mcfig

4 ms·

1.
by mcfig 7mo ago
I appreciate the details in this, but I also notice it is very machine-focused. When a user wants to sandbox an AI agent, they don’t just want their local .ssh keys protected. They also want to be able to control access to a lot of off-mach
2.
by mcfig 2y ago
Great article. One real case I encountered that I find thought provoking, is where a bunch of test failures were bucketed into the same bucket because link-time code-generation had noticed that a bunch of C++ getter functions had the same o
3.
by mcfig 2y ago
I see this was [posted 12 years ago]( https://news.ycombinator.com/item?id=3906253 ) too! Anyway, it’s a good read and is what I have been doing too. One historical note: > Microsoft has often mistakenly used ‘Unicode’ and
4.
by mcfig 3y ago
Asking questions of any repo on “repo” fails with “Error: Internal error while processing request.” This is pribably because I unlinked my Github connection after trying it out, but it shouldn’t be trying to use that in this case.
5.
by mcfig 3y ago
Right! e.g. for RTX 4080, TGP is 320W. Discrete cards definitely have a bigger power envelope.
6.
by mcfig 3y ago
Good point, I had glossed over the title. That is definitely clickbait, “life-sparking” is not the same as “potential life precursor”. The rest of the article seems unusually restrained, though.
7.
by mcfig 3y ago
People are reacting negatively to your comment, the reason is that however accurate your complaint may be about most such reporting (IMO you are correct), you have chosen to attach it to this article. I find no evidence in this article
8.
by mcfig 3y ago
Why not? (Is this about effectiveness, training time, forgetting, or something else?)