Hi! 👋 Here's our #introduction.
We're BBC Research & Development; we explore and test new technology to discover how the BBC can best make use of it in the future.
For 100 years our engineers have been at the forefront of developments in broadcasting.
We're now researching how everyone could get TV & radio via the internet – along with all the flexibility and creativity that brings.
5G, AI, next-gen audio, UHD, personal data… we are investigating all these – and more!
https://www.youtube.com/watch?v=rzztGFXYR1Y
reminder to myself. in order to plumb #kernel #config file for a particular #buildroot build:
make linux-patch
# edit output/build/<linux directory>/.config
make # continue with the build
code generation with AI. i do not get it and it has enormous legal risks for commercial use. i’d wait at least to wait for precedent legal cases before seriously using it.
also even when manually written code paths, which in some level address the functionality in a an software package, it will take about 5-10% of the overall project time. 90-95% of the time goes into QA and finding out ways of fitting the implementation to the constraints defined by all sorts of non-functional issues.
that’s why case tools never succeeded, and that’s why i’m skeptical of the whole idea of AI code generation. there’s always someone offering a magic wand to write the code for you but it never works - at least if you aim for the highest possible quality, i.e. aim for offering more value for customers than your competitors.
it would be cool if #rust #cargo had a method of generating self-contained binary that you could copy to a device and it’d do the same thing as running cargo test
on the host. cargo could select the code paths for the binary based on parameters that you would otherwise put after, such as --features