

GPT-5s Are Alive: Outside Reactions, the Router and the Resurrection of GPT-4o
Aug 12, 2025
The conversation dives into the mixed public reactions to GPT-5, highlighting a gap between expectations and reality. It contrasts the performance and capabilities of GPT-5 and GPT-4, revealing frustrations and user experiences. The discussion tackles challenges like hallucinations and complex task handling, while also reflecting on emotional connections users have with these AI models. A lighthearted segment showcases parental milestones alongside humor about AI's evolving landscape, creating an engaging blend of tech critique and personal anecdotes.
AI Snips
Chapters
Transcript
Episode notes
Launch Confusion Explained
- GPT-5 launch blurred reactions because people didn't know if they were reacting to the model, router, or specific variants.
- Rapid changes after release (limits, router, GPT-4O availability) amplified confusion and mixed impressions.
Better For Deep Subject Experts
- Tyler Cowen and others found GPT-5 notably better on their subjects and fast in complex queries.
- Improvements include anticipatory suggestions and richer interactive outputs like charts and diagrams.
Manually Pick Thinking Mode
- Use the explicit model switcher or pay for access to GPT-5 Thinking/Pro when you need stronger outputs.
- Don't rely on the automatic router for all tasks because it can be inconsistent and random.