• 0 Posts
  • 33 Comments
Joined 2 months ago
cake
Cake day: April 17th, 2026

help-circle


  • I thought, LLMs would never become able to write code. And now, I use Claude Code as the always available senior on coke.
    LLMs have a reliability problem. If that gets solved somehow, they can actually drive a worker bot - or a terminator.

    And the big money pits also don’t only do LLMs. Those just get all the press because they are usable by normal users right now. Of course, some of those money pits are just investor scams. It’s a fully corrupted society after all.



  • The thing, western governments fear is AI-powered terminators. They want the tech first, so they can win the war when someone attacks them. That is the arms race part.

    The unemployment explosion is obviously also happening. But that’s actually a pretty good thing in the long run as a society with 90% unemployment and the need to work to live is absolutely unsustainable. AI will basically force the end of capitalism by increasing the system’s volatility until it adapts.



  • That will not work long-term. When prices are high for a long time, it becomes more and more attractive for governments to start their own foundries for economic and strategic reasons. And while that is not easy, China is on its way, ASML and Zeiss actually are European companies, and no one starts at zero because there are a ton of patents which already expired or expire soon.
    A fully industrialized nation which really wants to make chips can make chips. Making the best chips is pretty darn hard, but making the chips from a few years ago is doable for China and the US right now, and the EU in ten years.
    On a bloc level it makes sense to have your own foundries independent of foreign influence just for military and infrastructure reasons alone.




  • In my experience with Claude Code Opus, analyzing or generating files eats the most tokens while thinking is actually surprisingly cheap. I guess, the token-counting is somewhat wrong on purpose to incentivize use of high-effort thinking mode because when you incentivize using lesser models or modes, people get disappointed by the output quality and stop using the service…
    So just let it analyze the code base for flaws and bugs in a loop using lots of sub agents for each type of bug or code smell.

    The good thing about that method is that it is technically malicious compliance; but it also offers a high degree of plausible deniability and likely yields some actual bug fixes to offer as justification.

    Doing it just once every once in a while without fanning out into tons of agents rereading the same files is the non-malicious-compliance way of using AI for bug hunting and usually worth it. Also let it write tests for the found bugs (and properly review those tests using the natural neural network in your head).










  • Computing was never as great as it is now. Never before did we have so much free open source software at such a high level of quality to use and tinker with. Never before was it this easy to find help for the most obscure problems. Never before was playing on Linux a viable option.

    But online and offline social networking really got enshittified a lot. 3rd spaces online like offline are fully commercialized. Online, everything is remembered and if it can be used against you, it will be used against you either by the government or some rando just because they can.
    But you can use a VPN and as many pseudonyms as you need to properly separate your community-specific personas. And if you live in a city of religious fanatics, the internet also is a great to find like-minded people. There are communities for everything.