Snorre Fagerland
@snoffle.bsky.social
Dad. Husband. Threat researcher at RSAC. He/Him. Maintainer of the Virus History Project.
#infosec #cats #nature #science #art #antifascist #drawing #istandwithukraine
#infosec #cats #nature #science #art #antifascist #drawing #istandwithukraine
Wonderful; just the sort of use it's intended for. I aim it to be as complete as possible, though there are still holes, unfortunately.
(Creeper on ARPANET/TENEX comes to mind grrrr...)
(Creeper on ARPANET/TENEX comes to mind grrrr...)
November 11, 2025 at 5:06 PM
Wonderful; just the sort of use it's intended for. I aim it to be as complete as possible, though there are still holes, unfortunately.
(Creeper on ARPANET/TENEX comes to mind grrrr...)
(Creeper on ARPANET/TENEX comes to mind grrrr...)
wow does the same thing. pet fights.
November 5, 2025 at 1:04 AM
wow does the same thing. pet fights.
Ja, lar meg gjerne overtale her, hvis det er betydelige trafikkbehov allerede tilstede (eller planlagt).
September 15, 2025 at 6:50 PM
Ja, lar meg gjerne overtale her, hvis det er betydelige trafikkbehov allerede tilstede (eller planlagt).
I'm not saying this is the ideal future of computing. But this is the future we seem to be headed towards. /end
August 4, 2025 at 11:35 PM
I'm not saying this is the ideal future of computing. But this is the future we seem to be headed towards. /end
Benefit?
- Ultra fast development/improvement cycles
- No translation overhead
- Self-optimizing
- Realtime resource optimization
- Autocorrecting for context changes
- No human programming needed, save for goals and boundary definitions
- Ultra fast development/improvement cycles
- No translation overhead
- Self-optimizing
- Realtime resource optimization
- Autocorrecting for context changes
- No human programming needed, save for goals and boundary definitions
August 4, 2025 at 11:35 PM
Benefit?
- Ultra fast development/improvement cycles
- No translation overhead
- Self-optimizing
- Realtime resource optimization
- Autocorrecting for context changes
- No human programming needed, save for goals and boundary definitions
- Ultra fast development/improvement cycles
- No translation overhead
- Self-optimizing
- Realtime resource optimization
- Autocorrecting for context changes
- No human programming needed, save for goals and boundary definitions
How do you debug or even KNOW what happens in a self-governing, code-generating, self-modifying system at machine speeds?
How do you constrain it? And where do you intervene?
How do you constrain it? And where do you intervene?
August 4, 2025 at 11:35 PM
How do you debug or even KNOW what happens in a self-governing, code-generating, self-modifying system at machine speeds?
How do you constrain it? And where do you intervene?
How do you constrain it? And where do you intervene?
It would for sure require new silicon designs and likely instruction sets to adapt to higher level semantics; eg CLASSIFY or TRANSFORM instead of MOV or JMP.
And a whole host of new problems, concepts and challenges.
And a whole host of new problems, concepts and challenges.
August 4, 2025 at 11:35 PM
It would for sure require new silicon designs and likely instruction sets to adapt to higher level semantics; eg CLASSIFY or TRANSFORM instead of MOV or JMP.
And a whole host of new problems, concepts and challenges.
And a whole host of new problems, concepts and challenges.
Enter AIOS. You give goals, it creates, tests, refines and schedules its own code for it at machine speed.
It would run AI-created code sequences, not executables. It would do scheduling and memory management based on learned patterns, with AI-handled exceptions and self corrections.
It would run AI-created code sequences, not executables. It would do scheduling and memory management based on learned patterns, with AI-handled exceptions and self corrections.
August 4, 2025 at 11:35 PM
Enter AIOS. You give goals, it creates, tests, refines and schedules its own code for it at machine speed.
It would run AI-created code sequences, not executables. It would do scheduling and memory management based on learned patterns, with AI-handled exceptions and self corrections.
It would run AI-created code sequences, not executables. It would do scheduling and memory management based on learned patterns, with AI-handled exceptions and self corrections.
We already have vibe coding. Most of that don't work well yet.
BUT IT WILL, 5 or 10 years from now. And then it's time to ask: Why is there a human in this loop? Source code is how humans tell computers what to do. When computers make code for computers, people are sand in the machinery.
BUT IT WILL, 5 or 10 years from now. And then it's time to ask: Why is there a human in this loop? Source code is how humans tell computers what to do. When computers make code for computers, people are sand in the machinery.
August 4, 2025 at 11:35 PM
We already have vibe coding. Most of that don't work well yet.
BUT IT WILL, 5 or 10 years from now. And then it's time to ask: Why is there a human in this loop? Source code is how humans tell computers what to do. When computers make code for computers, people are sand in the machinery.
BUT IT WILL, 5 or 10 years from now. And then it's time to ask: Why is there a human in this loop? Source code is how humans tell computers what to do. When computers make code for computers, people are sand in the machinery.