About Rhylthyme
Rhylthyme (“real time”) is a markup language and runtime for complex schedules.
The gap it fills
A cook bringing four dishes to the table together, a technician running a protocol whose incubations overlap, a stage manager calling cues and a trainer running supersets all face the same structure: several lines of work at once, dependencies between them, durations that are sometimes fixed and sometimes open-ended, and a few shared resources that only so many steps can use at a time.
Scheduling software mostly comes in three kinds, and none of them fits:
- Planners compute a schedule and hand over a static chart.
- Workflow engines run a dependency graph on machines, with no person in the loop.
- Timers and checklists walk a person through steps, with no idea of parallel work or shared equipment.
Rhylthyme is built for multi-track schedules with capacity limits, followed live by a person, where some steps end only when that person says so.
How it works
A Rhylthyme program is a plain JSON file. It describes parallel tracks of sequential steps. Each step has a duration (fixed, variable or open-ended) and a trigger that says when it starts: at the beginning, after another step, a set time before another step ends, or when someone presses a button. Resource constraints say how many steps may use the oven, the thermocycler or the squat rack at once.
Around that file there is a set of tools that agree with each other:
- a validator that catches schedules which cannot work, and an analyzer that reports total length, the chain of steps that sets it, and resource conflicts;
- players that run the schedule live as a timeline and an itinerary: this web app, a terminal runner, and the iPhone and iPad app;
- importers that turn recipes, lab protocols and run sheets into programs;
- a server for AI assistants (the Model Context Protocol), so that an assistant writes, checks and publishes a schedule through tool calls, and the validator refuses a broken one before anyone sees it.
The paper
Rhylthyme: A Declarative Language and Agent-Native Runtime for Human-Executed, Resource-Constrained Schedules
Jeremy Leipzig. Preprint, 2026. CC BY 4.0.
The paper describes the language, the tools and the assistant interface, and sets them against project scheduling, air-traffic flow management, temporal constraint networks, robot task scheduling, kitchen and laboratory schedulers, and workflow engines. Its claim is a narrow one: capacity limits and cross-track dependencies inside a timeline that a person can follow and control.
It also tests how well language models write these schedules from a text description. A structured, four-turn prompt was compared with a single message on 24 expert-written programs from kitchens, laboratories, events and fitness:
What remained hard was not the shape of the schedule but time: getting the durations and offsets right.
Who makes it
Rhylthyme was written by Jeremy Leipzig, PhD. He has spent more than 25 years between genomics research and software, with degrees in biology, computer science and information science, and more than 40 publications.
Rhylthyme is descended from a C++ program Leipzig wrote in 1998 that was used to coordinate rodent experiments.
More about his research, software and talks is on his profile and its Rhylthyme case study.
Feel free to reach out to him at leipzig@rhylthyme.com if you have interest in using Rhylthyme in your work.
Where to find it
- rhylthyme.com, with sites for the kitchen, the lab, events and the gym
- Rhylthyme for iPhone and iPad
- Documentation
- Connect an AI assistant
- Source code: the specification, the runner, the timeline engine and the examples
- Slack community
- support@rhylthyme.com