Ranked by today, not by who emailed last
Every task gets a score from your calendar, dependencies, and the work you actually finished. The top of the list is always what's next. Here's the scoring model.
Most prioritization systems quietly optimize for whoever shouted loudest. They sort by recency, by sender seniority, or by an unweighted "urgency" field a human had to fill in. None of those signals correlate with what you should actually do next.
What the ranking model uses
MetaSpark's ranker scores each task against three feature families:
- Calendar fit. How long does this task take, and is there a focus window in your calendar today that's big enough? Tasks that can't fit today get deprioritized for today.
- Dependency graph. If task A blocks task B and B is on a near deadline, A jumps. If C is waiting on someone else, it drops.
- Completion patterns. We watch which tasks you actually finished in the past four weeks. The ranker learns your real rhythm and stops boosting things you reliably skip.
Why no "urgency" field
Human-entered urgency is the most lied-about field in software. It gets used to mean "I want this done" and immediately inflates. Once everyone marks everything urgent, the field carries no information. We don't ask you to fill it in. The ranker infers urgency from the other signals.
Re-rank conditions
The ranker re-runs when any of these change:
- A calendar event is added, removed, or moved
- A dependency upstream of any task closes or is blocked
- An agent finishes a task and writes back to the source tool
- You manually close, snooze, or escalate something
The result is a list where the top is, with very high confidence, the next thing you should do. Open one tab. Start at the top.