“People don’t care about what you say, they care about what you build.”
That line is widely attributed to an interview Mark Zuckerberg gave before Facebook went public, back when building something people actually used was all that mattered.
And it’s worth revisiting now.
Because over the past few years, Meta has been saying a lot about artificial intelligence.
But until recently, the company hadn’t built anything that indicated it would be a legitimate contender in the AI race.
That changed last week.
Meta’s Reset
Last Wednesday, Meta introduced a new AI model called Muse Spark.
But to understand why it’s such a big deal, you have to rewind a bit.
You see, Meta’s previous AI strategy revolved around its Llama models, a family of open-weight systems that developers could download and modify.
That approach helped push open-source AI into the mainstream. But it didn’t really move the needle in the AI race.
While Meta was distributing models, companies like OpenAI, Google and Anthropic were pushing performance forward, especially in areas like reasoning and coding.
Eventually, developers started treating Llama as a capable but not best-in-class product.
Meta’s response was a full reset.
Last year, the company reorganized its AI efforts into a centralized group called Meta Superintelligence Labs, shifting away from its earlier open-source-first posture and focusing on building top-tier frontier models.
Muse Spark is the first clear product of this new focus.
According to Meta, the model is designed to be smaller and faster than previous generations, while closing much of the performance gap with leading systems.
Independent analysis suggests that’s directionally true.
On general reasoning and writing tasks, Muse Spark now sits within the range of top-tier models from OpenAI and Google.
That’s a meaningful change from where Meta stood just one cycle ago.
But that doesn’t mean it’s a world beater.
In coding — one of the most commercially important categories — Muse Spark still trails leading systems, particularly those from Anthropic.
That’s a gap Meta will need to overcome soon.
Because developers are using tools like Claude Code today to write software, automate workflows and connect systems together. Over time, those tools will turn into copilots, internal platforms and eventually full-scale automation across companies.
In other words, those AI systems will become something businesses depend on. And once that happens, they’ll become very hard to replace.
Right now, Meta doesn’t have a system that developers are building around.
But by only focusing on coding benchmarks, investors could be missing the bigger picture here.
Because Meta doesn’t necessarily need to win the AI race by building the smartest model. I believe it has the potential to do it by embedding AI into the largest distribution network in the world.
Think about it. Muse Spark is being deployed across:
- Messenger
- Meta’s AI app
- And its smart glasses platform
That ecosystem has a combined user base of nearly 4 billion active accounts.
No other AI company has that kind of instant reach. Not even Grok.
This gives Meta an entirely different path to winning. Because if it can successfully integrate Muse Spark deep enough into those products, users won’t need to go looking for AI.
They’ll already be embedded in it.
Of course, this strategy comes with a cost. Meta expects to spend as much as $135 billion in capital expenditures in 2026. That’s up from about $72 billion the year before.
A large portion of that increase is tied directly to AI infrastructure, including the data centers, compute and cloud capacity needed to support systems like Muse Spark.
At the same time, Meta has been aggressively competing for talent.
Reports indicate the company has offered compensation packages reaching up to $100 million to recruit top AI researchers. That’s a massive investment in both capital and human resources.
And it’s happening before Meta has clearly established how these systems will generate meaningful revenue.
That’s the risk here for Zuckerberg.
Meta no longer has a model problem. Because Muse Spark has proven the company can build systems able to compete at the frontier.
What Meta has is a product problem.
The company has launched AI features before, but none of them have become essential parts of daily behavior.
And until that changes, all of this spending is a bet that adoption will eventually catch up.
Here’s My Take
Muse Spark shows that Meta has re-entered the top tier of AI development.
Its models are now competitive across general tasks, and its infrastructure investment makes it clear the company is serious about competing in the AI race.
But leadership in AI won’t be decided by benchmarks alone. It’ll be decided by adoption.
Meta might be the only company in this race that can win without having the best model. If it successfully embeds AI across its ecosystem, it could turn distribution into dominance.
But Zuckerberg’s 2010 quote still applies today.
Because people ultimately won’t care what Meta has to say about its AI.
They’ll only care if it’s something that they actually want to use.
Regards,
Ian King
Chief Strategist, Banyan Hill Publishing
Editor’s Note: We’d love to hear from you!
If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to dailydisruptor@banyanhill.com.
Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!
