By Titik Puspa 
In a landscape rapidly populated by ambitious, research-centric AI laboratories, Flapping Airplanes emerges as a particularly compelling new entrant. Fueled by the vision of its exceptionally young and inquisitive founders, the startup is charting a course toward revolutionizing artificial intelligence by developing significantly less data-hungry training methodologies. This pursuit holds the potential to fundamentally alter the economic viability and expansive capabilities of AI models, a mission bolstered by a substantial $180 million in seed funding, providing ample runway for their groundbreaking research.
The impetus behind Flapping Airplanes’ inception lies in a profound recognition of the current limitations and an optimistic outlook on future possibilities within the AI domain. "There’s just so much to do," states Ben Spector, one of the co-founders. "The advances we’ve seen over the last five to ten years have been spectacular. We love the tools; we use them every day. But the question is, is this the whole universe of things that needs to happen? And we thought about it very carefully, and our answer was no, there’s a lot more to do." This sentiment is echoed by his brother, Asher Spector, who adds, "This question is just so scientifically interesting: why are the systems that we have built that are intelligent also so different from what humans do? Where does this difference come from? How can we use knowledge of that difference to make better systems?"
At the core of Flapping Airplanes’ strategy is a concentrated bet on the critical importance of data efficiency. While leading AI models like those developed by OpenAI and DeepMind have achieved remarkable feats through massive data consumption – often referred to as training on "the sum totality of human knowledge" – the Spector brothers and co-founder Aidan Smith believe this approach is inherently inefficient. "Humans can obviously make do with an awful lot less," Ben explains, highlighting the significant gap between current AI capabilities and natural learning processes. "There’s a big gap there, and it’s worth understanding."
This focus on data efficiency is not merely an academic pursuit; it’s underpinned by a belief in its commercial and societal value. "It’s a bet that this data efficiency problem is the important thing to be doing. Like, this is really a direction that is new and different and you can make progress on it," Ben elaborates. "It’s a bet that this will be very commercially valuable and that will make the world a better place if we can do it." This conviction extends to the team’s composition, with Ben adding, "It’s also a bet that the right kind of team to do it is a creative and even in some ways inexperienced team that can go look at these problems again from the ground up."
Aidan Smith, who brings a background from Neuralink, a company deeply immersed in understanding the human brain, views the biological model as a crucial point of reference. "If you look at the human mind, it learns in an incredibly different way from transformers," he observes. "That’s not to say better, just very different. So we see these different trade-offs. LLMs have an incredible ability to memorize and draw on this great breadth of knowledge, but they can’t really pick up new skills very fast. It takes just rivers and rivers of data to adapt." Smith further posits that the algorithms employed by the human brain are fundamentally distinct from contemporary AI training techniques like gradient descent. "That’s why we’re building a new guard of researchers to kind of address these problems and really think differently about the AI space."
The name "Flapping Airplanes" itself alludes to this philosophical divergence. While current AI systems are akin to sophisticated, large-scale aircraft like the Boeing 787, Flapping Airplanes isn’t aiming to replicate birds. Instead, it seeks to build an "airplane that flaps" – a novel approach inspired by natural principles but not bound by them. Ben clarifies, "My perspective from computer systems is that the constraints of the brain and silicon are sufficiently different from each other that we should not expect these systems to end up looking the same." He emphasizes that the vast differences in hardware, computational costs, and data locality necessitate distinct architectural designs. However, he firmly believes that these differences should not preclude drawing inspiration from the brain to enhance AI systems.
The company’s unique approach to talent acquisition further underscores its commitment to fresh perspectives. Flapping Airplanes actively recruits exceptionally young individuals, some still in college or even high school, who demonstrate exceptional creativity and a knack for teaching new concepts. "The number one thing we look for is creativity," states Aidan. "Our team is so exceptionally creative, and every day, I feel really lucky to get to go in and talk about really radical solutions to some of the big problems in AI with people and dream up a very different future." Ben adds that a key indicator for him is whether a candidate teaches him something new. "If they teach me something new, the odds that they’re going to teach us something new about what we’re working on is also pretty good." This deliberate cultivation of a team unburdened by established paradigms is seen as a strategic advantage in tackling fundamental research problems.
The significant seed funding of $180 million, secured in a challenging investment climate, reflects a strong investor appetite for groundbreaking AI research. "The market has been hot for many months at this point. So it was not a secret that large rounds were starting to come together," Ben acknowledges. "But you never quite know how the fundraising environment will respond to your particular ideas about the world." He credits their success to a clear message that resonated deeply with investors, many of whom expressed a desire for exactly the kind of research-driven approach Flapping Airplanes embodies. "We have been extremely fortunate to have found a group of amazing investors who our message really resonated with and they said, ‘Yes, this is exactly what we’ve been looking for.’" Aidan concurs, noting "a thirst for the age of research has kind of been in the water for a little bit now."
