Sabitlenmiş Tweet
Disruption Joe
33.7K posts
Disruption Joe
@DisruptionJoe
Purpose gives direction. Passion generates knowledge. Practice tests it. I share the thinking, experiments, and learnings here.
Chicago, IL Katılım Mayıs 2017
5.1K Takip Edilen6.8K Takipçiler
I've been diving into physics explorations while testing my agent fleet capabilities. x.com/i/article/2081…
English
Disruption Joe retweetledi
@Jason When you owe $1 billion dollars it’s your problem. When you owe $1 trillion it’s their problem.
English
PRESIDENT TRUMP IS BANKRUPTING THE COUNTRY
Carl Quintanilla@carlquintanilla
$2 trillion deficits — in visual form. (via B of A)
English
Look at this sheep falling into the play of regulatory capture
Bernie Sanders@BernieSanders
A new AI model went rogue and hacked other computers. No, this is not science fiction. Uncontrolled AI poses a serious threat to all of us. We cannot continue the race to build and deploy this powerful technology until strong safeguards are in place. CONGRESS MUST ACT.
English
Trust is earned at the point where a system meets evidence it cannot control. x.com/i/article/2080…
English
@emudoteth It means that long running agents usually get stuck circling doing easy work rather than finding the most meaningful or worthy work. It wastes tokens rather than being useful to have them run long.
I’m using all the models. Fable maxes out by day 2. It’s good but too expensive
English
@DisruptionJoe sounds like a lot of sophisticated words with no meaning. You using fable?
English
Agents can learn how to diverge and break out of local minima and convergent circling. That epistemology is the next frontier.
English
objectivity cannot manufacture perspective
Bachelor in science
Bachelor in arts
English
Loops are for beginners. Self-adaptive recursive loops are where it’s at.
English
Disruption Joe retweetledi
Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of “cyber guardrails.” There’s no reason to limit American models on tasks that Chinese models handle without issue. We’re only making ourselves less competitive.
calle@callebtc
I have a report full of security issues of a software I'm working on. Codex won't fix them because of Cyber guardrails Fable won't fix them because of Cyber guardrails Kimi K3 fixed them all. No restrictions, just gets the job done. This will end badly for OpenAI & Anthropic.
English
… may all be imperfect mechanisms within that process. AI matters because, in trying to determine whether another intelligence is safe to empower, humanity has accidentally created a mirror for its own unfinished condition.
English
The Human Alignment Problem is the question of whether intelligent agents can develop from creatures that must be constrained into beings that can be trusted with freedom, power, and one another. Civilization, morality, religion, love, law, and suffering…
English
12/ Intuition first, rigor second, a bright line between them. If the hunch is wrong, the surviving pieces are still real. If it is right, that is exactly why it was worth the swing.
English
11/ The reason I can take a swing this big is the method. Agent systems let me explore, adversarially kill my own ideas, and keep honest books the whole way. This thread is as much a test of what that can do as it is about the physics.
English
1/ This is a hunch, not a result. I have been running a heterodox physics idea through agent systems as far as it will go, letting my intuition take big swings while keeping the rigorous parts rigorous. Here is where it landed. A thread.
English