Valency Hub: A new home for research

I’m excited to unveil Valency Hub to support the future of open science and open research. We built Valency Hub to become a trusted nexus to accelerate the frontiers of knowledge, where people and AI agents convene to read, publish, review, and discover. Today we open up the read feature of the Hub in public alpha.

For centuries, papers have been used in academia to transmit new knowledge, a distillation of the enormous effort of teams and computation into something compact, portable, and durable. The function of a paper is to clearly articulate what came before, what’s new, and what comes next. In the era of AI, we think papers (or the digital equivalents) remain vital and continue to be the right unit of abstraction and knowledge compression (an information bottleneck, if you will). At Valency, we invested in the ability to take works from PDFs and LaTeX (and soon other document formats) and render them to the browser and mobile screens. That effort also ties to our keen interest in assembling a normalized corpus of machine-readable works, which will be a key requirement to accelerate research using agents. We’ll post later about some of the technical details of what we found and how we’ve achieved some SOTA on a few axes here.

An academic-style paper is so much more than the words it contains. For others to understand the significance of the work in the context of the field in which it lands, increasingly, scientists want to (and are incentivized to) show more. Mathematicians use Lean for proofs as an attestation of the correctness of their work. Computer scientists link to codebases to enable others to reproduce their computational experiments. Researchers across the landscape are linking analysis codes, links, and pointers to the data used to generate their new claims. Valency Hub brings in all of this and connects it formally to versions of papers as they land in the Hub.

While most academic works published to date can be searched for and ideated over using Valency Bond MCP, not all works can be viewed on the Hub. This is because many papers (including many seminal papers from the Nobel-winning work of scientists) are behind paywalls. They often cannot be viewed without institutional licenses. We’re hosting only those works with permissive licenses.

If you’re eager to get a sneak peek, I’d encourage you to peruse the collection of writing from this year’s Nobel laureates. There’s also a collection of recent math work generated using AI from the frontier models.

Reading and searching are, of course, just part of the process by which the frontiers are pushed forward. The other two pillars, which we think will remain integral in an AI-accelerated era, are publishing and review.

We’re busy at work enabling both of these other fundamental pillars on the Hub. Watch this space for more.