physicsguy

5 ms·

61.
by physicsguy 4mo ago
It's still better than it was 10 years ago (compiling everything from source wasn't uncommon even then!) but it's painful. Comparisons to other interpreted languages aren't really fair though, if JS/TS had as many c
62.
by physicsguy 4mo ago
`uv` is great but the biggest issue with Python packaging right now continues to be getting it right for scientific and ML packaging. Want to install PyTorch? which one? CUDA? Oh, OK, then you have to get the wheel directly from them becaus
63.
by physicsguy 4mo ago
Go is terrible for scientific/ML work though, the libraries just aren't there. The wrapping C API story is weak too even with LLMs to assist. Try and write a signal processing thing with filters, windowing, overlap, etc. - there&#
64.
by physicsguy 5mo ago
For e.g.: https://arxiv.org/abs/2603.04337 https://arxiv.org/abs/2603.05607 https://arxiv.org/abs/2605.01171 For a more detailed review: https://github.com/li
65.
by physicsguy 5mo ago
This is over a year old now, why post now? There have been other advancements in this area.
66.
by physicsguy 5mo ago
What I've seen in an AI-forward looking environment is that it's much more common for PM/POs to be knocking up at least a UI prototype now, and experimentation is happening often even before writing the tickets. Similarly whe
67.
by physicsguy 5mo ago
Most of the scientific libraries of note originate in academia where MATLAB is effectively free to users. The cross over to Python was well under way by ~2014
68.
by physicsguy 5mo ago
Plus NVIDIA clocked that it was also the developer library ecosystem and even now there just aren’t equivalents. The AMD rocFFT library wasn’t even complete compared to FFTW until very recently and cuFFT did that more than a decade ago
69.
by physicsguy 5mo ago
The people that write the libraries care, why do you think Python is where we’re writing ML code and not MATLAB?
70.
by physicsguy 5mo ago
Cython and PyBind and Nanobind are good for wrapping an existing library written in C++ and crafting an interface that doesn’t feel like it’s a C++ one. They were a big step from ctypes and SWIG
71.
by physicsguy 5mo ago
It also massively changes the value economics of the frontier models. In a lot of cases, you really don't need a general purpose intelligence model too.
72.
by physicsguy 5mo ago
Yes, I don't find AI generated documents are useful, they just add a ton of fluff. but it's removable fluff at least was my point.
73.
by physicsguy 5mo ago
I see a lot more Kia’s but I think it’s shifting because dealerships are shifting. The Toyota dealership near me now sells Jaecoo, Cherry, etc. and I am seeing tons of the Jaecoo SUV cars around. The Pentagon Vauxhall dealership I bought my
74.
by physicsguy 5mo ago
In my experience I'm pasting a lot more into AI to get the high level summary though.
75.
by physicsguy 5mo ago
I'm not based in London.
76.
by physicsguy 5mo ago
I'm no proponent of PCP or leasing but you have to admit there is a hassle factor invovled once they hit a certain age + mileage.
77.
by physicsguy 5mo ago
If you're at that income level, your employer pension contribution is already likely high and you've likely been stacking it for a while anyway. At some point there is a diminishing return to how much you should put in your pensio
78.
by physicsguy 5mo ago
People on higher salaries are disproportionately likely to be the ones doing it though - much much more likely to work for companies that implement the schemes for a start.
79.
by physicsguy 5mo ago
Yeah, but you're still taxed at 72% between £100 and £125k if you have a student loan (as most people in that age bracket will be), so even in that case the hit to your take home isn't that much.
80.
by physicsguy 5mo ago
The biggest seller of EVs here is the salary sacrifice schemes that give a huge discount to high earners, especially those with kids. Imagine you're on taxable income of £120k and have two chidlren in nursery. Currently you get no help
81.
by physicsguy 5mo ago
They changed their major release numbers too tbf. 4.x it was point release per year, now it's a major release per year.
82.
by physicsguy 5mo ago
> For a brand new app with no prior context, they move fast. But here’s what they’re all doing: making the most statistically common decision for your situation, not the best one for your situation. They don’t have your context. And your
83.
by physicsguy 5mo ago
Especially when it was £20 to move the whole family (kids were free) to Australia. My great uncle and aunt did just that, after national service he was in business working for Walls ice cream. Took himself and two kids off in early 60s, wer
84.
by physicsguy 5mo ago
We used bitbucket for years in our academic research group because GH didn't allow private repos. But otherwise it was better to use, so when they allowed that, we made the switch.
85.
by physicsguy 5mo ago
There have been critical vulns in nginx-ingress (the part which is deprecated) like this: https://kubernetes.io/blog/2025/03/24/ingress-nginx-cve-2025... If you're using it after it's dead, you
86.
by physicsguy 5mo ago
It's also difficult for data pipelines or data intensive things. At several companies we've run into the "Need to put ML model behind API and pods get killed because health checks via API are basically not compatible with con
87.
by physicsguy 5mo ago
Til there's a security issue, right? Nginx is a big target.
88.
by physicsguy 5mo ago
My current company has company managed boards, 6000 devs and we have about 250 custom fields. I work in a research team and we only need Kanban and I can't change the issue type if something is created. Hell.
89.
by physicsguy 5mo ago
I was joking.
90.
by physicsguy 5mo ago
As a British person, this is confusing since the game doesn't involve a sausage roll!

More ›