[AINews] Fal’s H3 Max Live breaks the infinite videogen barrier

https://www.latent.space/p/ainews-fals-h3-max-live-breaks-the

The biggest AI shift this week may be easy to miss:

AI video generation just crossed the “real-time” threshold.

Fal’s post-trained MiniMax H3 Max is reportedly fast enough to generate video faster than humans can watch it—enabling truly infinite, interactive AI video streams.

And that’s just the beginning.

The important point isn’t that today’s infinite AI video is good—it’s terrible.

It’s that the constraint has been removed.

Once generation is faster than consumption, entirely new products and interaction models become possible.

The future is arriving through falling inference costs and increasing model speed—not just bigger models.

Source: Latent Space $2026-09-01$.