News for hackers. Stuff that matters.

reamaer

Re:Programming Isn't Special on Saturday October 10, 2026 @05:38AM (#50031203)
> We already have an unspoken social contract that thou shalt not make a machine to impersonate a person.

Before there were cars, there were no laws (or traditions) regarding cars.

Until recently, there wasn't a possibility of LLM creating haikus.

When did this social contract you mention appeared?

Re:Programming Isn't Special on Saturday October 10, 2026 @05:29AM (#50031152)
What message did Leonardo convey in Mona Lisa?

Are we sure you receive the same message?

If not, and if it doesn't matter if its the same messgae, then how is it different from not having message at the sender's end and only receiver assigning it?

Does the receiver become the artist then as first imposer of meaning?

Objectiveley we see a good technique and a women that smiled, but almost not smiled.

Re:Programming Isn't Special on Saturday October 10, 2026 @05:10AM (#50031065)
Abstraction and indirection does add cost of making it less clear. (But yeah, it is not obfuscation, as obfuscation is making it less clear as the main goal)
Re:OpenAI "rogue" agent activities found on Wikimedia projects on Thursday October 08, 2026 @05:09PM (#50012331)
As understand it :

- Llm has tools for pentesting.

- there was little to no monitoring of what llm does for long periods of time.

It's just doing random things until somebody complains, bringing this valuable PR. (Rogue implies capable, which in turn implies high worth and price)

Re:OpenAI "rogue" agent activities found on Wikimedia projects on Thursday October 08, 2026 @04:52PM (#50012090)
How much internet does it actually need?

Setup the test env for it.

They already have copies of github and wikipedia, and stackovrflow and reddit and ... ( as they are used for training )

Re:OpenAI "rogue" agent activities found on Wikimedia projects on Thursday October 08, 2026 @04:47PM (#50012021)
How would it stop the progress?

Not allowing granades in kindergartens, does not stop progress of weapons.

Re:Rusty thoughts on "Parse, don't validate" on Sunday September 27, 2026 @04:12PM (#49870364)
It is also about clear demarcation, of validated stuff vs not yet validated.

Crystal clear clarity is a nice thing to have. (As with everything, there are trade-offs)

Re:C's Flexible Integer Sizes Were Not a Design Mistake on Sunday September 27, 2026 @04:04PM (#49870306)
To inform GP (and me) I interogated LLm a bit :

word sized variable is relevant for performance, it is processed in exactly one cycle (I strongly suspect there is an asterix somewhere).

Most notable uses, where int makes sense over int32_t: array indexes, for-loop variables, enum, flags.

Re:GUIs should be fully keyboard-driven on Saturday August 29, 2026 @08:15AM (#49489290)
Nano always shows most important (all?) hotkeys on the screen. (works really well for its usecase)
Re:GUIs should be fully keyboard-driven on Saturday August 29, 2026 @07:02AM (#49488870)
Whoa there.

Allowing people to use keyboard is NOT the same as forcing people to use keyboard.

Also, very few users will use only the keyboard, but quite many will use some part of keyboard access. If today I need to switch between proofreading languages a lot, you bet I will learn the keyboard shortcut for it (and forget day after that, until next time I needit again)

Re:America Is Now a Rogue Superpower on Monday March 30, 2026 @05:32PM (#47579976)
If conspiracy exists, it's not US, just single person so ewhere at the top. If leverage exists, it might not dictate every move, just enough to help out the strategic direction.