Cisco's Outshift unveils 'Internet of Cognition'
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Outshift by Cisco, led by SVP Vijoy Pandey, argues that vertical model scaling alone cannot unlock cross-domain agentic problem solving, and that a horizontal coordination layer is the missing connective tissue for "distributed artificial superintelligence"
- The unit's open-source AGNTCY project—now under the Linux Foundation—provides the connectivity layer that lets agents across companies and platforms discover one another, prove identity, and exchange messages through standardized protocols
- Mycelium, Outshift's open-source coordination layer for shared intent, raised decision-making rates from 33% to 93% across 14 scenarios in internal testing by making agents declare a goal, surface missing information, and resolve conflicts before acting
- Pandey cites research finding failure rates of 41% to 87% across seven open-source multi-agent systems, attributing the gap to architecture rather than prompting: "Without the right coordination layer, naive multi-agent setups can perform worse than a single agent."
- Continuous Agent Semantic Authorization (CASA), Outshift's open-source guardrail reference implementation, continuously checks each tool request against an agent's authorized task—addressing Pandey's claim that roughly 90% of the time an agent cannot confirm it is cleared for the job it was handed
Why it matters: With multi-agent systems failing 41–87% of the time in cited research, Outshift is positioning its Linux Foundation–hosted AGNTCY and Mycelium layers as open standards for cross-domain AI coordination—enterprises that pilot horizontal workflows now may end up defining the protocols the rest of the industry inherits.
