Graphon Labs raises $35M for persistent graphon intelligence memory it hopes will 'accelerate progress' towards AGI
retained text
AI/ML Graphon Labs raises $35M for persistent graphon intelligence memory it hopes will 'accelerate progress' towards AGI There are big claims and then there are BIG claims, and startup up Graphon Labs is making one of these; it says it is building the foundational infrastructure to help "accelerate progress towards AGI" and has raised $35 million to do it. This plan is to build persistent intelligence to enable AI to continuously access, understand, remember, and build upon everything that came before, to help speed up work towards Artificial General Intelligence. It says that, as AI evolves from models that answer questions to agents that run around the clock, the ability to retain context, learn from experience, and accumulate knowledge over time becomes essential. The technology is built on graphons (graph functions), a mathematical framework for analyzing very large graphs, which Graphon Labs’ founders developed during their PhDs. Arbaaz Khan, Graphon Labs co-founder and CEO, said: “The world is heading toward agents and physical systems that create and depend on more information than anyone has ever had to manage. In a world where we are moving to AI using computers, we need to fundamentally rethink how computers access data to pre-fill their context windows. The old way, using vector search, is like searching through the pages of a torn-up book. We take those pages and turn them into a connected map, unlocking seamless memory search.” Graphon is a persistent intelligence platform for multimodal data. It ingests videos, images, audio, and documents using patent-pending technology, goes deep into the data to automatically build an ontology, and gives AI the context it needs. It says vector search sees points, cutting our data into chunks, turning each one into a point, and returning the nearest few. Everything between the points (the reasons, the sequence, the relations) is gone. Graph RAG scores a chunk plus its neighbors, and is still limited. Graphon builds one large graph of your data, then fits a graphon to it: the limit object of that graph, a single function that maps the topology of the full graph. A graphon is the continuous limit of a graph: one smooth function over all of your data instead of a pile of points. Nothing between the points is lost. The other founders are CTO Clark Zhang and COO Deepak Mishra and the company was started up in July 2025 in San Francisco. The $35 million comes in an A-round led by RRE Ventures, with participation from HP, Energy Impact Partners, Hitachi Ventures, and Gaia VC. It follows an $8.3 million seed round. The graphon technology is derived from PhD research at the GRASP (General Robotics, Automation, Sensing and Perception ) Lab within the School of Engineering and Applied Sciences at the University of Pennsylvania. The founders say their technology rethinks each layer of the stack, across storage, search, retrieval, reasoning, and memory. The result preserves what an agent knows, what it has done, what changed, and what happened as a result, with relationships, sources, and timing intact. It also cuts the compute time it takes to represent and reason across enormous amounts of information. Graphon Labs says it’s building an intelligence architecture that scales from the enterprise to the individual. The goal is for it to sit beneath agents wherever intelligence needs to know, reason, act, remember, and learn. There are three products: - Graphon Enterprise: secure, governed memory across enterprise data across all modalities, so agents and teams can search and reason over it, deployed on hosted, Graphon-managed VPC, customer-managed VPC or on-prem infrastructure - Graphon Developer: builders can build over any data or modality without recreating the context stack - Graphon On-Device Memory: private, persistent multimodal intelligence on phones, laptops, cameras, wearables, and other devices Their technology achieved the highest accuracy among evaluated systems on three leading benchmarks for long-term context: LoCoMo, ConvoMem and ATM-Bench. On LoCoMo, Graphon scored 93.9 percent accuracy using 58 percent fewer tokens than loading the full conversation into the model. On ConvoMem, it achieved 90.2 percent accuracy with 95 percent fewer tokens. On ATM-Bench, a benchmark from researchers at the University of Cambridge testing question answering across video, audio, images and text, Graphon scored 81.7 percent, 5.0 percentage points ahead of the next system on the public leaderboard. LoCoMo and ConvoMem comparisons use baseline results published by Databricks. Initial users include the San Francisco 49ers and Hitachi Rail. Khan added: "This funding takes us from early experimentation to scaled enterprise deployments, opens our infrastructure to developers, accelerates our research, and brings Persistent Intelligence on-device. Those are steps toward a bigger goal. When intelligence becomes lossless and continuous, it compounds, it improves faster, and eventually superintelligence becomes available to everyone. We'll measure our success by whether we build the infrastructure that gets intelligence to its next stage and make it available to all.”
document fetched 2026-11-14 03:02 · canonical https://blocksandfiles.com/ai-ml/2026/11/13/graphon-labs-raises-35m-for-persistent-graphon-intelligence-memory-it-hopes-will-accelerate-progress-towards-agi/5309682