ephimetheus

I enjoy programming as a hobby and have done it for many years. I also enjoy building stuff with LLMs. Why is one pure and sanctioned and the other one is bad? Why do I need other people to decide this for me?

Or it’s really just plain old gate keeping maybe.

Funny this predates even the original paper first introducing the attention mechanism underpinning modern LLM.

I’ve noticed this too. I think it started around the time of the first round of big tech layoffs in the US, and when the US ratcheted up the price for H1B(?) visas.

Seems like economic uncertainty or fear of it breeds xenophobia. Who knew

sPHENIX uses software that we’ve worked on at CERN to do some of their reconstruction!

Oh yes and so much so! Electrons are point-like (not composite like a proton) and interact only electroweakly (not strongly).

There absolutely are direct neutron experiments, but they are much lower energy and have a different focus, partly because neutrons being neutral means they’re very hard to accelerate.

There’s an ultra cold neutron source at Paul Scherrer that is used to measure if the neutron has an electric dipole moment. This is complementary to high energy experiments.

I get what you’re saying, but the measurements are real. In some sense they are the truth.

In the article this refers to the finding that the quark is more complex than three valence quarks.

The measurements indicating that the three-quark-model is incomplete are overwhelmingly conclusive, so some degree of certainty in the language is warranted in my view.

Neutrons are just as complex, they’re much harder to study though.

Didn’t they just drastically increase the price for non US citizens? I think it almost went up by 10x

I appreciate the opinion, but that precludes it from being used in repos that are not jj-first and happen to use submodules.

I got into jujutsu recently for the mega merge workflow.

The only thing I’m missing now is support for git submodules, especially when working with workspaces.

This requires me to keep using git worktrees with collocated jj in each of them, which is suboptimal.

Is selecting text in the PDF supported? I can’t get that to work in iOS Safari.

So far whenever jujutsu came up I didn’t find its features that convincing relative to git.

I have to say however that the mega merge workflow seems intriguing and might in fact be a solution to having to serialize my work like I do in git for now.

I think I’ll look into that.

Exactly, I would've hoped someone could come up with a way to render markdown directly into a PDF, without roundtripping via tex and having to handhold the styling process in the way that's required now.

What always disqualifies these projects for me is the fact that they need to use a headless browser to export to PDF. PDF export is the primary feature I need from these, and it’s a shame the export mechanism is still this slow and unreliable.

This is a fantastic analogy!

Which package do you mean? Last time I tried this it was kind of clunky.

For experiment data, there is a layer on top of all of this that distributes datasets across the computing grid. That system has a way to handle replicate at the dataset level.

Paperless-ngx is fantastic! I’ve been running it for a while and it works great!

I wrote an iOS [1] app to connect to you instance and it’s open source [2].

[1] https://apps.apple.com/de/app/swift-paperless/id6448698521

[2] https://github.com/paulgessinger/swift-paperless

Doesn’t work in web views unfortunately.

iOS embedded safari doesn’t run them unfortunately.

I mainly read HN from an RSS reader, whose web view always shows light mode.

Especially meat is particularly expensive in Switzerland, however, compare to most of Europe. The marketing somehow managed to cement this a meaning higher quality, and the population largely accepts the higher price.

4000TWh of energy per year comes out to about 456GW continuous power. So pretty much in the right ballpark?

On macOS, I get the message that the disk image is corrupted.

+1 for RDataFrame for what it can do. Just be prepared to bail to C++ and for loops when you exceed what it can do without major headaches.

Yeah I think it has to do with the memberwise splitting. https://github.com/scikit-hep/uproot5/issues/38

I understand this has not been a priority so far.

It kinda works if you open a magic file with a specific on-disk representation which bypasses this, but that’s not a solution at all.

I can’t count the number of of times where a beginner did some stuff in pyroot that was horrifically slow and just implementing the exact same algorithm in C++ was two orders of magnitude faster.

If you don’t use RDataFrame, or it’s just histogram plotting, be very careful with pyroot.

All of our other code is C++. The data reconstruction framework writing ROOT files, the analysis frameworks doing stat analysis. The event data model is implemented in C++.

It has its rough edges, but you do get a lot of good synergy out of this setup for sure.

uproot to this day doesn’t properly implement reading TEfficiency, I believe, which is a bummer, to be honest.