Building AI-Native Biopharma: announcing its partnership with Incyte Corporation and Kosmos for R&D teams. Sam Rodriques May 19, 2026 At Edison, its goal is to use AI to accelerate the entire pipeline of new medicines, from basic discovery to clinical development. Drug development is fundamentally cumulative: every experiment, analysis, and clinical outcome generates information that should improve future decisions. Yet today, most of that knowledge remains fragmented across teams, systems, and stages of the R&D process. Edison Scientific Inc. believe the full gains from scientific AI will only be realized when AI systems are integrated across the entire research and development stack, continuously learning from the evolving scientific context inside an organization. Building AI-native biopharma organizations will require AI systems that participate directly in scientific workflows, decision-making, and organizational learning. Building AI-Native Biopharma Organizations Today, Edison Scientific Inc. is delighted to announce a partnership with Incyte Corporation (NASDAQ:INCY) to pioneer a new model for AI-enabled drug discovery and development. Kosmos will be embedded across the Incyte discovery and development lifecycle, enabling continuous learning from translational and clinical data, real-time synthesis of evidence and predictive models of therapeutic performance. Its initial deployment with Incyte will focus on integrating Kosmos into high-impact workflows in target discovery, target validation, and translational biology. By embedding its platform directly into Incyte's research environment, Edison Scientific Inc. aim to help Incyte's researchers more effectively synthesize experimental, clinical, and biomarker data, generate and evaluate hypotheses, and support critical R&D decisions throughout the development process. Over time, Edison Scientific Inc. expect these capabilities to expand across Incyte's broader R&D organization. Kosmos: The AI Scientist for R&D To meet the needs of Incyte and future biopharma partners, Edison Scientific Inc. has rearchitected Kosmos to cover the full biopharma R&D lifecycle, from discovery to development. Edison Scientific Inc. has made a number of improvements to Kosmos, aimed at making it more interactive, more powerful, and more useful for scientists throughout the pharma R&D stack. These changes will help scientists and researchers ask better questions, produce better study designs, and increase confidence around go and no-go calls. You can apply for access to the new Kosmos here. Edison Scientific Inc. will work on expanding access as capacity becomes available. The new Kosmos introduces several major advancements for biopharma R&D: * Built for and validated across clinical development: To extend Kosmos into clinical development, Edison Scientific Inc. introduced new data sources, skills, and reasoning models spanning toxicology, clinical trials, and regulatory documentation. Edison Scientific Inc. has validated Kosmos by reproducing key stages of clinical development for a novel checkpoint inhibitor, from lead identification through a Phase II proof-of-concept study. * A persistent scientific collaborator: Because scientific reasoning is inherently iterative, Edison Scientific Inc. rebuilt the experience of working with Kosmos around a persistent, synchronous interface on top of its native asynchronous agent architecture. Kosmos can now interact with researchers and take feedback from them as it goes about its work, making it much easier and more natural for researchers to collaborate with Kosmos on producing complex outputs. Researchers can chat with Kosmos in real time, receive updates as investigations progress, and collaborate with Kosmos through shared workflows, including Slack, Teams and email. Kosmos can also provision compute for GPU-intensive tasks such as AlphaFold and molecular docking calculations, author PDFs, and interpret figures from recent scientific papers. * Trained on your organization's data: Kosmos now integrates with local datasets, cloud storage, and warehouse systems including ELNs, assay data, biomarker data, and unstructured scientific notes, enabling researchers to reason across proprietary organizational knowledge throughout the R&D pipeline. * Leveraging scientific world models: Kosmos constructs internal scientific world models that integrate literature, experiments, clinical outcomes, and organizational context across discovery and development. * Natively trainable architecture: The models underlying Kosmos can now be trained directly with reinforcement learning and supervised fine-tuning on experimental and reasoning data provided by