We rewrote Prime Agent in Rust with 5x faster cold start, 8x less memory, 2x streaming throughput, and full behavioral parity. Prime Agent is a harness designed from the ground up for long-running agents. The RLM architecture allows many agents to coordinate across a session that outlives any single context window, and is what enabled 2,000+ agents to rewrite the harness they were running on.

concentration of power is the outcome but I am convinced many of them sincerely believe what they profess and don’t see the second-order effects of their beliefs (due to self-delusion, naivety, compartmentalization) the sincerity with which they believe the Wrong People controlling this technology could lead to mass extinction is the most dangerous part, because it grants them permission to do anything they think could prevent it, even horrific things sincerity is what drove Robespierre and the architects of the Terror

A nice thing about my industry is that a lot of people who grew up poor or squarely middle-class or didn’t go to an elite college now enjoy really nice lifestyles and are discovering a lot of consumer goods and experiences for the first time I think we should celebrate this

This was a huge push by our research team My favorite finding: direct A2A communication, scoped to the agent’s nuclear family (parent-sibling-child), created cooperative behaviors between subagents. This challenges the common assumption that subagents are stateless function calls and that an orchestrator needs to mediate everything.

A personal update: I’ve joined Prime Intellect to build out our marketing organization. Since moving to San Francisco 4 years ago, I've noticed my conception of time warping. People here often borrow the sentence, "there are decades where nothing happens; and there are weeks where decades happen" to describe the time compression caused by AI breakthroughs. Sometimes it seems we're irresistibly carried along a long arc of exponential progress. While not quite linear, the contemporary story of AI presents technological progress as a jagged frontier punctuated by thresholds, with its benefits inevitably diffusing to the masses. But civilizational progress doesn't neatly follow technological progress. If anything, the history of institutions is much more cyclical, not linear or exponential -- an infinite wheel of birth, growth, decline, and rebirth. And for all the talk of imminent/immanent eschatology, AI incumbents have failed to tell a story where most of humanity benefits from AI; what their words and actions increasingly point toward is a future where wealth and power are concentrated in the hands of a few organizations. But things don't have to be this way. Intelligence can be something people own and build upon, rather than rent from a handful of companies. Over the past 18 months, we've found that open-weight models, post-trained with specialized data, can surpass the capabilities of closed-source models. While this threatens centralized AI incumbents, it also means that eventually, every company can become an AI lab, and every individual can become an AI researcher. This is Prime Intellect's project: enabling sovereign intelligence through an open frontier AI stack. Now, you can own your intelligence. Your experience becomes your advantage. If this is a compelling vision to you, hit me up! We're looking for ambitious teams to build the new guard of open AI with us.

btw when she went to the hospital for her symptoms she was told it was all in her head and she was exaggerating her symptoms to gain attention and thats what motivated her to do this research and no traditional science lobby funded her so she relied on private funds/donation