Iteris is now open source for researchers, engineers, and students working on computational mathematics.
Iteris is an agentic research system for open problems in computational mathematics. It coordinates numerical evidence, construction search, proof drafts, review, and durable project state in one auditable workflow.
Use the iteris command to create projects, start or resume runs, and verify results.
Keep project structure, fact graphs, task boards, checkpoints, and logs visible across long runs.
Switch between Codex and Claude Code executors with ITERIS_EXECUTOR or CLI flags.
Three entry points for running Iteris projects
The open-source release focuses on practical interfaces for starting projects, managing active runs, and inspecting generated research state.
Command-line orchestration
Use the iteris command to create projects, start or resume runs, inspect status, manage checkpoints, run verification, and call tools.
iteris monitor
One command for setup, project creation, progress tracking, run management, and Q&A through structured menus and lookups.
Visual inspection for long runs
Inspect runs, fact graphs, task boards, project structure, and verification progress without digging through raw artifacts.
Built for repeatable research workflows
This release adds practical improvements for faster runs, broader branch exploration, and executor choice.
Optimized harness and execution engine
The execution framework and run orchestration were refactored for shorter wall-clock time and lower token usage on comparable tasks.
Evolve module
Iteris can generate, manage, and explore multiple branches on one problem direction, including extrapolations, variants, and extensions.
Claude Code and Codex support
Switch backends through environment variables or CLI flags to match each model’s strengths to the task.
Community
We encourage researchers to leverage Iteris for more open problems in computational mathematics. We will continuously publish updates on our advances in AI-driven frontier problem solving and Iteris harnessing techniques. We also offer technical support for large-scale experiments and welcome inquiries about potential collaborative work.
Contact us: chenlh@pku.edu.cn