its partners, helping those partners to build a sustainable intelligence advantage. * Enterprise-grade security and traceability: Built for enterprise deployment with SOC 2 compliant infrastructure and fully auditable workflows, Kosmos enables researchers to trace analyses, reasoning steps, and scientific decisions across complex investigations. How Individual Academics and Researchers Can Continue to Work with Kosmos The academic and grass-roots research community played an important role in shaping Kosmos, and Edison Scientific Inc. remain committed to supporting researchers working at the frontier of science. Many of the capabilities announced today were informed directly by how researchers used Kosmos in practice, and Edison Scientific Inc. intend to continue working closely with the academic community as the platform evolves. Here is what that looks like now: * Its core research agents will remain publicly available through the Edison Playground. This includes its Literature agent (formerly Crow), its Analysis agent (formerly Finch), and its Precedent agent (formerly HasAnyone/Owl). All users will continue to have 10 free credits per month for those agents. The legacy Kosmos interface will also be available there for a limited time. * Its free tier for academics is changing. Academics who previously received 650 credits/month will now receive a one-time block grant of 2000 credits instead of monthly refreshers. These credits can be used on the Edison Playground. Edison Scientific Inc. will launch an updated free tier for academics in a few months. * Edison Scientific Inc. will shortly introduce new Kosmos subscriptions. Founding Kosmos subscriptions will no longer be available. Edison Scientific Inc. plan to announce new subscriptions in the coming months. To thank all of its existing subscribers for their support, existing subscribers will receive priority access to Kosmos with significantly higher rate limits and heavily discounted access to higher-tier plans at the same price in the future. You can apply for access to the new Kosmos here. Edison Scientific Inc. will work on expanding access as capacity becomes available.

Accelerating science at scale. Jon Laurent, Siddharth Narayanan, Seenia Hong, Richard Magness 03.16.2026 Edison Scientific partners with NVIDIA to push the frontier of AI for science. Scientific research is facing an intellectual bottleneck. With more than 10 million research papers added every year and datasets growing exponentially, the frontier of science is becoming harder to push forward. This is the problem Edison Scientific, a spin-out of the non-profit FutureHouse, set out to solve: to accelerate discovery across fields by giving every researcher access to an AI scientist. Solving this at scale requires not just capable models and agents, but the infrastructure to train them and the benchmarks to measure them carefully. Its partnership with NVIDIA has been integral to both, enabling its teams to work closely together in adopting NVIDIA NeMo and Nemotron to develop solutions into its stack. Kosmos: its AI scientist. Kosmos, its AI scientist, can complete the equivalent of 6 months of research in a single day with 80% reproducibility. It reads 1,500 papers, runs thousands of lines of analysis code, and maintains coherence across tens of millions of tokens - pushing the boundaries of modern agentic systems. Building an intelligent system that can reason, plan, and execute across complex scientific workflows requires more than a single model. For Kosmos, Edison leverages the NVIDIA NeMo and Nemotron technology stack. NeMo Gym provides scalable reinforcement learning training environments. NeMo RL enables advanced RL training pipelines, including group relative policy optimization (GRPO) and end-to-end FP8 RL training, for models with hundreds of billions of parameters, such as NVIDIA's latest Nemotron 3 open models. Through Aviary, Edison's open-source framework connected with NeMo Gym for scientific environments, Edison Scientific Inc. train agents across biology, chemistry, and related domains for tasks including literature research, bioinformatic data analysis, and multi-step scientific problem-solving. A key component of Kosmos is the Literature Agent, which searches over 175 million scientific articles, patents, and clinical trials. Literature is powered by Nemotron Parse, a specialized vision-language model that transforms how Edison Scientific Inc. process scientific documents. Scientific papers are not clean text documents; they are dense, multimodal artifacts where the key finding is often hidden in a figure, not stated in the abstract. Rule-based PDF parsers fail in this