This substantial funding allows Flapping Airplanes to prioritize research over immediate product development, a deliberate choice to avoid distraction. "We want to try really, really radically different things," Aidan explains. "And sometimes radically different things are just worse than the paradigm. We’re exploring a set of different trade-offs. It’s our hope that they will be different in the long run." Asher emphasizes that while they are excited about commercialization, the initial focus must be on fundamental research. "If we start by signing big enterprise contracts, we’re going to get distracted, and we won’t do the research that’s valuable." Ben reinforces this, stating, "When you’re a startup, you really have to pick what is the most valuable thing you can do, and do that all the way. And we are creating the most value when we are all in on solving fundamental problems for the time being." He remains optimistic that sufficient progress will enable them to engage with real-world applications and gather valuable feedback.
The economic implications of data-efficient AI are profound. Asher posits, "Lots of regimes that are really important are also highly data constrained, like robotics or scientific discovery. Even in enterprise applications, a model that’s a million times more data efficient is probably a million times easier to put into the economy." This efficiency could unlock new AI applications in fields where data collection is inherently difficult or expensive. Ben envisions a future where AI transcends mere automation, becoming a catalyst for entirely new scientific discoveries and technologies that are beyond human conception. "The most exciting vision of AI is one where there’s all kinds of new science and technologies that we can construct that humans aren’t smart enough to come up with, but other systems can." He believes that models capable of deep understanding, rather than mere statistical pattern matching, are crucial for achieving this.
The debate surrounding Artificial General Intelligence (AGI) and the potential for singularity is approached with a pragmatic and grounded perspective by the Flapping Airplanes team. Asher states, "I really don’t exactly know what AGI means. It’s clear that capabilities are advancing very quickly. It’s clear that there’s tremendous amounts of economic value that’s being created. I don’t think we’re very close to God-in-a-box, in my opinion." Ben echoes this sentiment, emphasizing the vastness of the field and the ongoing work. Aidan, however, offers a forward-looking perspective: "The brain is not the ceiling, right? The brain, in many ways, is the floor. Frankly, I see no evidence that the brain is not a knowable system that follows physical laws. In fact, we know it’s under many constraints. And so we would expect to be able to create capabilities that are much, much more interesting and different and potentially better than the brain in the long run."
The question of computational costs, a significant barrier for scale-driven AI, is viewed differently by Flapping Airplanes. Ben argues that pursuing "really crazy, radical ideas" can paradoxically be cheaper than incremental work. "When you do incremental work, in order to find out whether or not it does work, you have to go very far up the scaling ladder." Radical new ideas, even if they fail quickly, provide rapid feedback without extensive computational investment. While scale remains an important tool, their focus on data efficiency means they can test many ideas at smaller scales, mitigating the need for immediate, massive compute. Asher concisely captures their philosophy: "You should be able to use all the internet. But you shouldn’t need to. We find it really, really perplexing that you need to use all the Internet to really get this human level intelligence."
The potential outcomes of this data-efficient approach are multifaceted. Asher outlines three key hypotheses: First, that reduced data reliance forces models into "incredibly deep understandings of everything it’s seen," leading to more intelligent reasoning rather than mere fact recall. Second, it could enable vastly more efficient post-training adaptation, allowing models to acquire new capabilities with minimal examples. Third, it could unlock entirely new verticals for AI, particularly in areas like robotics and scientific discovery where data constraints have historically hindered progress. Ben adds that the pursuit of deep insights, crucial for scientific breakthroughs, relies on models operating on the "creativity side of the spectrum," moving beyond simple interpolation of data.
The company is actively seeking engagement from the AI community and potential collaborators. They invite communication via email, even for those who wish to disagree with their approach. "We have [email protected]. If you just want to say hi. We also have [email protected] if you want to disagree with us," Asher shares. They are particularly keen to connect with "exceptional people who are trying to change the field and change the world." Ben emphasizes that unconventional backgrounds are welcome, stating, "You don’t need two PhDs. We really are looking for folks who think differently."
Looking ahead, the future of AI, shaped by the work of labs like Flapping Airplanes, is anticipated to be profoundly different from what is currently achievable. Asher predicts that future models will exhibit "smarter in even stranger ways," leading to "unknowable, alien changes and capabilities at the limit." Ben, while perhaps more tempered in his outlook on immediate consumer experience, broadly agrees that their research agenda aims to build capabilities "fundamentally different from what can be done right now." The ambition is not merely to create incremental improvements but to forge a new generation of AI that is more efficient, more adaptable, and ultimately, capable of unlocking entirely new frontiers of knowledge and innovation.