I wish I had the confidence of the 5-year-old who just walked up to me in a hospital hallway and said, “Bad news mister, my mother is lost.”

Postdoctoral Fellow/Research Fellow - Canberra / ACT, ACT, Australia Classification: Academic Level A/BSalary package: $89,313 - $137,870 per annum plus 17% superannuationTerm: Full-time, Fixed Term (Up to December 2027) The Position Up to two academic positions are av...

Come and join our team! We are looking for:

- Postdoctoral Fellow in Statistics
- Research Software Engineer (#rstats preferred)

This is part of the Analytics for the Australian Grains Industry project (see anu-aagi.github.io).

jobs.anu.edu.au/jobs/postdoc...

jobs.anu.edu.au/jobs/researc...

The units are Covid Genome copies per person per day in wastewater for the period (national aggregate), divided by number of people admitted to hospital for the period (national aggregate). So recently it has been taking 56000/40000 = 1.4 times the amount of covid out there to hospitalise someone.

Banner for Wellington WikiCon ’26, photo by Ulrich Lange. The 26 should have an apostrophe in front of it, not an open single quote, but I guess I am the only person who cares.

I’m going to be at the Wellington WikiCon in 10 days, and would love to meet the wiki-curious. Also I’m going to sneak into Commons and turn that quote mark into an apostrophe before it drives me nuts. en.wikipedia.org/wiki/Wikiped...

FWIW the whole population admissions rate from the data we have is lower- in July 2025 it was about 1 admission per 40000 covid samples in wastewater. For the latest data it is about 1 admission per 56000

I was quoting hospital admissions- so reported cases is not something I have reliable data on that handles underreporting. But for hospital admissions (so testing for respiratory admissions) the fall is in older ages with the number of under tens being at around the same level.

I think I will start carrying my camera, so having a bit of optical zoom on hand, and making sure the battery is charged

FWIW, without knowing what the exact variant people had,
in the 28 days to 30 July 2025 there were 50 under 10 hospitalisations (and 400 older for 11% <10).
in the 30 days to 19 April 2026 there were 55 under 10 (227 older folk for 20% <10) but driving that is the fall in older hospitalisations

Screenshot of admissions showing down generally except in Central, but in the specific DHBs matter more

Based on Hospital Admissions, it looks like this covid wave is now spread out to the Waikato/ Mid-Central area, after being in other places( with major centres having dropped).

Relevant to that dirty Australian plan to make urea from lignite in Southland, here's the PCE in 2010 saying its a terrible idea. Using their numbers, 1.5 MT urea means 2MTCO2 emissions every year. Which completely blows all future carbon budgets pce.parliament.nz/media/m1vcbt...

The cool thing about smart bike parking like Locky Docks; it comes with lots of data.

“Bike parking sessions across the Locky Dock network rose 36%, from 1,280 to 1,738 in the last month.”

Read more;

www.linkedin.com/pulse/fuel-s...

Uh-oh. Do we need a new version of @magnets.eurosky.social’s brilliant observation about how climate change will manifest?

In a surreal twist, people are now doubting footage that is both very real and very close to home—a result of AI imagery poisoning the waters:
www.tickaroo.com/e/GCRysAwNeu...

This is horrific. I feel physically ill.

(Reposted with ALT text)

I always suspected the bits of pasta shaped like shells.

Beautiful sunrise with a palm tree in foreground , sky tower in background with an orange pinky sky

What a sunrise today

Left-spiralling snail Ned dies before finding a mate The rare snail's quest to find true love has come to an abrupt end.

Ned is Dead. Dead on Wed. So it's said. Ned is Dead.

www.rnz.co.nz/news/communi...

Ah, but under the 102 Billion AI model, in the next (checks current year) 12 years AI will lead to 0 unemployment with the AI doing all the low productivity jobs, and (all) humans working high productivity jobs. That has to be worth a bunch of GDP.

Seriously, that is the model.

#Covid #BA.3.2

Confirmation by New Zealand's PHF Science:

BA.3.2 is more frequent in children.

bsky.app/profile/domi...

Editors should not upload manuscripts to AI without permission, says Authors Guild The Authors Guild (AG) in the US has warned editors not to upload submitted manuscripts to consumer-facing AI systems.

Ooooh, this is gonna be a mess

www.thebookseller.com/news/editors...

To speak about myself in the third person, "To the extent that numbers not matching observed reality makes people sad, this report makes David sad"

Road Snowfall Warning issued for Desert Road (SH1). Road Snowfall Warning issued for Lewis Pass (SH7). Road Snowfall Warning issued for Porters Pass (SH73). Go to https://metservice.com/warnings/home for more information

Perhaps they asked a Generative AI, in order to be more productive.

I've been trying to match the original numbers in the 2024 report that is quoted to make more detailed commentary on it, but measuring the key graph in the Microsoft and Acceture report it has 'past' NZ GDPs of 357 billion for 2023 and 362 in 2024. And that matches no record I can find.

GBB in the Ōamaru Mail!

We are always thrilled to share our local stories with local media, connecting to communities about our mahi in their neighbourhood 💛

Matching seasons by shifting things six months, it is the southern hemisphere's Halloween/Samhain in about 9 or 10 days

The whole web is built on a primary grace: that the people visiting you are trustworthy. It costs a little something for me to serve my webpage to everyone who goes there. When the number of scrapers outnumbers the number of humans by an order of magnitude, this grace is fragile.

I guess that is kind of "an announcement" on the subject of securing more fuel.
bsky.app/profile/thep...
(but more seriously, changes in supplier policy is yet another stated reason to change fuel alert levels)

Dozens of AI disease-prediction models were trained on dubious data The models are designed to predict someone’s risk of diabetes or stroke. A few might already have been used on patients.

In a new preprint, the authors identify 124 peer-reviewed papers which used provenance-unknown Kaggle data to train AI for stroke and diabetes risk. Some sets show impossible patterns, yet models of the models are already hitting clinics.
#MedSky #MedAI

For my bubble, R data traffic has been fairly constant, but I think it worth formally noting that social media takes place within a social world and the individual's placement within it. And neither of those things are constant, so change the relationship to social media
bsky.app/profile/smac...

Chris, you shall be visited by three spirits. The first spirit is the ghost of caucus past.