Who is building the agent operating system?

Dimension breakdown

• Assistant Intelligence

  7/10

  Llama is strong and improving, a notch below the very top frontier models.

• Agent Capability

  5/10

  Limited shipped agentic product; capability lives in the open ecosystem more than in Meta AI.

• Conversational UX

  7/10

  Meta AI in WhatsApp/Instagram and voice on Ray-Ban glasses make interaction ambient and natural.

• Cross-App Actions

  4/10

  Little orchestration beyond Meta's own apps.

• Personal Context

  6/10

  The social graph is rich context, but it isn't a personal-assistant memory in the productivity sense.

• Third-Party Integrations

  7/10

  Open weights mean the entire industry builds on Llama — enormous indirect ecosystem reach.

• Developer APIs

  7/10

  The Llama API and Llama Stack plus open weights give developers maximum control and portability.

• Model Strategy

  8/10

  Owns its models and an open-weights strategy that shapes the whole field; building its own silicon.

• Agent Platform Potential

  6/10

  Unmatched consumer reach, but no clear agent runtime, permissions model, or platform yet.

• On-Device AI

  7/10

  Quantized Llama runs on-device and powers the Ray-Ban glasses edge experience.

• Enterprise Readiness

  4/10

  Minimal enterprise sales motion, governance tooling, or compliance story.

• MyAGI Alignment

  5/10

  Open weights empower the ecosystem, but Meta itself isn't building a governable control plane.