@binarystate@mastodon.coffee Producer of research, software and odd sounds. Currently based at Rutherford Appleton Lab, Harwell. Opinions my own. (He/him)
UK
Joined May 2010
- Tweets2.1K
- Following438
- Followers363
- Likes3.7K
I want the figures wherever LaTeX puts them. It has better layout judgement than vast majority of scientists.
Adam Jackson is tooting on the other network retweeted
I will be attending the #SOOCON24 and #stateofopencon and speaking about my work. I will present preliminary research on applying the Do No Harm framework to open source/science interventions.
Grab a ticket - limited free tickets are also available. stateofopencon.com.
Adam Jackson is tooting on the other network retweeted
I asked Stable Diffusion XL to draw a hot-desking office in Hell and most of these look like UCL offices and some look better than UCL offices.
Maybe it's the sleep deprivation but I don't really understand term 3 of the T&C:
3) Additional Terms. You understand that we may present you with additional terms and conditions ... including on the JetBrains website ...
Can someone explain what this one means?
Adam Jackson is tooting on the other network retweeted
Applications now open for our next #Neutron Training course in May 2024. This is a practical instrument-base courses for postgrad + post-doc researchers who have little or no experience but whose future research program will make use of neutron techniques. bit.ly/3MuB6Xg
Yes one can reimplement from the paper but it delays community benefit, repeats effort on implementation "gotchas" and if code *is* available means wading through complex/irrelevant interfacing details. 2/2
It's so frustrating to see new publicly-funded methods published as a de-facto "exclusive" to one package when it isn't technically necessary. Something has gone wrong with the incentive structure. 1/2
To be clear, by "explanations" I mean "what does this paper say", "what does this figure mean", etc. I think that's a more exciting, open and democratic kind of Sci Comm than the usual "summarise this field in a few easy sentences that make it sound like one man did it."
But then I also see very highly liked/retweeted/viewed posts on LK99 that are much lower quality. Something is incentivising that. The cultists are haunting "For You", not "Following".
The issue is more that there's just a lot of rubbish in the timeline. I've found the *scientist-led* comment threads re: LK99 to be quite nice, mostly. Plenty of viewpoints grounded in reality, and interested public asking for (& receiving) explanations. 1/2
Differing viewpoints of scientists on science stuff, yes. Differing viewpoints of cultists on unrelated issues, not so much.
Here's a cheeky little blog post about bad tests from a @SoftwareSaved event a few years ago! "Five failed tests for scientific software" software.ac.uk/blog/2018-05-…
A _practical_ difference is that when a "unit test" fails it should make it pretty clear where the problem lies and what internal expectation/promise is being violated. Whereas "integration" tests check one path through a lot of code; they are easier to write but harder to fix.
70-80% is a good sign. I am slightly wary of code with 95+% coverage as there is a danger Goodhart's Law is at work. It is possible to write tests that technically execute the code branches but don't really stress the problem areas well.
Adam Jackson is tooting on the other network retweeted
The more I learn about devops, the more sure I am that networking computers was a huge mistake.
Yup, I remember this one whenever I meet a new cat
xkcd.com/231/
Adam Jackson is tooting on the other network retweeted
These ten recommendations from @DrHorrell apply to all neutron users as well 😇😉
#beamline #scientists
Import recommendations that probably apply to all synchrotron users. Remember: #beamline #scientists are your friends! #IUCr2023 @DrHorrell
