Meta Platforms Inc.’s Muse could pose a bigger strategic threat to frontier AI labs such as OpenAI and Anthropic than previously estimated, according to Stratechery analyst Ben Thompson.
In his latest column, Thompson called Muse “by a significant margin, the best and most approachable personal agent product I have tried.” His key observation: Meta’s Muse Spark 1.3 model is not state-of-the-art, yet it is capable of powering a compelling AI agent.
META stock was up 0.5% in early premarket trading on Tuesday.
Why Muse Matters
Thompson believes personal agents could have a much stronger moat than chatbots. Once users feed an agent their information and incorporate it into everyday life, switching to another service becomes considerably harder.
“Model capability is good enough that compelling products — products that actually have moats — can now be built,” he wrote.That could challenge the strategy of frontier labs, which have invested heavily in building increasingly powerful models while moving toward owning the user interface and applications built on top of them.
Thompson also argued that the AI industry is moving toward a separation between models and “harnesses” — the software layer that turns a model into a useful agent. OpenAI’s Codex and ChatGPT and Anthropic’s Claude Code are examples of this agentic approach, but their coding artifacts can ultimately be used with competing models.
Agents Are Becoming The Battleground
The argument comes as AI agents emerge as the industry’s next major battleground. OpenAI and Anthropic have both pushed agentic tools that can perform multi-step tasks, particularly coding and computer-use workflows
Thompson previously argued that agentic AI is powerful enough to justify enormous demand for computing infrastructure. But Muse suggests the value may increasingly accrue to whoever owns the end-user product, rather than whoever has the most capable underlying model.
Traders Watch Meta’s AI Push
Meta shares rose over 11% on Monday – adding nearly $200 billion in market capitalization – after reports said the Muse app topped download charts for iPhone and Android in the U.S.
Muse can handle multi-step tasks, including browsing websites, filling forms and carrying out actions on users’ behalf. The launch has also triggered pushback from platforms, with Amazon blocking Muse from accessing its site
On Stocktwits, the retail sentiment for META was ‘extremely bullish’ on early Tuesday.
Monday’s move in META is “on early signs its AI agent is working. A week ago the stock was priced on whether AI development would slow down. This is what repricing expectations looks like. The agent barely generates revenue yet. The market moved on what it might become,” said a trader.
Muse Could Give Meta A Deeper Read On Users
Muse could give Meta a far richer understanding of users’ interests and behavior, potentially strengthening its advertising and recommendation systems.
Instagram head Adam Mosseri recently said Meta’s systems have historically relied heavily on signals such as what users like and what other people with similar interests engage with. “Now, only now are we actually getting as sophisticated as I think people have assumed we’ve been for many years,” Mosseri had said during Lenny’s Podcast in July.
That could change further as users spend more time with personal AI agents such as Muse, which can remember conversations, preferences and details about users’ lives.
