Projects, programs, and tasks

There are three nouns to remember for organizing work in Jootle: projects, programs, and tasks. Each one has a precise meaning. Used right, they keep things organized for you and legible for your AI. Used carelessly, you end up with a long flat list of “stuff” and your AI has to guess at structure that should have been obvious.

The good news is that the distinctions are clean.

Quick definitions

A project is a piece of work with a defined start, a defined end, and a deliverable. “Renovate the kitchen.” “Launch the v2 of the product.” “Plan Jamie’s birthday party.” Projects end.

A program is an ongoing container for work that doesn’t end. “Marketing.” “Household operations.” “Client X.” Programs are arenas, not events. Inside a program you may run many projects over time.

A task is a single unit of work that needs to be done. “Email the contractor.” “Draft the press release.” “Reconcile last month’s expenses.” Tasks live inside projects, or inside programs, or sometimes on their own.

If you remember nothing else: projects end, programs don’t, tasks are the smallest unit.

Which one am I starting?

When you find yourself about to make something new, ask “does this finish?”

• If the thing finishes when a specific deliverable lands, it’s a project. Start a project.
• If the thing is “I’m responsible for keeping this area running”, it’s a program. Start a program.
• If the thing takes an hour or two and produces no further work, it’s a task. Just ask your AI to do it, or add it to your tasks.

A few examples to anchor the distinction:

What you said	Likely shape
”I’m redoing the kitchen this summer.”	Project (ends when the kitchen is done)
“I run marketing for my consulting practice.”	Program (doesn’t end)
“Run a campaign next month for the new offering.”	Project (inside the marketing program)
“Send Lily this week’s review.”	Task
”Keep my inbox triaged daily.”	Goal (inside whichever program owns it). See Playbooks and goals.
”Plan our wedding.”	Project (very large; will contain sub-projects)
“Manage our household.”	Program (will contain many projects: vacations, renovations, big purchases)

How they nest

A program can hold many projects. A project can hold many tasks. Tasks can also live directly under a program for the ongoing operational work that doesn’t fit in any specific project.

So you might have:

Household (program)
├── Kitchen renovation (project)
│   ├── Get contractor quotes (task)
│   ├── Pick countertop material (task)
│   └── Final walkthrough (task)
├── Beach trip 2026 (project)
│   ├── Book rental (task)
│   └── Make packing list (task)
└── Pay bills monthly (recurring task, no project)

You don’t have to set this up by hand. When you ask your AI to start a project or a program, it offers to thread the right pieces together. You can also let things stay flat and reorganize later. Programs don’t have to exist for projects to work; they just help when you have many related projects.

Screenshot needed

Suggested image: The Projects view in the sidebar, with a small mix of projects and programs visible (no real personal data needed; a demo instance works fine). The goal is to show what the hierarchy looks like at a glance. Crop ~800px wide.

What lives inside a project

When you open a project in the web app, you get a workspace. The workspace has multiple views and you can switch between them:

• Overview. Status, dates, owner, the people involved, the summary your AI keeps up to date.
• Tasks. The individual units of work, with status, owner, and dependencies.
• Files / Artifacts. The documents the project has produced. (See Artifacts.)
• Chat. A project-scoped chat with your AI, where everything you say is already in the project’s context. You don’t have to keep saying “for the kitchen project”.
• Activity. What’s been happening, who’s done what, when.

Programs have a similar workspace, with the same kinds of views, scoped to the program.

What “chat scoped to a project” actually means

This is the feature you’ll notice immediately.

When you open the chat inside a project, every message you send is automatically associated with that project. You don’t say “for the kitchen”. You just say “what did Mike quote us for the countertop install?” and your AI knows which Mike, which countertop, which budget.

If you ask “remind me what I decided about the dishwasher”, the answer is drawn from that project’s history, not from your entire Jootle.

If you ask something unrelated (“what’s on my calendar tomorrow?”), your AI handles it cleanly. It doesn’t refuse because you happen to be in the project view. It just answers and stays scoped.

How your AI routes messages to projects

When you’re chatting outside a project view (web home, Telegram, email), your AI has to figure out which project a message belongs to. It uses three signals, in this order:

1. Explicit context. If you’ve been chatting in a project view, that’s the project until you switch away.
2. Keyword matches. If you mention the project name or a unique identifier in your message, your AI routes to that project.
3. Semantic similarity. If the wording of your message clearly relates to a project (mentioning the people or details in it), your AI infers the right one.

If your AI is unsure, it asks. (“Is this for the kitchen project or the bathroom project?”) Be glad it asked, because the alternative is silent misfiling.

Limit, honestly

Today, the routing uses keyword and semantic signals but does not yet consult the knowledge graph for routing. So if you say “tell Jason I’ll have an update next week”, and Jason has a project but you didn’t mention the project name, your AI may have to ask. We’re improving this. Until then, naming the project the first time you bring up a thread saves a question.

When to start a project vs. just keep going

A common mistake: people delay starting a project because it “isn’t a real thing yet”. Don’t.

Starting a project does three useful things:

1. It tells your AI to treat the thread as a coherent body of work, so context accumulates in the right place.
2. It gives you (and anyone else on your instance) a place to find what was decided.
3. It lets your AI produce artifacts and run playbooks scoped to the project, instead of everything ending up in your general chat.

If you’re not sure something rises to “a project”, start one anyway. You can delete it. The cost is zero.

When to convert a project to a program (or vice versa)

Sometimes you start a project, and partway through realize the work is going to keep going forever. (You set out to “redo the website” and a year later there’s a steady stream of pages and campaigns.) That’s a sign you want a program. Ask your AI to convert it. The project’s contents become the program’s starter set, and you can run new projects underneath.

Going the other way is rarer. If a program you started has converged on a single deliverable and won’t keep generating new work, you can wrap it up as a project. Most programs don’t, though, because their reason for existing is the ongoing nature.

A note on tasks that don’t belong anywhere

It is fine to have tasks that aren’t in any project or program. Ask your AI to “remind me to take out the recycling on Sunday” and a task appears, unattached. The web app shows these in your task list.

The reason to put a task under a project is so the context stays together. The reason not to is when the task is genuinely one-off and putting it under a project would just be ceremony. Use judgment. Your AI doesn’t care which way you go.