adaption

Co-founder Sara Hooker will be joining Dr. William Tjhi (Head of Applied Research, AI Singapore), Dr. Shikoh Gitau (Founder & CEO, Qhala), Pak Mahmudi Yusbi (Head of Strategic Planning & Business Development, ASEAN Foundation) moderated by Darius Liu, CFA, CAIA (Head of Global Strategy & Business Expansion, Adaption). The conversation, organized by AI Singapore, will explore inclusive AI development, community-driven data, and building systems that work for real people in real contexts. Intelligence shouldn't arrive preconfigured. When AI is built for the average use case, the communities and contexts that don't fit that average get left on the sidelines. Sara will be talking about what it takes to change that, and putting the ability to shape AI in more hands.

On May 20, Co-founder Sara Hooker will be joining a panel at Asia Tech x Singapore on the future of AI agents. The discussion will explore how advances in memory, reasoning, and multi-agent collaboration are transforming AI from simple task execution into intelligent, autonomous partnership, and what enterprise leaders should prepare for next.

On May 20, our Co-founder Sara Hooker will speak at the Digital Frontier Forum in Singapore, hosted by the Founders Forum Group and IMDA, on the growing infrastructure challenges behind frontier AI. The panel will explore how compute, energy, and efficiency constraints are reshaping AI development, enterprise investment, and the future of scaling intelligent systems.

Today, we're introducing the Public Sector Grant. Public sector and civic institutions operate at the foundation of society, often across dozens of languages, systems, and constraints that most AI was never designed to handle. The Public Sector Grant provides access to Adaption for organizations working in these environments, where intelligence needs to evolve continuously, not stay fixed. Read the full announcement:  https://lnkd.in/eswwhwjj

Co-founder Sara Hooker sits down with Dr. Leslie Teo, Senior Director of AI Products at AI Singapore, on May 18 in Singapore. They’ll dig into the question reshaping AI’s new era: how do we build AI that keeps learning from the world it serves, instead of blindly scaling compute? 

We named it AutoScientist for a reason. Co-founder Sudip Roy breaks it down. It works the problem like a research team would. Co-optimizing the data and recipe, self-improving until quality lands on your goal. How frontier labs operate. Now in your hands.

On May 17th, our Co-founder Sara Hooker will be speaking at AI Engineer’s event in Singapore. Her talk explores a major shift happening in AI: moving beyond static, monolithic systems toward models that continuously adapt, evolve, and stay aligned with the world as it changes. She’ll cover: • Why the future of intelligence depends on adaption, not just scale • Principles behind continual, gradient-free learning • What it takes to build AI systems that are malleable by design • Why AI should work for people, not the other way around Looking forward to the conversations ahead.

The path to frontier AI is still slow, manual, and gated. AutoScientist changes that. It self-improves and automates the full research loop behind model training and alignment. Adaption Co-founder Sara Hooker put it well in TechCrunch: "the whole stack should be completely adaptable, and should basically optimize on the fly to whatever task you have."

Today, we're introducing AutoScientist. Model training and alignment are among the most powerful ways to shape AI, and among the hardest to get right. Most attempts fail for the same reasons: datasets that are too small or too noisy, training signals that conflict, and techniques that only exist inside frontier labs. AutoScientist automates the full research loop. Data and training recipes are co-optimized, iterating until the model converges on the behavior you describe. A solo developer. A small business owner. A researcher working on a language the big models have never heard of. The domain expertise stays human. The frontier research is handled automatically. This is what it looks like to own frontier AI. Read the full announcement: https://lnkd.in/eb8g6Kt8

Sara Hooker and Sudip Roy talk about why almost all frontier model training outside of a handful of labs has been a clear failure. Huge expectations. But poor results. There are a few key things that have gone wrong.