Today I learned that x86 has a floating point version of the `nop` instruction, called `fnop`. It does nothing, but unlike the regular `nop` it uses the floating point unit to do nothing.
This meeting should have been a duel.
“In everyone’s pocket right now is a computer far more powerful than the one we flew on Voyager. I don’t mean your cell phone — I mean the key fob that unlocks your car.”
— Rich Terrile, JPL scientist and member of the Voyager imaging team
Apropos of a conversation I had with my mother just last night:
Reflections on (somewhat inevitable) dysfunction, or how security teams fail: https://lcamtuf.coredump.cx/blog/secfail/?3
This is the best thing I’ve read in 2025 https://haterade.substack.com/p/i-tasted-hondas-spicy-rodent-repelling
America is sabotaging its own scientific research with cutting the NSF budget from $9 billion to only $3 billion, using AI to evaluate grant proposals and shift NSF funding from universities to private companies
As a result, applications for positions as research group leaders from the US at Germany‘s Max Planck have doubled 🚀🔬
https://www.tagesschau.de/ausland/amerika/usa-forschung-kuerzungen-100.html
When I handle errors in complex projects, I like to have something like "return error_false()" instead of just "return false" because I can set a breakpoint on error_false and catch it the first time it happens. This doesn't work if it's a C project and I have a "goto error" for cleanup instead. I came up with a mildly diabolical equivalent, which I would like to curse you all to see: https://godbolt.org/z/Yjx4YoWeq
@lcamtuf community coming together to help neighbors push to prod
Barberis, E. (2025). UNVEILING MICROARCHITECTURAL SECRETS: THE SECURITY IMPACT OF CPU IMPLEMENTATIONS. [PhD-Thesis - Research and graduation internal, Vrije Universiteit Amsterdam]. https://doi.org/10.5463/thesis.948
@lcamtuf exactly. Can't argue about implementation-defined behavior if there's no implementation.
@lcamtuf because pseudocode does not bitrot as fast?
https://cacm.acm.org/research/a-few-billion-lines-of-code-later/
As a society we've come to fetishize "intelligence" over experience, skills, and acknowledgement of complexity. The issue with the people that are currently at work rapidly altering the functioning of the US government is not how intelligent they may or may not be. Regardless of their intelligence, they do not have the experience or the skills working with a system of this kind.
Would you trust those same people (with the skills they have today) to perform major surgery on you? Would you trust them to make radical modifications to an airplane you were about to board? If yes, then I don't believe we are having a rational discussion. If not, then why would trust them with the drastic changes to the operation of the US government? Do you really think a socioeconomic system of 350 million people (embedded in a larger system of 8 billion people) is simpler to manage than an airplane?
New way to get customer support just dropped
The Trump administration "anti-DEI" efforts are now trying to erase contributions of women and people of colour in cryptology, starting at the National Cryptologic Museum by covering them with paper.
https://bsky.app/profile/larrypfeifferdc.bsky.social/post/3lh637hlitc2w
