Inside Meta, the company's all-in bet on AI is making life miserable for the engineers who actually have to build it.

According to reports from current employees, Mark Zuckerberg's pivot to making Meta an "AI-first" company has created chaos internally. Projects are being canceled mid-development. Teams are being reshuffled constantly. Engineers are being pressured to rebrand existing work as "AI-powered" even when the AI component is marginal at best.

It's a case study in what happens when executive strategy shifts faster than engineering reality can keep up.

The AI-First Mandate

Following disappointing metaverse results and intense competition from OpenAI, Zuckerberg declared AI the company's top priority. That's not inherently wrong—AI is reshaping the tech landscape, and Meta has strong AI research capabilities.

But according to engineers, the mandate from leadership is less "build great AI products" and more "make everything AI, immediately."

Teams that were building perfectly functional products using traditional software engineering are being told to integrate AI whether it makes sense or not. Machine learning is being shoehorned into features where simpler solutions would work better.

One engineer described it as "AI theater"—making things look AI-powered for exec demos and press releases, even when the underlying technology is conventional or when the AI components actively make the product worse.

The Reorg Churn

Meta has gone through multiple reorganizations in the past year, each one shifting resources toward AI initiatives. For engineers, this means:

- Projects they've spent months building get canceled with little notice - Teams get split up and reassigned to different initiatives - Priorities change weekly based on whatever exec leadership is excited about - Work that was previously valued becomes irrelevant overnight

This is morale poison. Engineers want to ship products and see impact. Constant reorgs mean work gets thrown away, institutional knowledge is lost, and nobody can build anything substantial because the goalposts keep moving.

I've been through this at a startup. It's exhausting. And Meta has the resources to do better.

The Talent Exodus

Predictably, good engineers are leaving. Why stay at a company where your work might be canceled arbitrarily, where you're pressured to add AI buzzwords to appease leadership, and where long-term planning is impossible?

Some are going to actual AI-focused companies like OpenAI, Anthropic, or Google DeepMind, where AI is the core mission rather than a reactive pivot. Others are joining startups or companies with more stable engineering cultures.

The irony is that Meta's AI strategy is driving away the exact engineers it needs to execute that strategy.

When Strategy Becomes Dogma

There's a pattern in tech where CEOs become convinced of a particular future and then force the entire company to align around it, regardless of evidence.

Zuckerberg did this with the metaverse. He was certain it was the future, renamed the company, and burned tens of billions of dollars on VR hardware and virtual worlds that users largely ignored.

Now he's doing it with AI. And to be fair, AI has more obvious product-market fit than VR metaverse experiences. But the execution—forcing every team to become "AI-first" overnight—creates the same problems.

Good engineering strategy is about making thoughtful tradeoffs. Sometimes AI is the right tool. Sometimes it's overkill. Sometimes it makes products actively worse by adding latency, unpredictability, or errors that wouldn't exist in deterministic systems.

But when leadership mandates AI everywhere, engineers lose the ability to make those tradeoffs. Build with AI or get deprioritized.

The Real Costs

The human cost is burned-out engineers and a degraded workplace culture. The product cost is features that use AI because they have to, not because it serves users.

The competitive cost might be worst of all. While Meta is churning through reorgs and forcing AI into everything, competitors with clearer strategies are shipping better products.

OpenAI focuses on frontier models. Google integrates AI into search and productivity. Anthropic focuses on safety and enterprise use cases. They have coherent strategies.

Meta's strategy seems to be "do AI loudly so investors know we're doing AI." That's not a strategy. It's theater.

What Meta Should Do Instead

Here's what a sane AI strategy would look like:

1. Identify specific products where AI creates genuine user value 2. Invest heavily in those areas 3. Let other teams continue using conventional engineering where it works better 4. Give engineers stability to actually ship things 5. Measure impact, not AI adoption

But that requires leadership willing to make hard choices about what not to do. And right now, the mandate seems to be "do everything, but make it AI."

The technology is impressive when applied thoughtfully. The question is whether Meta can slow down long enough to apply it that way.

Based on employee accounts, the answer is no. And that's a problem money alone can't solve.