The Biological Computing Co. (TBC) made its mischief 1 video model publicly accessible on October 30, presenting it as a neuron-derived variant of the open Wan 2.2 architecture. The company is exploring whether activity from living neurons can inform the way conventional AI systems are developed. The announcement is a commercial-access milestone for that experiment, not proof that biological computers have replaced digital video-generation infrastructure.
The initial service targets short image-to-video clips rather than general-purpose long-form filmmaking. The fal listing exposes a hosted inference interface and lists 480p and 720p output, with three- and five-second clips. Its posted usage prices begin at $0.005 per second for 480p and $0.01 per second for 720p. Those are API prices, not evidence of the company's operating costs or a universal price guarantee.
OpenRouter separately lists mischief 1 with an October 30 release date and image-to-video capabilities. That marketplace listing corroborates public distribution, although its latency and uptime indicators describe the service observed by the marketplace rather than independently testing the model's research claims.
TBC describes the system as neuron-derived and advertises speed and cost advantages over other generation workflows. The reviewed public material does not supply an independent controlled evaluation of the biological training process, the contribution of neuron signals, or the claimed performance improvements. The model remains based on a conventional open video-generation system.
The wider significance is the attempt to connect wet-lab computing research with a product that developers can actually call. Availability through third-party APIs makes the claim testable by outside users, but quality, consistency, biological contribution and economics still require reproducible comparisons. A demonstration should not be confused with an established new computing paradigm.