Episode 11 - When AI Becomes Too Smart to Listen

In Episode 11, Avyay and Sidd reflect on a year of rapid change in AI and explain why they’re stepping away from Claude. They compare their experiences with Claude, Grok, Codex, DeepSeek and other emerging models questioning whether benchmark scores really translate into better performance and why an AI that follows instructions can be more valuable than one that constantly overthinks. They also explore the growing problem of vendor lock-in, Cursor’s expanding ecosystem, the future of local versus hosted inference, and why users should retain control of their own workflows. The episode concludes with a compelling vision of AI becoming a utility like electricity with skills serving as the appliances that turn raw intelligence into reliable, repeatable results.