Male Whinchat perched on barbed wire, rusty brown with a bold pale eyebrow. iPhone 8 + Kowa TSN 773 scope
Lovely migration Sunday, cycling across the Saclay plateau near Paris, France
🎉100th species of the year: Garden Warbler.
More notable was #99 Whinchat, by a freshly cut field still full of appeal.
Also #98 Northern Wheatear, #97 House Martin, #96 Spotted Redshank and #95 Wood Sandpiper.
#birds
I’ve always asked myself why all storks except the European one look ugly to me, is it because I’ve long known only the European one, or the European is really special.
A light green shared bike on a city plaza
That feeling of freedom when unlocking my first shared bike in a foreign town.
Amusing to hear journalists (just now on BBC News) talk about transporting 500 kg of 50% enriched uranium like it’s just a 500 kg pallet that would easily fit in a helicopter.
Not an expert, but still a physicist, I’m pretty sure it’s not bubble-wrapped…
#Iran
The image compares first image generation of birds and flowers (very clumsy) with text to image GAN in 2016 (Reed et al), with near perfect generation using the exact same prompt with Nano Banana in 2026
Brushing up some slides for the #COMCHA school in Zaragoza, I realised it's been 10 years since the first modern text-to-image paper with GANs (Reed et al., 2016). I ran the exact same prompt through Nano Banana ... the field has come a long way...
@aishikghosh.bsky.social, remember? #hepml
Ah, copyeditors OK, but my point is to do it *before* the refereeing process starts
Interesting, which editor ? Mine was Springer for EPJC
In fact, publishers should really screen for inaccurate citations; it should not be more complicated to put in place than anti-plagiarism filters.
Absurd to leave humans to handle this manually.
Close up view of a bike frame, clearly broken
Vintage blue bike, "Jacques Anquetil" printed on the frame
Vintage blue bike, "Jacques Anquetil" printed on the frame, in front of Jacques Anquetil memorial at Villiers le bâcle
End of the road for my vintage "Jacques Anquetil" bike. Bought it on a whim at a bike flee market, just before Covid, entirely refurbished, rode probably 20k km since.
Jacques Anquetil was a french cycling legend in 50-60, with a messy personal life (look him up), I discovered later.
Yay! As an associate editor for #EPJC, I just rejected my first paper due to poor gen AI use.
The core work was honest and publishable, but the intro & state-of-the-art were botched, with hallucinated citations!
#academiclife
When I left home for uni, I got a cassette player + one tape: Dvořák’s New World Symphony.
By year’s end, I had many but now, any two seconds of that symphony freaks me out.
At least he says: « There is a problem, f* you »
Trump would say « f* you, there is no problem »
See?
Malacca is underrated, I had no idea until recently that crossing from Indian Ocean to Pacific Ocean was non trivial
Volkswagen dieselgate vibes
So when I hear people building “society of AI agents” …
Don’t we already know how hard it is to make a society work? Don't we see the evidence every day? 6/6
They kept the best parts of object orientation, especially interfaces, but clearly separated: Algorithms (and later Tools) that do things and Data that flows between them. 5/6
Thankfully, visionaries like Markus Frank and Pere Mato at CERN went against the trend. They introduced Algorithm / Data separation, which became the foundation of the Gaudi/Athena framework still used today by the ATLAS and LHCb experiments at the LHC. 4/6
My takeaway at the time: emergent complexity is a science of its own. How do ants build nests without an architect?
I even put a picture of an ant cemetery in my Habilitation.
How can you be confident you'll be able to specify small-object interaction rules and achieve what you want? 3/6
For particle physics:
Calorimeter cells were supposed to organise themselves into well‑calibrated clusters.
Tracking hits were supposed to assemble into well‑fitted tracks.
And properly identified photons and electrons emerge.
In practice, prototypes failed spectacularly.
2/6
#AgenticAI reminds me of the object‑oriented programming craze of the 2000s. Trained in Fortran/C, I dutifully attended two week-long OOP courses.
The mantra was: “There should be no God object”. No SomethingManager. Complex behaviour should emerge from the interaction of many small objects 1/6
A slightly blurry photo of a Jack Snipe in a swampy area among twigs. The picture was taken by holding a phone up to binoculars, and the phone focused on the twigs rather than the bird.
This blurry bird is a Jack Snipe. A notoriously hard bird to spot, it only flushes when you’re almost stepping on it, it’s called the “deaf” Snipe in French.
Detected with a thermal monocular, which is very effective for these cryptic birds.
First one spotted this spring in Paris region.
#birds
Yes but my gripe is that this is the one which was chosen for the olympics, to be compared to Olympic triathlon which is a ~2 hour effort.
This ski-mountaineering event at the #winterOlympics #olympics is completely ridiculous.
These athletes have insane fitness, but it all comes down to how fast they can put on/take off their gear and handle their equipment.
Fine for a 2-hour triathlon, but for a 3-minute event? Come on.
What’s great since the advent of LLMs is that now when I write a review, I don’t bother polishing my style, on the contrary, to show that this is 100% homemade, handcrafted review, folks.
Like leaving a few lumps in the mashed potatoes
#academiclife
Aha, to the gallon !! Glad you made it work…
Can’t wait for predatory journals to start using AI seriously.
I’m so done being invited to resubmit my "AI for Particle Physics" paper to the "Journal of Advances in Civil and Mechanical Engineering"
#AcademicLife
Note that the paper is about applicants, only some will be successful and they will be very welcome, but it is a tiny fraction of what would be needed to compensate the loss
No they are not. Already not enough positions for the talented who graduated here
Ah it might be a data artefact, as if the location is sent much less frequently at some point.
Just found this paper, from this author, which seems to be for this study (this is really not my area of expertise)
www.researchgate.net/publication/...
⚛️ @kylecranmer.bsky.social , have you seen this brewing ?