Three Buddy Problem On AI’s future, security’s failures, and what comes next...
30 snips
Aug 15, 2025 The hosts dive into the challenges of scaling tech products within large corporations like Microsoft while navigating corporate politics. They explore the AI startup boom and its risks, likening it to the dot-com bubble. The conversation shifts to the geopolitical landscape of GPU technology and export controls, particularly regarding China. They scrutinize recent cybersecurity vulnerabilities and their implications for national security, highlighting urgent privacy issues. The hosts also tease an upcoming live event, blending anticipation with humor.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Lets Few Experts Serve Many
- Cybersecurity has underdelivered as a societal project and needs autonomous, distributed capabilities to scale protection.
- Commoditized evaluative power (AI) can let few experts serve many more organizations if properly applied.
Treat GPU Drivers As Telemetry Sources
- Monitor NVIDIA drivers and telemetry because drivers phone home and can reveal deployed GPU counts if machines are internet-connected.
- Treat telemetry endpoints as a vector for asset visibility rather than assuming GPUs are invisible once shipped.
Manufacturing Ecosystems Matter As Much As IP
- The West leads conceptually in AI but lags in advanced manufacturing capacity tied to China and Taiwan's supply chains.
- Building fabs locally requires rebuilding entire surrounding ecosystems, not just plants.
