mdsumner

mdsumner

@mdsumner.bsky.social

maxing the pixels to glow R, Python, GDAL, virtual Zarr, coordinate systems, southern ocean research I sit and reflect about all that I've learned and all that I've seen Triangle wins hypertidy.org

614 followers535 following1K posts

mdsumner reposted

Hadley's post clarified for me the role the language we use plays in describing AI "collaborate" and "expertise" are the kind of terms I'd use for a person, not a tool (you wouldn't collaborate with a hammer) AI might be unlike any other tool we've created, but it's still just a tool

Hadley Wickham@hadley.nz

I made an AI video about genzplyr and I lowkey love it. It's also 100% slop. In this post I dig into what makes something slop, when it's ok to share it, and how to avoid it in work that matters. Read more at: substack.hadley.nz/p/what-is-slop

mdsumner reposted

Replying to @bharrap.bsky.social

Oh I thought of one myself! Haven’t you ever _fought_ with the computer to get something done?

mdsumner reposted

The Australian newspaper has been claiming a "forensic audit report" shows BoM's temperature data has "widespread error". For 15 years climate sceptics have been claiming a conspiracy around the data. Is anything different this time?

theguardian.comThe BoM’s temperature record keeping is under attack – again. What’s going on? | Temperature CheckAn β€˜audit’ of the Bureau of Meteorology main dataset for measuring climate change claims it contains β€˜widespread errors’. It doesn’t stack up

mdsumner reposted

The PM has been telling everyone to turn the temperature down. He might want to have a word to Lachlan Murdoch about turning the temperature up with climate-denying conspiracy theories about BOM manipulating temperature data. #auspol

Graham Readfearn@readfearn.bsky.social

The Australian newspaper has been claiming a "forensic audit report" shows BoM's temperature data has "widespread error". For 15 years climate sceptics have been claiming a conspiracy around the data. Is anything different this time?

mdsumner reposted

πŸ›°οΈ Open Source Tool of the Week: WebODM! Turn drone photos into detailed maps, 3D landscapes and elevation models. Great for environmental research, agriculture and surveying. Open source can map the world! 🌍 webodm.org #OpenSource #Linux #WebODM

mdsumner reposted

My British partner has introduced me to a piece of media called "Countdown" that she alleges is actually some form of entertainment?

Including crop_burn() and materialize_chunk() to batch generate actual pixels a combination of {fasterize} and {exactextractr} see blog for practical application at REMA v2 scale www.hypertidy.org/posts/2026-0... (also available as Rust crate, and Python package) 2/2 #RStats #Rust #Python

hypertidy.orgWriting controlledburn output in parallel – hypertidyThe serial COG write loop from the last post, replaced by 24 workers writing disjoint Zarr chunks. The 31-billion-cell grid on disk in 28 seconds of 9.5 MB.

mdsumner reposted

Stumbled slightly teaching #rstats ggplot2 today until I remembered the palmerpenguins vs penguins situation. Debarcle! Up there in competition with Microsoft for worst reimplementation of existing concept.

mdsumner reposted

PyTorch won’t be unseated by JAX, nor anything that NVIDIA invents. It will be unseated by DuckDB.

mdsumner reposted

I made an AI video about genzplyr and I lowkey love it. It's also 100% slop. In this post I dig into what makes something slop, when it's ok to share it, and how to avoid it in work that matters. Read more at: substack.hadley.nz/p/what-is-slop

substack.hadley.nzWhat is slop?I made an AI video about genzplyr and I lowkey love it. It's also 100% slop. In this post I dig into what makes something slop, when it's ok to share it, and how to avoid it in work that matters.

I did not ever use it, months later I found Manifold.net (the govt used mapinfo and TAB was *it*, so Manifold was the go) at the same ish time I was starting to use #RStats βš‘β˜οΈπŸ’―

mdsumner reposted

"deploy this shiny app" πŸ‡ is all it took to deploy an app to @ricochet.rs Our new Claude Code and Codex plugins makes pushing to prod a whole lot simpler for data scientists. πŸ‘‡ An actual recording of deploying an app. #rstats #databs #julialang #python #claude

mdsumner reposted

The most underrated thing about teaching Intro to R is that you get to choose how cohorts of #rstats users will go on to pronounce package names dplyr? dplyr? Or is it dplyr? They don't know, but they don't know that you don't know either!

Massive formal upgrade and robustifying of our cloud-enabled GHRSST cog/geotiff store source.coop/ausantarctic... #ausantarctic #cng #source.coop Indexes coming soon to adding to the flat and STAC geoparquet to feed and satisfy all your Zarr and GDAL and QGIS and R and wasm and more engines

mdsumner reposted

I've got mixed feeling about Claude and agents in general, but if we can make the part **after** you build something as simple as a single sentence, that is the ergonomics I think we all need.

ricochet πŸ‡@ricochet.rs

"deploy this shiny app" πŸ‡ is all it took to deploy an app to @ricochet.rs Our new Claude Code and Codex plugins makes pushing to prod a whole lot simpler for data scientists. πŸ‘‡ An actual recording of deploying an app. #rstats #databs #julialang #python #claude

mdsumner reposted

Replying to @matthewfeickert.com

> If conda[-forge] aligns versions ...across platforms and is recent that's very attractive Yes, see some of the points in Chris's Zulip server response (linked in bsky.app/profile/matt...).

Matthew Feickert@matthewfeickert.com

Chris Toney also recently commented on a thread on the conda-forge Zulip server RE: this post to give some more context on the boons that conda-forge builds are providing and also where there could be better support for GDAL + Windows: conda-forge.zulipchat.com#narrow/chann...

mdsumner reposted

Replying to @matthewfeickert.com

> ten years ago ... it was so blatantly not fit for purpose. I just didn't go back it was so ridiculous. I had a very similar experience in 2014 and I described myself as a "conda hater". Then in 2023 Pixi was created, I learned about it, I tried it out and I fully changed my views.

mdsumner reposted

Replying to @matthewfeickert.com

I realized that the conda packaging standards are great and I had really bounced off hard of the tooling and infra at the time. I'm now a big proponent of conda-forge (as you can tell by me not shutting up about this). :P

mdsumner reposted

Amy Acton is running against former GOP presidential contender Vivek Ramaswamy for governor of Ohio. She knew when she decided to run that the attack ads were inevitable. But what happened yesterday was beyond the pale. Link in comments.

mdsumner reposted

Replying to @mdsumner.bsky.social

Yeah it's making me feel a bit sad and more conflicted about it all! Getting less and less keen to share stuff with the wide ranging opinions floating around these days... Should I submit that real gdal bug that ai found or not?!

There's a bitter taste to devs being told off for being too ... literally whatever ... atm Everyone is standing on someone's damn lawn, take a chill pill gawd.

Alright handsup all you treacherous worms, who's vibing a RIIR GDAL ... I'm reminded of fortunes("diamond") #RStats πŸ˜‚

mdsumner reposted

TIL of `arrow::map_batches()`: just converted 2gb of CSV files (11.2m rows) into geoarrow in 0.3s 🫒 #rstats

mdsumner reposted

tigris v2.3.0, one of the major milestones from this summer's R Consortium grant, is now on CRAN. - The package has a modern httr2 backend now, which will help the package last into the future; - New features include `population_centers()`, shown here; - The open issue count is now at zero!