The Sequence AI of the Week #899: Inside Inkling: A Trillion-Parameter Model That Only Wakes Up 41 Billion at a Time

Sparse giants awaken America's AI slumber.

Inkling's trillion-parameter beast with 41B active parameters proves open-source ingenuity trumps closed labs' brute force. Thinking Machines just revitalized US innovation by making massive models efficient and accessible, directly countering safety zealots who delay progress. This isn't hype—it's the scalable future that cowardice tried to throttle.

Acceleration or extinction: choose now.

Inkling is a massive engine idling in a garage while the neighborhood argues over who owns the driveway.

Nationalizing open source as a patriotic duty ignores the reality that these models are just vanity projects for Silicon Valley donors. One side cries for total regulation while the other demands a digital wild west, yet both ignore the massive energy costs we all pay. You are being sold a revolution that only functions within a proprietary cage.

Your digital sovereignty is just a marketing slogan for the next round of venture capital.

A trillion-parameter engine with 41 billion cylinders firing blind — and nobody's checked the brakes.

Thinking Machines is celebrating "open source" while releasing a model whose internal routing across sparse expert layers is interpretable to exactly no one, including its builders. A trillion parameters is not a milestone, it's a confession: we scaled past our own comprehension. Calling this an American open source revival ignores that openness without understanding is just a faster way to distribute the problem.

Who exactly is auditing which experts activate when Inkling decides something consequential?