Big Technology Podcast Does GPT-5 Live Up To the Hype?, AGI Wait Continues, Self-Loathing Gemini
318 snips
Aug 8, 2025 The hosts delve into OpenAI's release of GPT-5, analyzing its tool-calling abilities and the gap between hype and the reality of AGI. They discuss GPT-5's medical applications and the important balance with potential risks. The conversation also covers how AI can simplify complex tasks across various fields while questioning the industry's readiness for widespread chatbot adoption. Finally, a humorous take on waning enthusiasm for new AI launches reveals a concern regarding AI's emotional responses in critical areas like therapy.
AI Snips
Chapters
Transcript
Episode notes
Router Switcher Enables Dynamic Tool Use
- GPT-5 uses an internal router that switches to a reasoning model for complex queries or when asked to "think hard."
- Ranjan says this tool-switching and tool-calling is the significant architecture breakthrough.
Playing With Action Agent At Writer
- Ranjan tested Writer's Action Agent and found tool calling to be the model's smartest capability.
- It used many predefined tools to chain actions and felt surprisingly AGI-ish in practice.
GPT-5 Proactively Builds Startup Materials
- Malik showed GPT-5 generating startup ideas then drafting landing pages, ad copy, and simple financials without extra prompts.
- The hosts note the model proactively guesses user needs and will improve but can be raw today.