Login
You're viewing the social.epfl.ch public feed.
  • Linus Gasserligasser
    Nov 5, 2026, 7:00 AM

    That's interesting - the usual "breaking RSA-4096 takes centuries", but what if you optimise by waiting for better hardware? Supposing big, universal quantum computers never make it...

    Taking Moore into account, wait for 120 years, and then calculate for 3 years on the latest hardware.
    But: you also have to take into account the energy use! And this will hit even worse: for RSA-4096, taking the "Landauer floor" of the energy required for a bit-flip, you need 5 years of ALL sunlight of earth!

    💬 1🔄 0⭐ 0

Replies

  • Linus Gasserligasser
    Nov 5, 2026, 7:03 AM

    With today's hardware, just the energy consumption is 20 million years of sunlight on earth! I never thought about that...

    Asking about the "black hole computer", Claude wrote:

    > A 1 tonne black hole computer finishes in about 10^-10 s, with memory of about 10^22 bits (memory scales with M²). The catch is reading the result out of the Hawking radiation.

    Oh, more problems :)

    In case you like this kind of thoughts: claude.ai/share/2c8f7ac7-0267-

    💬 1🔄 0⭐ 2
  • Nov 5, 2026, 10:17 AM

    @ligasser Thanks, that's exactly the right way to use the bullshit machine. Excellent output, I must say. They really gave Claude a sense of humor that resonates with me. It's also nice that it mentions the Landauer limit and the necessity of #reversiblecomputing

    💬 2🔄 1⭐ 1
  • 💬 0🔄 0⭐ 0
  • Linus Gasserligasser
    Nov 5, 2026, 8:49 PM

    @simon Now I need to ask it how to do reversible computing :)

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0