Goju Tech Talk

Today's AI Market is a Reminiscent of the 1999 Dot Com Bubble (Here Comes the Market Crash)

This is the full unedited livestream. Goju spends the evening reading the tea leaves on where AI is actually headed, and the through-line is that the current moment rhymes with 1999. He starts with a viewer clip from Asha on Opus versus subscription pricing, then makes his case that Anthropic's move to API-metered billing looks like a pre-IPO revenue play rather than a product win, and that the real future is local models running on your own hardware once the large models plateau. From there the stream widens out. He does a live fact-check on the report that the acting CISA director uploaded sensitive documents to ChatGPT, and uses his own background in the classified sector to explain why that would be a serious lapse from the top cybersecurity role in the country. He answers James on whether AI can ever touch top-secret or patent-worthy work, and lands on keeping it out entirely. He walks through why he will not publicly endorse a Chinese open-weight model, using the XZ Utils backdoor and a sleeper-trigger thought experiment to show how one download can quietly own a machine. He closes on Prof G's dot-com clip, the Cisco 92 percent drawdown, and the argument that transformers are public-domain work out of Google Brain, so there is no real moat for the frontier labs. His read across the whole night is trust but verify. He puts numbers to OpenAI's burn, sketches a six-to-nine-month window where a Microsoft-Anthropic or Google-OpenAI outcome could play out, and keeps returning to the same advice: keep a small-local-model backup plan, read the primary sources, and do not confuse hype for demand. 🔍 Topics covered: - Asha's clip and the case for local models over metered subscriptions - Anthropic's API pricing read as a pre-IPO revenue play - The acting CISA director's ChatGPT upload and a live fact-check - Why AI has no place near top-secret or patent-worthy IP - The China-specific risk in open-weight models and the XZ Utils backdoor - Prof G's 1999 parallel and the Cisco 92 percent drawdown - Why public-domain transformers leave the frontier labs without a moat - OpenAI burn math and a six-to-nine-month reckoning - Keeping a small-local-model backup plan 💬 Which of these do you think plays out first, a lab acquisition or a real shift to local models? 🔔 Subscribe for no-hype tech analysis: https://youtube.com/@gojutechtalk 📺 Related: Why Dario Amodei's Argument Against Open Source AI Misses the Point https://youtu.be/JR22KE6kLMA 📺 Related: The Dangers of Over-Reliance on LLMs and AI https://youtu.be/U1Dhfij4Uy0 📺 Related: The Problem With Today's AI (In Simple Language) https://youtu.be/Cl7x2OhbPwU #AIBubble #LocalAI #Anthropic #OpenAI #ChinaAI #CISA #XZUtils #ProfG #GojuTechTalk