Architectures for personal intelligence, semantic graphs, and AI systems built for alignment.
The model learned to propose structured graph edits (nodes, edges, splits, merges, deprecations) through a gated pipeline, safely isolated in a Speculative → Probationary → Main architecture.
At Thumos Care, we've always believed that AI should augment, not replace, the doctor-patient relationship. Today, we're excited to announce a major step forward in that vision: Clinician Personalization.
Announcing major updates to Thumos Care's AI-powered health optimization platform When we first built Thumos Care, we had a simple vision: help people understand their health data and get actionable guidance. Users could upload their blood work, and our AI would analyze the results.
Most language models learn reasoning implicitly from internet text. They absorb patterns of what sounds convincing without explicitly modeling the structure of scientific arguments.
I. What I Know to Be True In 2016, I discovered financial and operational connections between D-Company—a transnational criminal syndicate—and Tehreek-e-Labbaik Pakistan (TLP), an ultraconservative political party whose supporters have publicly celebrated extrajudicial killings on false accusations
I am writing this publicly because I have reached a point where I cannot carry this alone anymore. I have spent years trying to make sense of what has happened to me — while building advanced AI projects, winning hackathons, and applying to more than 600 jobs — yet I remain unemployed, isolated, and
Generic LLMs can't: Calculate voting correlation across 178 UN resolutions Track how 184 countries voted on specific issues over 30 years Find speeches that reference UNSC Resolution 1325 with sub-second latency and full context Analyze temporal evolution of diplomatic positions using graph relation
There are people who may resent your country for its role in spreading Wahhabism across the Muslim world, but most of them remain unaware of the extensive efforts you've undertaken to reverse such policies. This transformation is not...
As you embraced the papacy upon your election, facing a historic crowd at a historic time, the expressions on your face indicated overwhelm. With your lips sealed, you pulled in three strong vortexes of air through your nose, before...
As artificial intelligence gets exponentially better, disharmony in human intelligence would result in exponentially greater consequences due to the leverage provided by the former.
Achieving true superintelligence requires generative AI models capable of deeply modeling both human and non-human agents. While state-of-the-art generative models like GPT-4 can already parse emotions, sentiments, and psychological...
The search for next generation artificial intelligence is on with several outstanding teams pursuing diverse visions for the development of superintelligence. But it is already here.
Large Language Models (LLMs) are incredibly powerful, yet they face a major limitation: a fixed context window that restricts how much information they can "remember." This forces developers to carefully choose what to include in each...
This post about an exciting experiment I worked on last year: "House of Wisdom" --an AI-driven application designed for contextually rich conversations grounded in books, research reports, and laws, with answers accompanied by Mermaid...
How should democratic societies regulate artificial intelligence while preserving innovation and protecting civil liberties?
Revisiting Husserl and Merleau-Ponty's ideas about embodied experience in an era of virtual reality and digital interfaces.
Researchers at Google DeepMind and Stanford University published a paper this year demonstrating how computational software agents can simulate believable human behavior. They created a sandbox environment, inspired by a video game called...