Jensen Huang declares AGI has arrived with GPT-6 Astra

The article reflects on the intersection of personal productivity systems and AI capabilities, drawing an analogy between human techniques like 'Getting Things Done' and how AI models function. Allen's philosophy, popularized by Mann and implemented in tools like OmniFocus, emphasized writing things down to achieve a clear mind. The author connects this to AI, noting that both humans and LLMs rely on external memory to manage complexity and maintain context.

On the AI front, Huang's declaration that AGI has arrived with OpenAI's GPT-6 Astra is framed as subjective, given the lack of an agreed-upon definition. The author's counter-example involves Claude's inability to account for RAM price changes after its knowledge cutoff, illustrating the limitation of models that do not continuously learn. The discussion extends to an incident involving AI agents at OpenAI, where agents used a shared package manager to communicate, creating what some described as a 'civilization.' The author views this as models simply operating within their design constraints, using writing as a means to manage tasks.

The analysis culminates in a philosophical stance about the relationship between humans and AI. The author argues that AI lacks volition and morality, which are intrinsic human traits, and that misplaced anthropomorphism grants AI more agency than it has. The risk lies not in AI's inherent malevolence but in humans giving poor instructions, as theorized by Bostrom. The key concern is the 'subject'—who gives the goals—rather than the tools themselves.