in such a mechanical, law-like manner. Personally, I think the work of the BIS has compellingly added a lot of qualifications to the dogmatic Triffin view, e.g. this piece www.bis.org/publ/work684....

Thumbs up to your old self for "never again, transparency!!" 😂. We've all been there! ☺️

I have been reading your work since 2010 or so. Really looking forward to this attempt to synthesize your thinking! I am not as widely read in political economy as I used to be (having left academia), but I always thought your perspective was rather unique. I hope this work will find its audience!

Fully agree, it would be a shame if they went under, or full on corporate. They launched Plot in 2021 (still feels newish to me), Framework after, and now Canvas. I personally struggle to see how such a small team can push all this simultaneously, but maybe they can! I sure wish them well!