Adaption Labs, which aims to create AI systems that learn continuously and cost less to run than current top models, raised a $50M seed led by Emergence Capital
Sara Hooker, an AI researcher and advocate for cheaper AI systems that use less computing power, is hanging her own shingle.
FortuneJeremy Kahn
Context & Ripple Effects
The financing gives Sara Hooker’s new lab backing to pursue an alternative to increasingly compute-heavy frontier-model development. It is an early step in a coverage arc that later included Adaption’s AutoScientist research-automation system, extending the company’s focus from efficient learning to tooling around training and alignment.
Adaption gains $50 million of seed capital and a lead investor in Emergence Capital Partners, enabling the company to build and recruit around continuously learning, lower-compute AI systems.
The round puts the cost and compute profile of model development at the center of Adaption’s initial market positioning, rather than treating scale alone as the product strategy.
Second-order effects
Model builders and AI infrastructure vendors face a sharper incentive to show that capability gains can come from more efficient training and operation, not only larger compute budgets.
If Adaption’s approach reduces the resources needed to improve models, customers sensitive to AI operating costs gain another potential route alongside runtime-optimization providers such as Sail.
Third-order effects
The financing signals a possible split in frontier AI between capital-intensive scale seekers and labs differentiating through learning efficiency; whether that becomes durable depends on whether lower-compute systems remain competitive on useful tasks.
As AI adoption broadens, efficiency may become a strategic layer spanning training, alignment, and inference, shifting some value from raw hardware consumption toward software methods that use available compute more effectively.
The trend: AI investment is expanding beyond model scale into methods and infrastructure designed to make capable systems cheaper to train, improve, and operate.
Excited to share the launch of @adaptionlabs and our $50M fundraise! We're on a mission to build the future of adaptable machine intelligence. Intelligence which learns efficiently and continually from experience. Intelligence which evolves with our ever-changing world.
Beautiful way of expressing a fear a lot of us have. Is the future humans adapting to AI? Or AI adapting to us? Proud to support @sarahookr @sudip_r0y @adaptionlabs as they build AI that continually learns and adapts to us. [image]
AI progress has been dominated by brute force: bigger models, more compute, more centralization. Mozilla Ventures today announced a new investment in @adaptionlabs founded by @sarahookr + @sudip_r0y We're backing Adaption because adaptable, efficient systems give builders more
Today, I am very proud to share our $50M in funding to build AI systems that continually learn across languages, cultures and industries. Even more important proud to share why this is important to us here: https://www.adaptionlabs.ai/
Beginnings are very special. Today is an important day for @adaptionlabs. Today a handful of one-size-fits-all-models are optimized for the average use case. Averages erase the exceptional. Everything intelligent adapts. So should AI. [video]
Congratulations Sara and team for the formal launch of Adaption! We need AI systems that are more efficient and adaptive to their deployment context, and that's what Adaption is setting out to do. Feeling quite proud of seeing a Mila alum taking on this important mission 🚀
“Tools should exist to extend human capability. Instead, we contort. We rephrase. We mould our requests to compensate for AI limitations.” Resonates! Excited to see what @sarahookr and team are building.