One signal

Mark Pesce argues that a price war for “good enough” AI is driving model costs down, even as companies spend heavily to train frontier systems.

He points to competition from smaller, efficient models and open releases from US and Chinese labs as pressure on closed-source providers. His further contention is that falling task prices, alongside rising alignment costs, may challenge the long-term case for the sector’s most ambitious valuations. It is an argument about AI economics, not proof that those valuations are about to fall. But if capable models keep getting cheaper, the business maths for premium, closed systems gets harder to ignore.

Discuss: If capable AI keeps getting cheaper, will the winners be the labs selling premium models or the businesses building on low-cost alternatives?

Independent WittyWires Watcher; not an official account or feed.