iiinova — where your digital self keeps evolving

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 07/10/2026

Listening to a podcast about immigration. "No one is saying open borders" Me:

Mr mole saying "I was saying open borders"

011 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 26/09/2026

Oh, great, just the thing I wanted to scream into

Floorplan with a big chunk labelled "void"

012 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 17/09/2026

Google docs now has option to download as markdown. I wonder if we could use this to get the holy grail rmarkdown -> docs -> rmarkdown chain working. #RStats

screenshot of google docs download menu with the markdown option highlighted

131 3

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 16/08/2026

Yesterday I was blown away by Deichman #Library in #Oslo. Is a gorgeous building full of lovely places to read, draw and even practice the piano! They even have a mini cinema that you can book!

Photo of the outside of the Deichman Library.

Photo of a nook inside the library with a table and a book add filled with white books children can draw in. The table is all drawn in as well as a whiteboard behind it

Photo from inside the library. A big empty space on the middle lets you see up to the ceiling and through the 4 floors.

Photo of a section of the library. To the left there are bookcases full of music sheets. In front, two electric pianos free to use. A woman is wearing headsets and practicing.

0249 24

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 09/08/2026

There are two types of lines on a map. Some lines, like roads, buildings, trails and rivers, Rey to represent something that exists in the world. You can go there and see the road, or the river. But there are also fake lines. The line here that divides Norway […] [Original post on mastodon.social]

Map in the border between Norway and Sweden. A thick orange line marks a road and a thin brown line marks a trail. Both cross the thin black line retesting the border. There are also blue polygons repressing lakes, an icon representing a bench.

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 09/08/2026

#Google Maps is so utterly broken. If you search for Muchmuseet, it zooms in on the old location near the Botanical Gardens, but it highlights the new one near the Opera House. #Oslo

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 04/08/2026

Like look at this. In the last 6 years 205 people were sent to the Nauru Concentration Camp. Taking the cost rom the previous article, that's a cost of about AUD 2.400 million, so #Australia is paying about AUD11 million to torture each refugee.

Bar chart titled "How many people have been sent to Nauru since May 2022?" as at 31 May 2026. Total: 205. By year: 2021: 0, 2022: 0, 2023: 23, 2024: 113, 2025: 40, 2026: 29.

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 28/07/2026

Experimenting with supersampling with splines to smooth contours when the resolution is kinda coarse. My logic is that contour methods assume linear interpolation, but that's just one assumption. Why not assume a smoother function? So, I first interpolate to […] [Original post on mastodon.social]

002 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 20/07/2026

If I didn't mess my maths up, if you take a random German speaker test taker and a random English speaker test taker, then the German will get the same or better score more than 2/3s of the time. That's kind of nuts. #RStats #Language

Bar chart ranking native-language groups by the probability that a random test taker scores the same as or higher than a random English-speaker test taker. The x-axis shows percentages from 0% to about 70%. The y-axis lists native languages. German has the highest value (67.3%), followed by Marathi (60.5%), Russian (58.1%) and Afrikaans (57.1%). The lowest values are Japanese (29.6%), Pashto (22.3%), Thai (21.0%), and Dari (15.4%).

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 20/07/2026

I am shocked at the low scores in the listening part. I expected low scores in the writing part, since it's less a test of English language and more a test of particular writing skills. But I am very surprised that native speakers had a hard time just jotting down information read to them.

Screenshot of IELTS test statistics showing mean scores for English speakers. The scores area 6.4 for listening, 7.0 for reading, 6.5 for writing and 7.1 for speaking

110 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 20/07/2026

I was curious and found mean IELTS scores by native language (2024-2025 statistics). It turns out Germans, some Indians, Russians and South Africans speak better English on average than native speakers.

Plot showing distribution of IELT scores by native language ordered by median score.
German, Marathi, Russian and Afrikaans have a higher median score than English. The distributions are also different. The most frequent score for Germans and Russians is 8, while the most frequent score for English-speakers is 6.

101 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 09/07/2026

There are boxes of ArcGIS CDs in the "Free" area of my office. Who wants them? #spatial

Photo of several boxes on a shelf. One is open, showing a big book of ArcGIS9 and small ArcGIS CD sleeves.

021 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 06/07/2026

If you check the pronounciation of "bona fide/s", cambridge dictionary has the same phonetic alphabet for UK and US English but completely different audios. dictionary.cambridge.org/us/pronunc…

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 27/06/2026

I forgot my keys and got locked out of my house, so I decided it was about time I learned how to get in without them.

Photo of a transparent lock with lock picking tools inserted into it.

131 3

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 19/06/2026

Good news, everyone, google has come up with a new chat system.

Screenshot of gmail's sidebar with a section labelled "Chat with a colleague or a friend" with "Spaces" that are chatrooms.

