2025 is the year most Big AI released their first real thinking models
Now we can create new samples and evals for more complex tasks to train up the next gen, more planning, decomp, context, agentic oriented
OpenAI has largely fumbled their early lead, exciting stuff is happening elsewhere