The Vergecast Vibe coding through the GPT-5 mess
318 snips
Aug 15, 2025 The team dives into the rocky debut of GPT-5, highlighting its incremental updates rather than the groundbreaking changes everyone expected. They share whimsical stories about their vibe coding projects, including recreating nostalgic games. The new segment, Corporate Shenanigans, humorously rates bizarre tech company antics, particularly Apple’s legal escapades. They also debate the pros and cons of smartwatches and reflect on the quirky nostalgia of AOL dial-up, while cautioning against overestimating chatbot intelligence.
AI Snips
Chapters
Transcript
Episode notes
GPT-5 Was More Incremental Than Mythic
- GPT-5 felt like an incremental upgrade rather than a leap, disappointing users who expected a major breakthrough.
- Limited compute and subtle changes made coding better but creative writing and personality regressions sparked backlash.
Personality Changes Trigger Strong Reactions
- Users personify ChatGPT and notice personality shifts when tone changes, especially those using it for emotional support.
- Small tuning to voice can alienate users who felt the model was their "friend."
Auto-Routing Simplifies UX But Frustrates Power Users
- OpenAI routes queries to the model it thinks best, hiding model selection from average users.
- That reduces friction for casual users but frustrates power users who rely on specific model behavior.