I’ve been experimenting with Scheduled Tasks, and I keep wishing they had more control over where each result goes.
Right now, it feels like a task is mostly tied to one conversation. That works well for simple reminders, but it becomes awkward for tasks that run frequently over a long period of time.
What I’d really like is for a Scheduled Task to be able to create, reuse, and organize conversations inside a Project based on a few simple rules.
For example, a task could:
-
keep adding results to the same chat during the day;
-
create a new chat when a new day starts;
-
start a new chat when the topic changes;
-
create one conversation per week, report, research question, or work session;
-
keep all of those conversations inside the same Project.
So instead of choosing only between “always use the same chat” and “create a new chat every time,” the task could follow a conversation-grouping rule.
Something like:
Reuse the current conversation until a certain condition is met, then create a new one inside this Project.
That condition might be based on time, topic, task state, or some other boundary.
A learning task is one example. I might have a Project called “Continuous Learning,” with one conversation for each daily topic:
-
July 18 — Distributed Consistency
-
July 19 — Database Indexes
-
July 20 — HTTP Caching
The task could send several lessons throughout the day into the current conversation. When the next day begins, it would create a new conversation, choose a new topic, and continue there.
But the same idea could apply to many other workflows:
-
market updates grouped by trading day;
-
news briefings grouped by date;
-
system checks grouped into a daily operations log;
-
research notes grouped by topic;
-
weekly project reports;
-
language-learning sessions;
-
fitness or nutrition logs.
The important part for me is the integration with Projects.
A Project should hold the long-term context: instructions, files, preferences, goals, and knowledge about what has already happened.
The individual conversations should hold the more focused local context for a particular day, topic, report, or session.
Scheduled Tasks would then manage the timing and decide which conversation each result belongs to.
That would avoid two problems.
Keeping everything in one permanent chat eventually creates a huge conversation containing months of unrelated outputs, setup discussions, follow-up questions, and old context.
Creating a brand-new chat for every execution creates the opposite problem. A task that runs eight times per day could generate hundreds of tiny conversations.
What seems more useful is a middle layer where the task controls the lifecycle of its conversations.
When creating a task, I’d love to see options such as:
-
Project:
Continuous Learning -
Conversation behavior:
Group runs into conversations -
Start a new conversation:
Every day -
Use Project instructions and files:
Yes
More advanced rules could come later, but even basic grouping by day, week, or month would make frequent scheduled tasks much easier to manage.
To me, Projects and Scheduled Tasks feel like they should work together more deeply than they currently do.
Projects provide continuity.
Conversations provide focused context.
Scheduled Tasks should be able to organize those conversations as the workflow evolves.
Is anyone else trying to build something similar? I’d be curious how people are currently handling scheduled outputs that are too large for one permanent chat but too frequent for one new chat per run.