Couch-2-5k

Herculean is perfect for scheduling personal regimes such as regular exercise or personal challenges, goals or accomplishments. As it’s built for the individual it’s ideal for managing the routine and intensity required for the individual to plan, track and complete their goals.

In this example we set up a full Couch-2-5k plan over 16 weeks, including the objective to find and attend a 5k run in Bristol, UK at the end of the training. Yes, real research into real runs are part of our full project plan.

The prompt was passed to OpenAI Codex running GPT-5.6 Luna on High reasoning settings. As a straight forward plan and task a simple, disgustingly cheap model like Luna can easily create the full project plan and link to the easily available research.

Our Prompt

Create a realistic Couch to 5K project for someone who currently does very little running and wants to be able to complete a continuous 5K comfortably after roughly nine weeks.

Build a progressive plan that averages out to three running sessions per week, appropriate recovery/rest between sessions, simple mobility and preparation work and weekly progression milestones. Use recurring tasks or recurrence where it makes sense rather than manually creating repetitive generic tasks.

Include initial preparation such as choosing suitable running shoes, selecting safe local routes, deciding how progress will be tracked and optionally choosing/registering for a 5K event near the end of the programme. The location of the runner is Bristol, England. Research the appropriate running events available at 16 weeks from the current date of 24/9/2026 and add them into the task.

Keep the project approachable for a beginner and focused on habit-building and completion rather than competitive performance.

Resulting board

Within 3m 8s the AI had integrated with Herculean and produced a full project with 20 tasks, all dependencies and 3 regular recurring training runs, each set to recreate once the prior training run had completed.

Additionally, the dependencies were fully linked allowing for a full dependency view starting with the initial work setting up the schedule, to mapping each milestone and linking the dependencies between booking the end run and attending the event.

Recurring tasks were added to repeat runs. They’re set to allow for you to complete each A, B or C run independently rather than sticking to a strict calendar event. If the beginner slips on a run or completes one early then the board will adjust dynamic rather than continuously stack up reminders.

In the final move, our AI has created a task to determine which of the races we’re going to enter, having researched and implemented the various options available in Bristol, UK around 14th January 2027.

The Advantage

This example took less than 5 minutes to generate the full Couch-2-5k training plan using a budget model. All the research was done for you and a full plan including your milestones set up in Herculean with the appropriate dates and deadlines made available.

Now on the plan you can move your milestones over to done as you hit them, track your runs as individual tasks whilst the recurring tasks keep your backlog topped up with your next set of runs.