Project Synapse: Blind Spots, ADHD Thinking, and the Data Center Backlash (Plus Chips, Open Source, and China's Video AI)

The hosts discuss using ChatGPT with memory to identify personal blind spots, concluding that multitasking is a myth and linking attention styles (process vs options) and ADHD tendencies to starting many projects without finishing. They compare language fluency to building separate internal "models," relating it to how LLMs work. In a lightning round, they cover rising political backlash to data centers over water, heat, grid strain, and taxes; high electricity costs; and how US corporate structures affect public outcomes. They review responses like more efficient cooling and chips (Nvidia, IBM sub-1nm), IBM's quantum timeline, OpenAI's rapid chip design, chip price lock-ins, and China's innovation under restrictions, including advanced video generation (ByteDance's "CDance 2.5"). They also debate open-source models, orchestration pools like Sakana/Fugu, and concerns about governments restricting access to frontier models such as Anthropic's Fable/Mythos and future GPT releases. 00:00 Weekend Banter 00:28 AI Finds Blind Spots 01:23 Why Multitasking Fails 02:17 Process vs Options Minds 05:24 ADHD and Getting Things Done 08:59 Thinking in Languages 11:32 Data Center Backlash 20:42 AI Distrust and Real Risks 22:52 Efficiency Chips and Retrofits 26:37 Musk Grok and Model Wars 29:27 Open Source Model Boom 32:43 Sakana Fugu Orchestration 36:00 OpenAI Chip and Chip Wars 42:58 Chip Profit Windfalls 44:13 PC Prices Rising Again 45:00 China Reinvents Memory 46:45 China Leads Video AI 50:56 Open Source Momentum 52:34 AI Writing Style Fight 57:55 Model Access Crackdowns 01:07:16 AI Economics and IPO Risk 01:17:49 Media Influence Teaser 01:20:26 CDance Demo and Wrap01_project_synapse (1)