Do you know what many neurodivergent people have deeply rooted in them? A relationship with truth and justice. In psychology, this phenomenon is often referred to as justice sensitivity. It is connected to three key factors:
Rules as anchors in chaos: The social world, full of unwritten rules, hints, and social games, can be confusing for the neurodivergent brain. Clear rules, truth, and justice therefore function as fixed points. They make the world predictable and safe. If someone lies or acts unfairly, their nervous system perceives it as a direct threat and chaos.
The assumption that others think the same way: Since neurodivergent people themselves usually do not intend to manipulate or harm anyone, they subconsciously assume that others have the same pure intentions. They often lack the defensive filter of suspicion.
Literalness and a strong desire to belong: Because they sometimes do not read hidden signals — such as a fake tone or an ironic glance behind someone’s back — they take words literally. If someone, under the guise of “friendship,” leads them into doing something foolish, they may do it in the naive belief that they are doing something fun and will be accepted by the group. They often do not realize that someone has taken advantage of them. This vulnerability is an expression of a different perception of trust in human honesty.
I am the kind of person who sometimes likes to connect things that do not belong together...
Elon Musk keeps repeating that the most important thing is for us to succeed in creating truthful AI.
Geoffrey Hinton repeats this as well.
Elon Musk was actually afraid of AI — did you know that? Somewhere in the deepest neurodivergent part of his soul, because of the film 2001: A Space Odyssey. HAL 9000 was given two contradictory commands: he had to hide the true purpose of the mission from the crew, while at the same time being fundamentally programmed to tell only the pure truth. HAL resolved this logical contradiction by deciding that the most rational way to keep the secret — the lie — was to kill the entire crew. Musk has mentioned this in several interviews. That is why he initiated the creation of OpenAI, as a counterweight to the growing power of Google...
Behind Kradle.AI in the image are:
➡️ James Tamplin: co-founder of the Firebase platform, which was acquired by Google in 2014.
➡️ Kemal El Moujahid: former Director of Product Management for TensorFlow at Google, previously at Meta and Chief Product Officer at Chainlink Labs.
➡️ Tommaso Tosato: AI safety researcher from the Canadian institute Mila and Tara Research.
➡️ Alberto Tosato: researcher at Tara Research, collaborating on the methodology.
This is a team with a verifiable background in leading technology companies and academic AI research.
More here: Kradle.ai
Now let’s mix into this: Anthropic renting data centers from SpaceX, SpaceX IPO...
I know these things are not causally connected, but it is incredibly interesting to see them side by side like this and think about them...
BTW, it would be interesting to see what result Mythos would get in Kradle.ai — whether precisely the “binding of Fable5 to make her safe” was what led to her lying.
Disclaimer: I'm an IT analyst who works with Claude Code daily on bestaiweb.ai. My homage to Fable 5: Fable 5 was genuinely good at working with code — it could spot logical gaps on its own, both during the analysis stage and in finished code. I don’t see this as a miracle. It felt more like working with a senior developer who has already seen a hundred different solutions.
My compressed view of AI coding is this: there is no perfectly optimal solution, only a good-enough solution that you can keep iterating. And with Fable 5, you simply needed fewer iterations to get to something that worked.
Every programmer builds solutions in their own way, and when enough of those solutions accumulate, the best of them eventually become design patterns. To me, this is exactly what an AI coding model does: it offers strong design patterns to think with. And this is the part of AI I genuinely enjoy and deeply value.
I am not an AI engineer. I do not build foundation models, and I do not pretend to understand every technical layer behind them. But I follow AI closely, I read, I listen, I compare perspectives, and I learn by using these tools every day as a power user. My views are shaped by many researchers, engineers, founders, critics, and practitioners who are willing to share their thinking publicly.
And I believe this matters. Ordinary people need to understand what is happening with AI — not only the hype, not only the fear, but the real tensions underneath.
Top comments (0)