The Pathway to AGI Isn't Intelligence, It's Shared Cognition (blog.reframetech.com)

The author argues that the real bottleneck on the road to useful, collaborative AI isn’t higher benchmark scores or ever-smarter LLMs, but continuity—today’s models are “idiot savants with amnesia,” brilliant in the moment but unable to carry goals, reasoning, or institutional memory across sessions, tools, and teams. That lack of persistent context forces repeated framing, wastes time, and limits ROI for knowledge work. Framing AGI as only an intelligence problem, the piece contends, is a red herring; what people really want are systems that compound context over time and participate meaningfully in ongoing human workflows. To solve this, the author proposes a “Cortex Layer”: open infrastructure for shared cognition that makes context first-class, structured, transportable, secure, and human-owned. Key technical elements include Streams (versioned, addressable units of meaning capturing goals, decisions and rationale), log-anchored and cryptographically signed records, local-first sync, and an open Agent Shared Cognition Protocol (ASCP) intended as an IETF-style standard. The implications for AI/ML are big: move focus from isolated model improvements to interoperable system design, enable auditable multi-agent collaboration, preserve institutional knowledge across tools, and make agent-assisted workflows reliably composable and verifiable.