Tech Talks Daily 3557: MythWorx Explains Why Reasoning Matters More Than AI Scale
Jan 17, 2026
Wade Myers, a tech entrepreneur and founder of MythWorx, discusses a paradigm shift in AI from size to reasoning. He critiques the current scaling practices of large language models, highlighting the challenges of power, cost, and latency. Wade emphasizes how GPUs create a 'tax' on tech progress and introduces MythWorx's innovative reasoning models that learn efficiently and provide clearer audit trails. He also raises concerns about data sovereignty for enterprises, advocating for low-footprint, private AI solutions to meet growing demands.
AI Snips
Chapters
Transcript
Episode notes
Scaling Hits Physical Limits
- Large language models scale by ingesting ever more data and compute, which becomes unsustainable over time. Wade Myers argues this approach runs 'off the edge of the world' as compute, power, and data growth collide with physical limits.
Reasoning Beats Memorization
- Reasoning-based AI models mimic human learning and can deliver answers without carrying vast corpora of data. MythWorks claims such models can outperform huge LLMs using a tiny fraction of the power.
Learning Enables Generalization
- Predictive LLMs require training on expected questions and still hallucinate; reasoning systems can generalize from learned discipline. Wade Myers contrasts blind benchmark training to learning-based problem solving like a human taking a math test.