Podcast #11

The AI landscape is shifting rapidly with major breakthroughs announced today, February 13, 2026. OpenAI has launched GPT-5.3-Codex-Spark, a new ultra-fast coding model designed for real-time applications. This smaller yet powerful model is optimized for Codex and boasts an impressive generation speed of over 1,000 tokens per second on low-latency hardware, promising to revolutionize how developers interact with AI assistance.

In the race for funding and valuation, Anthropic has secured a massive $30 billion in Series G funding, pushing its post-money valuation to a staggering $380 billion. The round was led by major institutional investors like GIC and Coatue. Meanwhile, Google continues to refine its capabilities with a major upgrade to Gemini 3 Deep Think. This specialized reasoning mode is now better equipped to handle open-ended scientific and engineering problems characterized by messy or incomplete data, expanding access to Ultra subscribers.

DeepMind is also making waves with the introduction of Aletheia, a math research agent powered by an advanced version of Gemini Deep Think. This agent can iteratively generate and verify long-horizon proofs using intensive tool support, marking a significant step forward in automated mathematical research.

Additionally, Minimax has released two new variants of its M2.5 language model, offering near state-of-the-art performance at a fraction of the cost—dropping inference expenses by as much as 95% compared to top-tier models. In the infrastructure space, leading inference providers are leveraging NVIDIA Blackwell GPUs to cut AI costs by up to 10x using open-source models, benefiting sectors from healthcare to gaming. These developments collectively signal a leap towards faster, more efficient, and highly reasoned artificial intelligence.