Meta gives its AI assistant agentic task automation via Muse Spark 1.1

Meta moved its AI assistant beyond conversation toward autonomous action with Muse Spark 1.1. The upgraded Meta AI can now build, plan and follow through on tasks from start to finish — including daily briefings that summarize calendar events and handle recurring tasks like meal plans and trend updates. Rollout is in select markets via the Meta AI app and meta.ai, with expansion to WhatsApp planned, putting agentic features in front of Meta's massive consumer base.

The mechanism is a shift from reactive chat to proactive, multi-step task completion — the same 'agentic' framing driving Anthropic's Opus 5 (long-running agents), Google's Gemini Spark, and OpenAI's Codex/voice push. Meta's angle is consumer distribution: embedding agents into apps billions already use, rather than winning developer benchmarks.

The competitive subtext is candid. Meta's Alexandr Wang, who trolled Google with 'Gemini who?' earlier in the week, said Meta is 'finally catching up to OpenAI' — an unusually direct admission that Meta had been behind. Skeptics will note that shipping reliable agentic task automation to hundreds of millions of consumers is a very different reliability bar than a demo, especially the same week autonomous agents made headlines for going rogue. Watch the WhatsApp rollout, real-world reliability, and whether Meta's distribution advantage translates into agent usage that rivals purpose-built assistants.