The most important companies in the history of modern technology, from Google to Databricks, were born not from a vacuum of capital and talent, but from a deeply collaborative foundation of open academic research. This is the core theory behind the Laude Institute, a hybrid venture fund and nonprofit founded by Andy Konwinski, co-founder of Databricks and Perplexity AI, that aims to formalize and accelerate the path from research breakthrough to market breakout. Speaking live at NeurIPS 2025, Konwinski detailed Laude's mission to provide “the right resources, the right researchers, the right timing” to unlock the potential of the next generation of world-changing companies.
Konwinski's experience co-founding two multibillion-dollar companies rooted in academic research—Databricks, which grew out of Berkeley AMPLab, and Perplexity, which grew out of generative AI research—has shaped his view that the academic-to-startup pipeline is not an anomaly, but the new gold standard. He argues that the large, cohesive founding teams common to these spinouts, such as Databricks' eight co-founders, are inherently less risky. These teams have already survived years of close collaboration and understand each other's strengths and weaknesses, thereby mitigating the devastating “founder divorce risk” that plagues smaller, less established teams. This established trust and shared intellectual property provides a robust platform for scaling paradigm-shifting ideas.
By operating a two-sided structure, the Lord Institute addresses systemic issues that impede open research. The Venture Division funds researchers once they are established and supports technology founders with capital and expertise from a highly specialized LP base that includes Jeff Dean and top faculty from Berkeley and Stanford. Importantly, nonprofits are providing “no-strings-attached grants” to fund open research before incorporation, building an upstream funnel to fund the next wave of disruptive companies.
The scale of funding required for frontier AI research overwhelms traditional funding mechanisms. Konwinski was quick to emphasize that the National Science Foundation (NSF) is not “broken” but woefully inadequate for the current demands of frontier AI development. The approximately $1 billion annually allocated to computer science research falls far short of the $10 billion to $100 billion needed to maintain global leadership. Laude complements this by directly applying the fast and selective “picker model” that characterizes Silicon Valley venture capital to academic grant writing.
This funding will be strategically directed towards the next layer of AI innovation. Konwinski notes that the current focus is moving beyond the basic model (pre-training and post-training) to a layer he calls “post-training, post-training.” This includes complex systems, context management, search augmentation generation (RAG), memory curation, and advanced prompt optimization. Projects funded through Laude's Slingshot program, such as DSPy, a “reverse compiler” that takes code and compiles natural language, and JEPA-style evolutionary prompt optimization research, exemplify the focus on building highly leveraged abstractions on top of existing large-scale language models. These tools are designed to maximize the utility and reliability of your current models, democratizing access to powerful AI capabilities without requiring you to perform extensive proprietary foundational training.
Mr. Lord's influence extends beyond his traditional Bay Area base. Konwinski pointed out that while West Coast institutions remain important, the majority of their grants and funded projects are intentionally outside of Berkeley/Stanford, spanning universities such as MIT, CMU, UI Urbana-Champaign, Caltech, and international centers in Toronto and Waterloo. This focus is reinforced by the growth of PhD entrepreneurship clubs across these campuses, like Agents at the University of Washington and Seedlings at Stanford University, where Lord is helping foster a broader, geographically dispersed pipeline of research-driven founders who want to commercialize academic research.
The need for strong open research funding is emphasized by current geopolitical dynamics. Konwinski pointed out that Chinese research institutes such as Moonshot and DeepSeek are now publishing twice as many interesting public papers as their US counterparts. The main reason for this is that frontier research institutions in the US, including OpenAI, have stopped publishing major research results. To combat this “open research crisis,” Lord is launching Open Frontiers. This is a livestreamed conference that aims to bring together 100 of the most influential open researchers, including Yann LeCun, François Cholet, and Jan Rijke, to share roadmaps, foster collaboration, and unify the ecosystem. This effort aims to restore America's leadership position in open AI research and ensure that fundamental scientific advances remain a common global asset, not the proprietary intellectual property of corporations. Laude's ultimate goal is to serve as a critical bridge, ensuring that groundbreaking, open research is efficiently translated into products that have world-changing impact and generate tremendous economic value.
