An essay on the history, theory, progress, and potential of world models, a prominent theme at Nvidia GTC 2026, co-written by General Intuition CEO Pim de Witte

Welcome to the 458 newly Not Boring people who have joined us since our last essay! Join 260,170 smart, curious folks by subscribing here:

Not Boring by Packy McCormick 2026-03-22

Context & Ripple Effects

Nvidia had already put world models into its product portfolio through Cosmos World Foundation Models, while its 2026 GTC preview centered on new infrastructure aimed at agentic workloads. The essay places General Intuition and Pim de Witte within that expanding Nvidia-led conversation rather than presenting a standalone product move.

Later coverage characterizes world models as technically promising but still unsettled, with well-funded startups pursuing the category. That makes a history-and-theory treatment relevant as the field works to define both its capabilities and its practical boundaries.

First-order effects

The essay gives General Intuition and CEO Pim de Witte a visible role in framing world models at a major Nvidia industry gathering.
It reinforces world models as a named GTC theme alongside Nvidia’s existing Cosmos model release, increasing attention on the category’s technical narrative.

Second-order effects

World-model developers will face greater pressure to distinguish concrete capabilities and deployment paths from broad claims about simulation and intelligence, especially as the field remains unsettled on key questions.
Nvidia’s ecosystem partners gain another shared vocabulary for connecting model development with the compute and robotics-oriented platforms Nvidia has been building.

Third-order effects

If world models become a durable workload category, competitive advantage may increasingly depend on pairing models with specialized data, simulation environments, and the infrastructure needed to run them—not just training larger language models.
The category’s eventual structure remains uncertain, but its prominence at Nvidia events points toward frontier AI becoming organized around distinct model classes and their supporting platforms.

The trend: World models are emerging as a frontier-AI category in which model research, simulation tooling, and AI infrastructure are increasingly marketed as one stack.

Discussion

@gen_intuition General Intuition on x
The real world is - or was - uncomputable. If language and code, two of mankind's most powerful inventions, are inadequate to represent our real world, what do we have left? The Answer is World Models They offer a way to do non-deterministic compute efficiently, and to run [image…
@notboringco @notboringco on x
Everything you ever wanted to know about World Models but were too afraid to ask.
@packym Packy McCormick on x
This is very cool. In the World Models piece today, Pim and I wrote that Pi's VLAs are a pragmatic approach to embodied AI and that the company seems to be making a very strategic bet. They keep unhobbling VLAs. [image]
@pimdewitte Pim de Witte on x
Adding a few things that didn't make the final cut as well: Game and physics engines remain hugely important - they offer determinism, are designed for entertainment, and are optimized for delivering on mass consumer hardware. World Models are probabilistic and currently designed
@pimdewitte Pim de Witte on x
Packy and I spent the past month unpacking world models from first principles. This piece is the result of that exploration. We go into why, what, how, and look out into the future on the implications of our work.
@max_rimpel Max Rimpel on x
This is the best primer on arguably the most important frontier in AI: World Models. LLMs are exquisite manipulators of symbols — but they operate entirely in the realm of representation. They can describe clapping, but they cannot clap. The core insight behind World Models is
@pmarca Marc Andreessen on x
Self recommending.

Chronicles