TBC Launches mischief 1, an AI Video Model Built With Insights From Living Neurons

What TBC is releasing

TBC, a San Francisco company that describes itself as an applied biological computing company using living neurons to improve AI models, announced mischief 1 on October 30, 2026. The model is available on fal and OpenRouter, two marketplaces where creators and developers can access AI models.

mischief 1 is built on Wan 2.2, a widely used open-weight video model. According to TBC, it matches the quality of that base model while generating video 5x faster and at a fraction of the cost. The company says the speed and price gains come from efficiencies it discovered in living neurons and applied directly to the model.

The launch follows TBC's announcement, five weeks earlier, of a collaboration with Amazon Web Services (AWS) to bring the model to enterprise customers. Availability on fal and OpenRouter extends access to individual creators and developers.

Pricing and early user feedback

TBC prices mischief 1 at half a cent per second. At that rate a five-second clip costs 2.5 cents, and 100 clips cost $2.50. The company argues this lets creators test more ideas and compare takes within the same budget, and makes it easier for developers to build video into products people can afford to use repeatedly.

Wonder Studios, an AI-native creative studio, tested the model in its own production workflow. Co-founder Justin Hackney said that mischief 1's speed and cost-effectiveness meant the studio "could push the work much further within the same time and more cost effectively." He added that the approach, as it reaches more advanced models, "could reshape how films and commercials are made."

How the neuron-based approach works

TBC says its discovery platform grows living neurons in the lab, measures how they process information and identifies computational principles that can improve AI. Those findings become proprietary software that makes existing models more efficient.

The neurons themselves stay in the lab. Customers use software that runs on standard AI hardware and, according to TBC, adds less than 0.1% to the base model's parameter count.

The company contrasts this with the techniques behind most efficiency gains in AI today, which it lists as quantization, distillation and caching. TBC says its optimizations come from direct measurements of living neurons, so they are derived from biology rather than only inspired by it, which it argues makes them hard to replicate.

What comes next

CEO and co-founder Alex Ksendzovsky said video is the company's first commercial application and that TBC is already working to bring its discoveries to language and physical AI, with the goal of building more capable models "without simply adding more compute." TBC says it is already applying its discovery process to newer open-weight video models, with early results showing greater efficiency gains. Applications in language models and physical AI are in development.

Co-founder and COO Jon Pomeraniec described mischief 1 as the company's "first commercial proof point" and said the larger opportunity is applying these discoveries across models and workloads.

mischief 1 went live on fal and OpenRouter on October 30, timed to Mischief Night. For video creators and developers, the immediate offer is a low-cost image-to-video option. The performance and price claims come from TBC itself, and the broader promise of efficiency gains in other model types remains in development.