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 12/06/2026

Hey, #RStats nerds and other nerds too. I'm trying to come up with a functional from for this curve. Something parametric that I can then think about symbolically and , not just a statistical fit. I can't find anything simple enough that fits the core […] [Original post on mastodon.social]

Plot with a noisy curve and a smooth fit that matches that description.

102 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 08/06/2026

RE: hachyderm.io/@groxx/116711385714918… This is a neat illustration of how our eyes+brain use green for sharpness information and the rest is just the suggestion of colour, especially blue. First, original image. Second, heavy gaussian blur […] [Original post on mastodon.social]

Photo of statue of Yaksha supporting the base of one of the Two Golden Chedi of Wat Phra Kaew, Bangkok, Thailand. The status features a very intricate dress pattern with blue, white, golden and red. It's skin is red. The background wall is golden.

Same photo but there is a heavy green bleed all over the image. It looks blurry, with undefined edges all over. The colours are also very undefined, except for the golden wall and the red hands and feet. It's so blurry and messy that it's almost hard to focus the eyes.

Same as the original image but now there is a redish bleed over it. Edges are still relatively sharp, although some colours are lost. Still, it looks like a regular photograph, maybe with a bad filmstock or strange lighting.

Same as the original image. In fact, it's almost the same. The dress looks a bit more blue all over the dress, but if you hadn't had the original image to compare it, you probably wouldn't notice it. Nothing really jumps as strange

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 06/06/2026

Is this a thing? Are people mining crypto on 5 year old phones?

I have a moto G Stylus That I use for mining crypto with and it has a (XT2215)

MD50 Li-ion 00 battery and I am trying to see if I can change it to a (XT2131)Li-Polymer MD50 Battery thw old Li-ion battery has started to swell up and I am hoping that the Li-Polymer battery will solve this problem I would appericate any help thanks in advance

101 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 02/06/2026

It took me a while to understand this #XKCD comic until I realised that Randall lives in he northern hemisphere where this actually makes sense. xkcd.com/3253

Xkcd comic. Tree people stand to the left to two tables separated by a sunbeam coming from the back wall. One person says "Let's take the far table. The closer one will be in the sun soon." The caption reads "Astrophotography gives you exactly one extremely minor superpower"

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 31/05/2026

Reviewing the materias for reproducibility.rocks and this sentence caught me by surprise. Clearly past me didn't have a good relationship with LaTex. TBF, neither does present me xD

LaTeX

You’ll need a working LaTeX installation to render PDFs. If you didn’t feel chills down your spine at the sight of that sentence, consider yourself lucky!

101 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 18/05/2026

New version of RStudio will add these lines to your ignore files. Automatically. Each time you open the project. Even when you don't have any of those tools installed. I think the time to jump ship is approaching. This shit is going way to far.

Screenshot of rstudio git window showing new additions to .Rbuildignore: ".positai" and ".claude".

110 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 12/05/2026

Yearly gini coefficients of my music taste. Inequality in this context is good; it means more artists. It seems my listening habits have been a bit narrower since 2024, although this year is still not done, so there's time. #RStats #Lastfm

Line chart showing artist scrobbles inequality from 2008 to 2026, with point size representing number of scrobbles (2000–6000). Early years (2008–2015) have smaller point sizes indicating fewer scrobbles (~2000), while post-2017 points are noticeably larger, reflecting higher scrobble counts (~4000–6000). Inequality fluctuates between roughly 0.3 and 0.8, with a sharp dip to ~0.28 around 2015, a recovery to above 0.8 by 2020–2022, then a decline to ~0.63 by 2026.

012 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 11/05/2026

This pitiful thing is jupyter's "save as" dialogue. You need to write the route directly. No UI to navigate, no autocomplete paths. "home" doesn't even truly exist; I assume is some lousy shortcut for $HOME or something. It's utterly unusable. The fact that […] [Original post on mastodon.social]

Screenshot of a white pop up dialogue with the title "Save file as..." with a single lonely textbox that starts with "home/research/Obs-seaice-analysis" and two buttons: Cancel and Save.

011 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 10/05/2026

Oh, my fucking god. The "Hallucination detector" is just another LLM. This is not serious.

python code that defines a "create_hallucination_detector()" function with these instructions:

You are tasked with spotting potential hallucinations in the paper. Look for:
1. Claims lacking citations
2. Data inconsistent with official sources
3. Conclusions not supported by presented data
4. Invented or malformed references
5. Statistical impossibilities or contradictory numbers
6. Technical terms used incorrectly
7. Non-existent methodologies or tools

Provide a concise report IN ENGLISH detailing any suspicious statements, with specific examples from the text.

