Semantics become evidence
SageMath behavior is a specification to test, not an impression to imitate. Compatibility claims are backed by executable examples, regression suites, and explicit scope.
Active research / 2026
Sage.js
I am working with Codex to rewrite Sage from scratch: a vastly cleaner open-source system intended to give computational agents the best possible foundation for doing serious mathematical research.
The research question
What should Sage be if we designed it now?
SageMath proved that a broad, free mathematical system could support research and teaching. Sage.js asks how to preserve that mission while changing the architecture around a new reality: agents can now read, write, test, benchmark, and document substantial mathematical software.
The goal is not just faster code generation. It is a system whose mathematical contracts are explicit enough that agent work can be trusted: claims have evidence, compatibility has a measurable scope, and difficult algorithms can be checked against independent implementations.
Design commitments
Software built for verifiable progress
The live dashboard deliberately separates whether a capability exists, how broad it is, and how strongly its behavior has been certified.
Independent mathematical truth
PARI/GP, Magma, Oscar, Hecke, Nemo, and other systems provide independent semantic and performance oracles instead of a single circular reference implementation.
Readable by agents and mathematicians
Algorithms, interfaces, and documentation are structured so a person or an agent can locate a gap, understand the contract, implement it, and verify the result.
Research software everywhere
Node, native executables, Jupyter, and WebAssembly make the same open mathematical system available in terminals, notebooks, servers, and browsers.
Continuity
The same threads, in a new system
Sage.js brings together work that previously appeared as separate parts of my career.
Explicit mathematicsModular symbols, elliptic curves, number fields, and exact algebra provide demanding research-level test cases.Open mathematical softwareThe original SageMath commitment remains: powerful mathematics should not depend on closed systems.Collaborative computationCoCalc made serious computation available through browsers and shared documents; Sage.js makes the mathematical runtime native to that world.Agent collaborationStable gap identifiers, executable examples, differential tests, and readable APIs turn the roadmap into parallel, checkable work.