Analyst Ben Thompson of Stratechery suggests that the strategic danger Meta Platforms Inc (NASDAQ: META)'s Muse presents to frontier AI labs like OpenAI and Anthropic might be more significant than previously thought. Thompson has called Muse "the best and most approachable personal agent product I have used, winning by a clear margin." His central observation is that while Meta's Muse Spark 1.3 model is not the most advanced, it is already sufficient to power a compelling AI agent.

"This is precisely the bearish signal," Thompson wrote. He argues that AI model capabilities may have reached a point where companies can build sticky consumer products without needing to lead the model performance race.

The Importance of Muse

Thompson believes the moat for personal agents could be notably stronger than that of chatbots. Once users feed their personal information into an agent and weave it into their daily routines, switching to another service becomes far more difficult. "Model capabilities are now sufficient to build attractive products that possess a genuine moat," he wrote.

This dynamic could challenge the strategies of frontier labs, which have heavily invested in constructing increasingly powerful models while moving toward controlling the user interfaces and the applications built on top of them. Thompson also notes the AI industry is trending toward a separation between models and the "harness," the software layer that turns a model into a useful agent. Both OpenAI's Codex and ChatGPT, as well as Anthropic's Claude Code, exemplify this agentic path, yet the programming outputs they generate can ultimately work with rival models.

Agents Becoming the Main Battleground

This argument arrives as AI agents emerge as the industry's next major arena. Both OpenAI and Anthropic are advancing agent tools designed for multi-step tasks, particularly coding and computer-use workflows. Thompson has previously stated that agentic AI is potent enough to justify the enormous demand for compute infrastructure. However, Muse indicates that value may be shifting increasingly to the side that controls the end-user product, rather than the one holding the most powerful underlying model.

Traders Eye Meta's AI Moves

Meta shares climbed over 11% on Monday, adding nearly $200 billion in market value, after the Muse app topped U.S. download charts on both iPhone and Android. Muse can handle multi-step tasks, including browsing websites, filling out forms, and executing actions on behalf of users. The app's launch also triggered pushback from platforms, with Amazon blocking Muse from accessing its site.

One trader noted that Monday's move in META stemmed from "an early signal that its AI agent is working. A week ago, the market was pricing in uncertainty about whether AI development would slow. This is what repricing expectations looks like. The agent is barely generating revenue yet, but the market is moving on what it could become."

Muse Could Deepen Meta's Understanding of Users

Muse might give Meta a far richer understanding of user interests and behaviors than ever before, which could strengthen its advertising and recommendation systems. Instagram head Adam Mosseri recently said Meta's systems have historically relied heavily on signals like "what users like and what people with similar interests are engaging with." In a July appearance on Lenny's Podcast, he stated, "Only now are we truly reaching the level of granularity people have assumed we had for years."

This situation could shift further as users spend more time with personal AI agents like Muse that remember conversations, preferences, and life details.