Last year, the paper Rel: A Programming Language for Relational Data was presented at  SIGMOD/PODS International Conference on Management of Data. The paper details the main technical innovations of Rel implemented as part of RelationalAI's relational knowledge graph management system. This year, the paper was announced as one of the SIGMOD Research Highlights of 2026. 

‍

SIGMOD Research Highlights are awarded to “a set of research projects that exemplify core database research. In particular, these projects address an important problem, represent a definitive milestone in solving the problem, and have the potential of significant impact.” For the 2026 highlights, 10 papers were selected out of roughly 1500 papers from different databases conferences.

‍

Recently, Viktor Leis and Thomas Neumann wrote a technical perspective championing the Rel language, which succinctly highlights how Rel “charts a compelling path toward simpler architectures, more reusable data logic, and more principled relational systems.”

Leis and Neumann discuss the 50-year-old "two-language" status quo, and argue that perhaps the friction between declarative SQL and imperative host languages is a fundamental design flaw, rather than a necessity. The Rel language was designed and implemented with the goal of abolishing the so-called impedance mismatch. While Rel is a declarative language for relational data, it also has key functionalities that enable it to capture the semantics encoded in general-purpose imperative programming languages. By adopting a semantics-first approach grounded in Datalog and first-order logic, Rel introduces a unified model, where the relations serve as the primary abstraction. 

Furthermore, the Rel language encourages a philosophy of "growing a language", which prioritizes a small core and user-defined extensibility over the more rigid, committee-driven expansion of traditional standards. Rel provides the core tools—modularity and abstraction—that allow users to build the language outward through libraries, and other reusable components. Since the system is rooted in formal semantics, Rel supports reasoning about programs and is highly optimizable.

Overall, we are proud and thankful that the innovation behind the Rel and its dream of abolishing the impedance mismatch, prioritizing extensibility and abstraction, and relying on a semantics-first foundation, are being recognized by the SIGMOD and broader database community.  

‍