AI Applied: Covering AI News, Interviews and Tools - ChatGPT, Midjourney, Gemini, OpenAI, Anthropic OpenAI's Pivot: Shutting Down Sora & What Comes Next
Mar 30, 2026
A fast breakdown of OpenAI's recent pivots, product shutdowns, and why certain apps failed to find a purpose. A look at perceived declines in ChatGPT performance and how model routing and paid tiers affect quality. Discussion of GPT-5.4 Pro’s strengths for advanced tasks and why compute is being shifted toward enterprise use cases.
AI Snips
Chapters
Transcript
Episode notes
OpenAI Still Leads At The Cutting Edge
- OpenAI still leads at the absolute cutting edge with models like GPT-5.4 Pro for very hard tasks.
- Conor says top users (e.g., Ethan Mollick) find GPT-5.4 Pro uniquely effective compared with alternatives like Opus or Claude for complex prompts.
Sora Felt Hollow Compared To Real Viral Clips
- Conor says he never liked Sora; it felt meaningless because unlimited synthetic short-form content dilutes real human moments.
- He describes short social clips' appeal (real human surprise) versus Sora's cheap, twilight-zone creations.
Model Router Makes ChatGPT Seem Worse
- ChatGPT feels 'dumb' often because OpenAI automatically routes queries to older, cheaper models by default.
- Jaeden explains the model router picks instant/legacy models instead of 5.4 unless users explicitly change settings.