Adaption Labs

Adaption Labs is an American artificial intelligence research startup founded in 2025 by Sara Hooker, the former head of Cohere's research lab Cohere Labs, and Sudip Roy, Cohere's former senior director of inference computing. The company is headquartered in San Francisco and develops AI systems designed to adapt and learn continuously from real-world interaction rather than scaling through ever-larger pretraining runs. As of February 2026, Adaption Labs had raised $50 million in seed funding led by Emergence Capital Partners.

At a glance

• Founded: 2025 in San Francisco. Public announcement October 7, 2025.
• Status: Private. Reported valuation approximately $500 million.
• Funding: $50 million seed round closed February 2026, led by Emergence Capital Partners. Co-investors include Mozilla Ventures, Fifty Years, Threshold Ventures, Alpha Intelligence Capital, E14 Fund, and Neo.
• CEO: Sara Hooker (co-founder, former vice president of AI research and head of Cohere Labs).
• Other notable leadership: Sudip Roy (co-founder, former senior director of inference at Cohere).
• Open weights: None disclosed.
• Flagship products: No public model release as of April 2026. Research focus on adaptive data, adaptive intelligence, and adaptive interfaces.

Origins

Adaption Labs was founded in 2025 by Sara Hooker and Sudip Roy following their respective departures from Cohere. Hooker had led Cohere Labs (formerly Cohere for AI), the research division of Cohere, from April 2022. Under her leadership, Cohere Labs ran the Aya Project, a multilingual large-language-model program covering 101 languages with contributions from more than 3,000 researchers, and shipped the Aya Expanse 8B and 32B multilingual models in 2024. Hooker's prior career included six years at Google Brain (2017 through 2022) as a research scientist focused on model interpretability, where she coined the "Hardware Lottery" framing of how compute-architecture constraints shape AI research directions, and she helped establish Google's first AI research center in Accra, Ghana. She earned a B.A. from Carleton College and a PhD in computer science from Mila in Quebec.

Roy joined Hooker as a co-founder. His prior role at Cohere was senior director of inference computing, with responsibility for the lab's inference infrastructure and deployment systems. Both Hooker and Roy previously worked at Google before Cohere.

The company was announced quietly on October 7, 2025 with a public framing that emphasized contrarian positioning against the dominant scaling thesis of frontier AI labs. Hooker stated publicly that "the formula of just scaling these models hasn't produced intelligence that is able to navigate or interact with the world." The seed round was closed in February 2026 at $50 million led by Emergence Capital Partners, and the company has stated plans for worldwide hiring through 2026.

Mission and strategy

Adaption Labs's stated mission is to build "thinking machines that adapt and continuously learn." The strategic premise is that current frontier AI systems, despite their growing capability, are structurally limited because they freeze their parameters after pretraining and rely on prompt engineering to adapt to new tasks, with each adaptation requiring expensive fine-tuning runs.

The strategy is organized around three thematic pillars:

• Adaptive data. AI systems generate the data they need on the fly through interaction with their environment, rather than relying on static pretraining corpora.
• Adaptive intelligence. Systems automatically adjust their computational spending based on the difficulty of the problem, in contrast to the fixed inference cost of current frontier models.
• Adaptive interfaces. Systems learn from user interactions over time, building persistent context and adjusting their behavior to match user preferences.

The technical research direction includes work on "gradient-free learning," with specific approaches such as "on-the-fly merging" (selecting from a population of adapter models at inference time) and "dynamic decoding" (adjusting output probabilities during generation without modifying core weights). The company has stated an objective of "eliminating prompt engineering" by building systems that adapt automatically to new tasks rather than requiring human-engineered prompts.

The competitive premise is that AI systems with smaller parameter counts but stronger continuous-learning capabilities can outperform larger frozen-parameter models on real-world tasks, while operating at substantially lower compute cost. Hooker's framing positions Adaption Labs as a counterweight to the scaling-and-compute strategy of OpenAI, Anthropic, and Google DeepMind.

Models and products

• Foundational research output. Adaption Labs's primary output as of April 2026 consists of research toward adaptive AI techniques, including gradient-free learning approaches.
• No shipped models, APIs, or open weights. The company has not released a model, an API, or open weights as of April 2026.

The commercial distribution strategy beyond research output has not been publicly stated. Whether Adaption Labs intends to ship its own consumer or enterprise product, license adaptive techniques to other labs, or pursue a hybrid posture has not been disclosed.

Benchmarks and standing

Adaption Labs has not released a model and is not represented on the standardized capability leaderboards as of April 2026. The company's standing rests on the founders' research credentials at Cohere and Google Brain, the lead-investor signal from Emergence Capital, and the contrarian-thesis positioning against the dominant scaling strategy.

The standardized leaderboards measure pretraining-and-fine-tuning capability rather than continuous-learning adaptation, and may not be the appropriate evaluation framework for Adaption Labs's research output even after a public model release.

Leadership

As of April 2026, Adaption Labs's named leadership consists of its two co-founders:

• Sara Hooker, co-founder and chief executive. Former vice president of AI research at Cohere and head of Cohere Labs. Prior six-year tenure at Google Brain (2017 through 2022). Named one of Fortune's top 13 AI innovators in 2023 and included in TIME's 2024 list of influential AI figures. Public face for the company on capability claims, fundraising, and the contrarian-thesis framing.
• Sudip Roy, co-founder. Former senior director of inference at Cohere. Prior tenure at Google. Brings systems and inference-infrastructure depth to the founding team.