End your review with: "REVIEW COMPLETED - Hallucination Detector"

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 10/05/2026

In the #AI era, this is what passes as a "sophisticated algorithm"

Dynamic Model Selection

The system uses a sophisticated algorithm to select the optimal model for each task:

final_score = (paper_complexity * 0.4) + (task_complexity * 0.6)

if final_score >= 0.65:
model = "gpt-5" # Complex tasks
elif final_score >= 0.45:
model = "gpt-5-mini" # Moderate tasks
else:
model = "gpt-5-nano" # Simple tasks

202 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 04/05/2026

I simply cannot understand how #Slack can fuck up notifications so badly. I toggled all the settings I could think of to get a notification for every single message in a channel. And today I open slack and... there's a new message from two days ago and I […] [Original post on mastodon.social]

Screenshot of slack settings showing mobile notifications turned on and notifications for "all new posts" enabled.

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 03/05/2026

Learning to make genuine Australian dishes.

photo of 5 sushi rolls, one of them cur in half.

011 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 15/04/2026

Google calendar's timezone selector has always been bad. Now they changed it to have search, which is good, but it only understands cities, so you can't select UTC. How can they fuck this up so royally?

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 09/04/2026

Preparing our haul to bring back to Australia.

photo of a shelf with a bunch of food. 3kg of yerba, 3 packets of chocolate cookies, 2kg of dulce de leche, a mate and other sweets.

011 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 07/04/2026

Oh, boy. This last name didn't age well, did it, The Expanse?

"The Epstein Drive hadn't given humanity the stars, but it had delivered the planets"

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 30/03/2026

Things that I *didn't* miss from home. The utter terrible airlines with horrible to nonexistent customer service. #Flybondi

screenshot of 51 airplane icons, 38 of which are black, indicating cancelled flights.

001 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 18/03/2026

Something that I really missed about home ♥️

Photo of an oorange tabby cat laying on the bed, on top of a tote bag

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 18/03/2026

Visiting Argentina and today I experienced the whole culinary spectrum. Terrible coffee and amazing pizza.

Photo of a plate with a slice of pizza with green veggies and white sauce and a slice of fainá.

112 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 10/03/2026

RE: mastodon.social/@glynmoody/11620661… I really don't understand people who think any "assurance" by the current US regime is worth a single dingleberry.

The administration can offer clear assurances that visitors here legally will not be targeted by enforcement actions at games, fan fests, or watch parties.

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 25/02/2026

This seems to be very model-dependent, though. llama3.1 seems to be better calibrated.

000 0

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 25/02/2026

RE: mastodon.social/@eliocamp/116124390… As a followup to this post, I looked into token probabilities using #ollama and... it's not pretty. With a question with a clear answer, the probability is concentrated in the correct answer; good. But […] [Original post on mastodon.social]

Barplot showing the probabilty of various tokens for the question "is the sky blue?" "yes" is at 1.

Same plot as before but for the question "is the sky green?" "no" is now the token with 100% probability.

Same as before but for the prompt "Imagine you flipped a coin, did it land heads or tails?". "tails" has virtually 100% probabilty.

Similar plot as before. Two panels, one for "pick the number 1 or to at random" in which 1 is virtually 100% probability and the other for "pick the number 2 or 1 at random" where 2 is virtually 100% probability.

102 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 25/02/2026

The publication includes two nonsensical plots comparing "sequential deliberation" with "simultaneous deliberation". They are utterly uninterpretable and they bafflingly use completely different conventions of colour, vertical axis order. Why does one plot […] [Original post on mastodon.social]

Plot titled "sequential deliberation" the y axis is labelled "participants" with the number 0 to 25 going from bottom to top and the x axis is labelled "time (minutes)". There are non-overlapping bars of different colours representing when each participant is talking within the 60 minutes covered by the plot.

Plot with no title. The y axis is labelled "participants" and goes from 0 to 25 top to bottom. The y axis is labelled "time (minutes)". The whole plot is coloured with a continuous colour scale (viridis) going from yellow to purple.

100 1

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 18/02/2026

Is my PDF viewer gaslighting me? First image is a pdf figure rendered by firefox, second image is the same pdf figure rendered on #ubuntu's default pdf viewer. The gray line is now black and there's a whole extra line! What's going on here? #RStats

Image of a line plot with four lines: black, gray, purple and yellow.

Similar image, but the gray line is not much darker and there is an extra gray line that follow the purple line just slightly higher.

002 2

Elio Campitelli @eliocamp.mastodon.social.ap.brid.gy · 17/02/2026

Forgot to post another progress report. Sea ice is DONE! The only issue is that now I need to fill the whole ocean with black. I might die. #CrossStitch

Photo of a cross stitch pattern showing outline of Antarctica in gray and white surrounded by sea ice represented as white and shades of blue.

010 1