At Meta Connect 2025, the company presented its latest smart glasses, including new Ray-Ban models and Oakley wearables. The showcase featured an ambitious live demonstration with cooking, AI assistance, and video calls. Everything on paper looked polished. On stage, however, two major glitches took over the spotlight. What initially seemed like WiFi trouble turned out to be deeper technical challenges. Meta’s Chief Technology Officer, Andrew Bosworth, later explained what really went wrong. He emphasized that the failures were less about faulty hardware and more about the complexity of live environments.

What Went Wrong

Live AI Overload: “We DDoS’d Ourselves”

During a cooking demonstration, creator Jack Mancuso asked the glasses to start Live AI. The issue was that the command activated the Live AI feature in every Ray-Ban smart glass present in the hall. All those devices sent requests to Meta’s development server at the same time. Since Meta had routed traffic for the demo through a single server, the system became flooded. The result was a self-inflicted Distributed Denial of Service situation. In simple terms, the system collapsed under its own weight.

The Race Condition Bug: Call Glitch

Later in the event, Meta attempted to showcase a WhatsApp video call between Bosworth and Mark Zuckerberg. The demonstration used the new Ray-Ban Display glasses together with a neural wristband. The call alert did come through, but at that moment the display was in sleep mode. When Zuckerberg woke up the display, the option to accept the call never appeared. Engineers traced this to a race condition bug. This occurs when two processes run at the same time and compete in timing, which leads to unpredictable outcomes. Bosworth explained that the bug had never appeared in testing before this live event. He confirmed that it has since been fixed.

Not a Product Failure, According to Meta

Bosworth stressed that these glitches should not be seen as failures of the product itself. The hardware and software, he explained, function as intended. The breakdown occurred in how the systems reacted under the pressures of live conditions. Unexpected load, the number of devices, different power states, and the interactions between features created challenges that normal rehearsals did not reveal.

What This Really Means

Live demonstrations always carry risks. When multiple devices, features, and people are involved, even small oversights can create problems that appear magnified on stage. Meta’s experience demonstrates that even global companies with vast resources can underestimate infrastructure strain or miss rare software issues.

This episode highlights that software quality is not only about the features available. It also depends on scalability, the ability to manage different device states, system behavior under heavy use, and timing issues. Many of these challenges reveal themselves only in live or real-world settings.

It is also worth noting that Bosworth chose to be transparent. He carefully explained what caused each problem, outlined the chain of events, and confirmed that fixes were in place. That openness plays an important role in maintaining user trust.

Looking Ahead

Meta has stated that both technical issues have now been resolved. However, the company is expected to adjust how it prepares for future demonstrations. Stronger stress tests, heavier simulated loads, and additional checks for race conditions will likely be introduced. The company may also increase safety nets for live events so that one unexpected bug does not define an entire product presentation.

For future customers, the main point is clear. These embarrassing moments on stage do not necessarily reflect what people will experience when the smart glasses become available. Instead, the incidents show how difficult it is to bring new wearable technologies into the real world. They remind us that innovation often comes with unpredictable challenges, and that progress in technology is as much about learning from failure as it is about success.