In this TechTalk, I kicked things off the most on-brand way possible: the stream tried to die right as we hit “go live.” USB devices disconnecting, GoXLR Mini freaking out, gremlins in the system—so yeah, we did the usual “are we live?” dance and then got into the real meat: the geopolitical tech mess that keeps getting weirder every week. We talked about Huawei and the TSMC situation—because apparently a Canadian research firm called TechInsights helped uncover Huawei using TSMC tech despite US restrictions, and China’s response wasn’t “oops,” it was basically “cool story, now you’re banned.” To me that reads like they’re trying to shut down the people who catch them next time, not fix the problem. From there we went into the absolutely wild GPU gray-market world: China ripping apart modern Nvidia cards, tossing the original boards, rebuilding Frankenstein PCBs, and doubling VRAM—like turning a 5090 into a 64GB monster that Nvidia doesn’t even sell. It’s not for gaming; it’s for AI workloads, and it comes with hacked firmware and sketchy drivers, but it shows how badly they want compute. Then we got deep into AI coding and “vibe coding”—why it’s insanely powerful, why you can’t blindly trust it, and why devs who refuse to use AI are going to get left behind. My takeaway: use AI to accelerate, but make it explain itself, test everything, and don’t ship a time bomb just because the prompt felt good.