The New AI Scaling Axis

After scaling pre & post training, what's next?

Beyond agent harnesses and fine-tuning, there is a third axis to scale that we call Test Time Cognition. This is the science of improving performance and reducing cost by getting more from existing frozen models at inference.

Hosted by Voaige, an AI lab at the intersection of neuroscience and AI, this event explores what is becoming the next AI scaling law: efficient cognition at inference time.

We’ll dive into the inspiration, research, and technology behind Test-Time Cognition, including how Voaige pushed GPT-5.2 beyond GPT-5.4-level performance without fine-tuning or agent harness engineering.