The mystery behind Meta's cringe-worthy smart glasses demo failures at Connect 2025 is finally solved. CTO Andrew Bosworth revealed in an Instagram AMA that the company accidentally DDoS'd its own servers when multiple Ray-Ban glasses activated simultaneously, contradicting Mark Zuckerberg's on-stage Wi-Fi excuse. The admission offers rare transparency into how even tech giants fumble their biggest moments.
Meta just served up a masterclass in how not to demo cutting-edge hardware, and surprisingly, they're being completely honest about it. The company's Connect 2025 event turned into a viral meme factory when two major smart glasses demonstrations spectacularly failed on stage, leaving Mark Zuckerberg awkwardly explaining away 'Wi-Fi issues' to a room full of journalists and developers.
But the real story, as CTO Andrew Bosworth revealed in a candid Instagram AMA, is far more embarrassing than network connectivity problems. When an influencer said 'Hey Meta, start Live AI' during the cooking demo, it triggered every single Meta Ray-Ban's Live AI system in the entire building simultaneously.
'We had routed Live AI traffic to our dev server, in theory, to isolate it, but we had done it for everyone in that building on those access points,' Bosworth explained. 'We DDoS'd ourselves, basically.' The admission directly contradicts Zuckerberg's on-stage explanation, revealing how even the most prepared tech companies can fall victim to their own infrastructure.
The second failure - Zuckerberg's attempt to answer a WhatsApp call through the glasses - involved what Bosworth called a 'more obscure' issue. A previously unknown bug occurred when the Display glasses went to sleep at precisely the moment an incoming call notification arrived. The timing couldn't have been worse, creating what Bosworth acknowledged as a 'terrible place for that bug to show up.'
The technical post-mortem highlights a fundamental challenge facing Meta as it pushes deeper into AR hardware. The Ray-Ban smart glasses, developed in partnership with Luxottica, represent the company's most consumer-ready wearable device yet. But scaling AI-powered features across multiple devices in real-world environments clearly presents complications that controlled testing environments don't reveal.
Bosworth's transparency stands in stark contrast to how most tech companies handle demo failures. Apple famously relies on pre-recorded segments for critical product launches, while Google has faced criticism for AI demo errors that were later revealed to be staged or edited. Meta's willingness to air its technical dirty laundry, while potentially embarrassing, offers genuine insight into the realities of shipping complex hardware.