environment: they truncate panels, lose spatial context, and ultimately contaminate context downstream in the retrieval pipeline. Nemotron Parse changes this, unlocking the multimodal pipeline for Edison. It processes documents page-by-page, producing classified bounding boxes and parsed text in meaningful structural categories such as figures, tables, formulas, and captions. Compared to the best rule-based parsers, integrating Nemotron Parse into its Literature Agent improved figure understanding by 15%, text understanding by 3%, and table understanding by 7%. It also unlocked a new capability to have equations accurately converted to LaTeX, substantially improving formula understanding across its pipeline. Measuring what matters. To properly measure the progress of its AI scientist, Edison Scientific Inc. need robust benchmarks. Existing benchmarks are not up to the task. Most existing AI benchmarks measure knowledge regurgitation on clean, structured questions. Real scientific work is different. Evidence can be scattered across dozens of sources and is often ambiguous. A system that aces standardized tests can still fail when asked to synthesize a body of literature, let alone tackle end-to-end research tasks. Building Kosmos has made Edison Scientific Inc. rigorous about measurement - not just because Edison Scientific Inc. want to know if it works, but because the field needs better benchmarks to make progress. Previously, Edison Scientific Inc. released benchmarks to measure this progress: LAB-Bench (and subsequently LABBench2) was the first agentic benchmark for research task capabilities, moving beyond pure knowledge to real-world work. BixBench was the first dedicated benchmark for measuring the bioinformatics capabilities of AI systems, and has since become the de facto standard in the field, cited by multiple teams developing agents for bioinformatics. Introducing BixBench-Hypothesis. BixBench was effective at measuring an agent's ability to carry out correct analysis given explicit instructions. What it did not assess was analytical judgment under ambiguity or the ability to pursue a hypothesis with an open-ended goal, which are closer to how scientists actually work. BixBench-Hypothesis (BBH) addresses this. Derived from the original BixBench capsule framework, BBH reconfigures each task as a (data, hypothesis, protocol) tuple paired with an evaluation rubric. The protocol is a step-by-step guide for carrying out an analysis on the data to address the hypothesis. The rubric is a structured set of expected outputs used to score an agent's analysis, consisting of 1- or 2-point subtasks plus a 5-point final objective tied to correctly supporting or rejecting the hypothesis. There are 51 hypothesis-driven tasks comprising BixBench-Hypothesis (split between Python and R). BBH-Train: an open training dataset, built with NVIDIA. Frontier models currently struggle on BBH, demonstrating the need for continued improvement in analytical judgment and execution. To accelerate progress in this area, Edison Scientific Inc. collaborated with NVIDIA to produce and release BBH-Train - an open-source collection of 250 hypothesis capsules similar in scope to BBH, designed for RL-based training. Edison Scientific Inc. built BBH-Train using a combination of de novo domain expert assembly and AI-assisted construction. For expert assembly, Edison Scientific Inc. engaged a panel of PhD-level bioinformaticians to create capsules from scratch, each contributing an expert-written analytical trajectory that was graded against a rubric through an iterative cycle of review and refinement. For AI-assisted assembly, Edison Scientific Inc. built a pipeline that, given a published study, identified 2-3 sub-hypotheses relying on publicly available data and derived a protocol and rubric for each. These auto-generated capsules were reviewed and revised by the same expert panel and verified as solvable by a purpose-built agent pipeline before inclusion. Pilot experiments have shown that this new dataset is effective at improving the model performance on the BBH benchmark. BBH-Train is available now on HuggingFace. To train agents on BBH-Train, Edison Scientific Inc. also released hypotest, a REPL-like environment that integrates with NeMo Gym and NeMo RL. This work - from the NVIDIA NeMo and Nemotron powered training infrastructure behind Kosmos to the benchmarks Edison Scientific Inc. use to measure it - reflects what Edison Scientific Inc. believe is necessary to build AI that genuinely accelerates science. Edison Scientific Inc. is excited to keep building and scaling Kosmos with NVIDIA technologies across many layers of this stack.