The Up Next Engine: How FlowBoard Suggests What to Work on After You Finish a Task

Introduction

You just finished a task. You feel good. You have momentum. And then... you stop. You scroll through the backlog, scanning dozens of items, trying to figure out what deserves your attention next. By the time you decide, five minutes have passed and your flow state is gone.

This is the "what now?" problem, and it happens every single time a developer completes a piece of work. Research on flow state shows that it takes roughly 15-25 minutes to regain deep focus after an interruption. Choosing the next task shouldn't be an interruption - it should be instant.

That's exactly what the Up Next engine does. The moment you mark a task as done, FlowBoard analyzes your completed work and surfaces up to three related tasks you could tackle immediately. No scrolling, no scanning, no decision fatigue. Just click and keep going.

How It Works

The Up Next engine runs a scoring algorithm against all your open tasks the instant you complete one. It looks at multiple signals to determine which tasks are the most natural continuation of the work you just finished.

1. Task Links (Strongest Signal)

When tasks are linked together, the engine knows they're related. This is the most powerful signal:

• Related links get the highest score - these are tasks you explicitly connected as part of the same effort
• Blocked-by links score high too - if you just unblocked something, it's probably ready to work on
• Any other link still gets a meaningful boost

The engine checks links in both directions. If Task A links to Task B, completing either one will surface the other as a suggestion.

2. Shared Tags

Tags represent the "what" of your work - a feature area, a module, a client. If the task you just finished is tagged "payments" and there are other open tasks with the same tag, they'll rank higher. Each matching tag adds to the score, so tasks that share multiple tags rise to the top.

3. Title Similarity

Sometimes tasks are related by name even when they're not explicitly linked. The engine compares meaningful words in task titles (filtering out short words) to detect overlap. Tasks like "Add email validation to signup" and "Add email validation to checkout" will score as related.

4. Same Assignee

If you were assigned the completed task, other tasks assigned to you get a small boost. This keeps suggestions focused on your own work rather than suggesting someone else's tasks.

5. Same Project

Tasks in the same project as the completed one get a minor boost. This makes sense - if you're working on the "Mobile App" project, you'll likely want to continue there.

6. Priority Score

Higher-priority tasks get a subtle boost on top of all other signals. This ensures that when two tasks are equally related, the more important one surfaces first.

What You See

After completing a task, a compact panel appears directly in your flow view showing up to three suggested tasks. Each suggestion shows:

• The task title
• A "linked" badge if the task has connections to other work
• A tag badge showing relevant context

You have two options:

• Click a suggestion to immediately start working on it - FlowBoard auto-assigns it to you, marks it as started, and opens the task details
• Dismiss the panel if none of the suggestions feel right, and pick from the backlog yourself

The panel stays visible until you either pick a suggestion, dismiss it, or complete another task (which triggers a fresh set of suggestions).

Getting Better Suggestions

The Up Next engine is only as good as the connections in your data. Here are three habits that dramatically improve suggestion quality:

Link Related Tasks

When you create a task that relates to existing work, link them together. This is the single biggest factor in suggestion quality. Use link reasons to be specific:

• Related - part of the same feature or effort
• Blocked-by - can't start until the other task is done
• Duplicate - same work, different description

Use Consistent Tags

Tags like "frontend", "api", "onboarding", or "payments" help the engine group tasks by domain. The more consistently you tag, the better the engine understands which tasks belong together.

Write Descriptive Titles

Titles like "Fix bug" or "Update thing" give the engine nothing to work with. Titles like "Fix payment webhook retry logic" or "Update checkout validation rules" create word overlap that the engine can detect.

Why Not Use AI?

You might wonder why the Up Next engine uses a rule-based scoring system rather than a machine learning model. There are three reasons:

• Speed - suggestions appear instantly because the algorithm runs locally with no API calls. There's zero latency between completing a task and seeing what's next.
• Transparency - you can understand why a task was suggested. It's linked, it shares tags, it's in the same project. There's no black box.
• Privacy - your task data never leaves your browser for suggestion purposes. No external service sees your backlog.

The engine is intentionally simple. It doesn't try to predict your behavior or learn patterns over time. It looks at concrete relationships in your data and surfaces the most connected tasks. This keeps suggestions predictable and trustworthy.

The Flow State Payoff

The real value of Up Next isn't saving five minutes of scrolling. It's preserving your flow state. When you finish a task and immediately see three relevant options, you can make a split-second decision and keep moving. There's no context switch, no backlog scanning, no "where was I?" moment.

Over the course of a day, a developer might complete 5-10 tasks. If each transition costs 5 minutes of decision-making and 15 minutes of regaining focus, that's up to 3 hours lost to transitions alone. Up Next compresses each transition to seconds.

Combined with FlowBoard's continuous flow model and priority formulas, the Up Next engine ensures that your most valuable work is always one click away.

Summary

The Up Next engine bridges the gap between finishing one task and starting the next. By analyzing task links, shared tags, title similarity, assignee, and project context, it surfaces up to three relevant suggestions the moment you complete your work. The result: less time deciding, more time building.