EP38: 2025 Year in Review, A Look Back at Our 2025 AI Predictions, The Reasoning Revolution

Co-hosts Mark Thompson and Steve Little present a special year-end episode, comparing their 2025 AI predictions to what actually unfolded in 2025. This episode is a great review of the top AI advancements in 2025!

The hosts examine which predictions hit the mark, including the agentic AI hype cycle, plummeting AI costs driven by DeepSeek, and the dethroning of OpenAI as the top AI model. They also explore predictions that proved partially accurate, such as the shift to local language models and the adoption of AI in social media.

Mark and Steve have a good laugh over their biggest misses, including breakthroughs in text-in-image generation, image restoration, and vibe coding. They highlight how reasoning models were the transformative force behind nearly every major AI advancement in 2025.

The episode closes with a preview of next week's 2026 predictions episode.

Timestamps:03:30 Agents, Agents, Everywhere: Deep Research and Agentic Browsers

09:49 Cost of AI Drops Like a Rock: DeepSeek Disrupts the Market

12:26 OpenAI Dethroned: Gemini and Anthropic Rise

18:46 Local Language Models

23:54 AI Invades Social Media

28:24 AI-Enhanced Writing: From Grammar Checking to Ghost Writers

32:30 Family Tree Diagrams: Possible But Not Practical

36:03 Handwriting Recognition: Reasoning Improves Results