Tech Talk Y'all GPT-5, Tilted Earth, and the Robot That Might Kill Me
Aug 8, 2025
Exciting advancements in GPT-5 showcase its capabilities, igniting discussions on AI evolution and competition. Ethical concerns arise as Google's AI fabricates medical information, while subscription fees for smart home devices frustrate users. The quirky twist of Earth tilting unexpectedly adds humor to tech talk. Innovations like Gamma.app revolutionize presentation creation, and a deep dive into local AI models highlights cost-effective options. Plus, beware of data breaches that jeopardize privacy, making authenticity verification ever more crucial.
AI Snips
Chapters
Transcript
Episode notes
GPT-5 Coding Leap
- Adam Walker relays Sam Altman’s claim that GPT‑5 acts like a PhD-level assistant for coding.
- GPT‑5 generated a cleaner, multi-difficulty Minesweeper with the same prompt.
Run Models Locally
- Sanjay Parekh recommends running open-source models locally with Olama’s desktop app for privacy and offline use.
- Olama lets you disable internet access and share models with apps.
Light-Based Anti-Deepfake
- Sanjay Parekh explains embedding invisible light codes to detect doctored video timestamps.
- Multiple coded light sources make AI-driven fakes harder to splice convincingly.