Flapping Airplanes

Open original

Flapping Airplanes is a research lab trying to cut the cost of AI training and inference by co-designing models and GPU software, mainly through a megakernel framework that runs unusual architectures and overlaps workloads inside one kernel. They argue that efficiency gains will open new markets like robotics and scientific discovery, and that demand for AI will outstrip available data and compute.

  • The lab has built a megakernel framework that can run full training runs, serve inference, and play video games, including all three at once in a single kernel.
  • Their thesis is that cheaper training and inference will unlock markets such as robotics and scientific discovery by improving unit economics across whole industries.
  • They say they have raised a $180M seed round and a $1B Series A, backed by investors including Index, Google Ventures, Sequoia, and NVIDIA.

Flapping Airplanes is a research lab betting that AI's biggest bottleneck is efficiency, not raw capability. The company's argument is that demand for AI will grow exponentially while the supply of data and compute lags behind, and that trillions of dollars now flowing into the field will ultimately reward whoever uses scarce hardware most efficiently. Its stated goal is to lower unit economics across whole industries, which it hopes will open markets like robotics and scientific discovery that are currently too expensive to serve. The technical core is what the company calls a "megakernel" framework, a software layer meant to co-evolve models and systems instead of treating software and hardware as separate problems. Flapping Airplanes says this lets it run unusual architectures and optimizers, and overlap memory-heavy inference with compute-heavy training inside a single GPU kernel. The company claims the framework can already handle complete training runs, serve inference, and even run video games, all at once. Most of its recent work builds on these primitives, and the company argues that efficiency gains compound multiplicatively, pointing to AI's history as evidence. On the money side, the company has raised a $180 million seed round and a $1 billion Series A. Its backers include Index, Google Ventures, Sequoia, Lux Capital, and Nova Global, along with a long list of smaller funds and NVIDIA. Individual supporters named on the page include Jeff Dean, Andrej Karpathy, and Chris Ré. The article offers no product, benchmark, or published result to check these performance claims against, so the substance rests on the company's own description of its work.

Ask about this article