The senior research and engineering team has not been disclosed by name. The company has stated plans for worldwide hiring through 2026.

Funding and backers

Adaption Labs's funding history through April 2026 consists of a single closed round: the $50 million seed announced in February 2026. The round was led by Emergence Capital Partners, with Mozilla Ventures, Fifty Years, Threshold Ventures, Alpha Intelligence Capital, E14 Fund, and Neo participating.

The Emergence Capital lead is consistent with the firm's enterprise-software focus and its track record of leading early rounds for Salesforce, Veeva, Box, and other enterprise platforms. The Mozilla Ventures participation signals alignment with open-source and accessible-AI positioning, given Mozilla's editorial focus on alternatives to closed-weights frontier models.

The company has not disclosed cumulative compute commitments, cloud-partner relationships, or follow-on financing plans. The reported valuation of approximately $500 million is consistent with mid-size 2025-vintage Insurgent labs founded by senior-team founder-departure cohorts.

Industry position

Adaption Labs occupies a distinctive position within the 2025-vintage Insurgent cohort. The combination of an established research-leader founding pair from Cohere, the contrarian-thesis positioning against scaling-and-compute, and the adaptive-AI research framing produces a profile that partially overlaps with Flapping Airplanes and Humans& but is differentiated on technical thesis.

The closest peer comparators are research-first Insurgents pursuing alternatives to frontier-model scaling. Flapping Airplanes shares the contrarian-thesis positioning and the "data is the bottleneck" framing, though with a different mechanism (data-efficiency in pretraining rather than continuous learning at inference). Reflection AI and Magic pursue novel architectures for frontier-model training. Humans& shares the multi-lab senior-departure pattern but emphasizes human collaboration rather than adaptive systems.

The strategic risks are substantial. The continuous-learning thesis has not been validated through a shipping product or public model release. The gradient-free learning research direction is novel and has limited prior demonstration at the parameter scales required for useful task performance. The valuation depends on team credentials and lead-investor signals rather than capability evidence.

The strategic strengths are distinctive. Hooker's senior research-leadership tenure at Cohere Labs and Google Brain, including the Aya multilingual program, provides credibility on building research output at scale. The contrarian-thesis framing is differentiated within an Insurgent cohort largely populated by labs that pursue scaling-leading approaches. The Emergence Capital-led syndicate signals interest from enterprise-software-focused investors that could become strategic-partner channels for any future product release.

Competitive landscape

Adaption Labs competes with several Frontier and Insurgent labs:

• OpenAI, Anthropic, and Google DeepMind. The dominant Frontier labs whose scaling-and-compute strategy Adaption Labs's continuous-learning thesis is positioned against. Adaption Labs does not compete head-to-head on shipped capability as of April 2026.
• Cohere. Hooker and Roy's previous lab. Cohere pursues a retrieval-augmented-generation and enterprise-distribution strategy distinct from the continuous-learning research direction.
• Flapping Airplanes. Closest peer Insurgent on contrarian-thesis grounds. Both founded in 2025 by research-first teams pursuing alternatives to the dominant scaling strategy. Flapping Airplanes emphasizes data efficiency in pretraining; Adaption Labs emphasizes continuous adaptation at inference.
• Liquid AI. Architecture-research peer. Liquid AI pursues alternative state-space models; Adaption Labs pursues gradient-free learning. Different mechanisms, similar contrarian positioning.
• Humans&. Closest peer on the multi-lab senior-departure founding pattern, though differentiated on technical thesis.
• University-based AI research programs. Mila (Hooker's PhD-granting institution), Stanford AI Lab, and the broader academic ML community produce some of the same continuous-learning research that Adaption Labs targets.

Outlook

Several open questions affect Adaption Labs's trajectory in 2026 and 2027:

• The first published research output, including the company's specific gradient-free learning techniques and any empirical demonstration at scale.
• The first model release, if any, and the capability profile relative to frontier-tier comparators.
• The commercial strategy beyond research output, which has not been publicly stated.
• Whether the company accepts follow-on capital at a higher valuation, and on what timeline.
• Hiring momentum, particularly from Cohere, Google, and the broader academic ML community.
• The durability of the continuous-learning framing as competitor labs incorporate similar adaptive techniques into their inference pipelines.

Sources

• Fortune: Former Cohere execs Sara Hooker and Sudip Roy secure $50 million seed round for their new startup Adaption Labs. Primary source on the $50 million seed round.
• TechCrunch: Why Cohere's ex-AI research lead is betting against the scaling race. October 2025 founding announcement and contrarian-thesis framing.
• BetaKit: Ex-Cohere execs Sara Hooker and Sudip Roy unveil new AI startup. Founding details.
• Sara Hooker on Wikipedia. Career history including Carleton, Mila, Google Brain, and Cohere Labs.
• Yahoo Finance: Former Cohere exec Sara Hooker has raised $50 million for her AI startup Adaption Labs. Funding round coverage.
• The Logic: Former Cohere executives launch startup Adaption Labs. Founding context.