Flapping Airplanes launches with $180 million for AI research

A new artificial intelligence lab called Flapping Airplanes launched on Wednesday, securing $180 million in seed funding. The seed funding round featured participation from investors Google Ventures, Sequoia, and Index. The newly launched lab is entering a highly competitive artificial intelligence sector, but it is doing so with a distinct operational philosophy that departs from the industry’s current trajectory.

This philosophy centers on a fundamental disagreement over how to achieve Artificial General Intelligence (AGI). While much of the tech sector has focused on a “compute-first” scaling paradigm, Flapping Airplanes is aligning itself with a “research-first” paradigm. The scaling paradigm, which currently dominates the industry, argues for dedicating as much of society’s and the economy’s resources as possible toward scaling up existing Large Language Models (LLMs) in the hope of reaching AGI.

In contrast, the research paradigm suggests that the path to AGI requires fundamental scientific discoveries rather than just larger computer clusters. David Cahn, a partner at Sequoia, outlined this perspective, noting that the research paradigm argues the industry is still 2-3 research breakthroughs away from achieving AGI. Because of this, proponents of this approach argue that resources should be directed toward long-running research projects rather than immediate scale.

As Cahn wrote:

The scaling paradigm argues for dedicating a huge amount of society’s resources, as much as the economy can muster, toward scaling up today’s LLMs, in the hopes that this will lead to AGI. The research paradigm argues that we are 2-3 research breakthroughs away from an “AGI” intelligence, and as a result, we should dedicate resources to long-running research, especially projects that may take 5-10 years to come to fruition.

This distinction directly shapes how AI labs allocate their capital and time. A compute-first approach heavily prioritizes cluster scale and favors short-term wins that can be achieved on the order of 1-2 years. Conversely, a research-first approach, such as the one adopted by Flapping Airplanes, is designed to spread its bets temporally. This means investing in long-term projects that may take 5-10 years to come to fruition, and remaining willing to back numerous projects that have a low absolute probability of working but collectively expand the search space of what is technologically possible.

Why it matters

Flapping Airplanes is positioning itself as a research-first lab, contrasting with the industry’s prevailing “compute-first” scaling paradigm. By focusing on long-term research bets rather than immediate cluster scale, the company represents a counter-trend to the server buildouts currently dominating the AI sector.