Skip to main content

​
The Problem

You sit down to work and face 50 open tasks. Some have due dates. Some are tagged “urgent”. Some have been sitting for weeks. Where do you start? Traditional task managers show you a list sorted by due date or priority. You’re on your own to figure out what actually matters.

​
The Solution

Stardust’s AI doesn’t just sort tasks—it understands your work. It analyzes urgency, impact, context, and risk to surface what you should work on right now.
Context-aware prioritization: AI sees patterns across all your tasks, not just individual fields.

​
How It Works

​
1. Multi-Factor Analysis

When you run stardust tasks prioritize, AI evaluates every task across four dimensions:
What it measures:
  • Due dates and deadlines
  • Dependencies blocking other work
  • Time sensitivity of the work
Examples:
  • Task due today → higher urgency than task due next week
  • Task blocking 5 others → higher urgency than standalone task
  • Security patch → higher urgency than UI polish
What it measures:
  • Explicit priority level (critical, high, medium, low)
  • Business value and outcomes
  • Number of users or systems affected
Examples:
  • Production outage → higher impact than feature request
  • Customer commitment → higher impact than internal improvement
  • Core feature → higher impact than edge case
What it measures:
  • What you’re already working on
  • Your work patterns and history
  • Related tasks and themes
Examples:
  • Continuing work in the same area → lower context switch cost
  • Switching between unrelated tasks → higher friction
  • Building on recent work → momentum advantage
What it measures:
  • How long tasks have been stuck
  • Likelihood of getting blocked
  • Patterns from past tasks
Examples:
  • Task stuck in progress for 2 weeks → needs attention
  • Task requiring external dependency → higher risk
  • Similar tasks failed before → elevated risk

​
2. Pattern Recognition

AI learns from your work patterns:

Completion Patterns

Which tasks do you finish quickly vs. which drag on?

Context Switches

When do you switch contexts and how does it affect velocity?

Blocking Patterns

Which types of tasks tend to get blocked?

Time Estimates

How accurate are your implicit time expectations?

​
3. Smart Recommendations

AI doesn’t just rank tasks—it explains why: 
🔥 Start Now
│ Fix production authentication timeout (task-456)
│
│ Reason: Critical priority, affects all users, blocking
│         3 other features. High urgency, high impact.
│
│ Estimated effort: 2-4 hours based on similar fixes

⏰ Do Next
│ Review Q1 roadmap with leadership (task-123)
│
│ Reason: Due tomorrow, dependencies on quarterly planning.
│         You have context from last quarter's review.
│
│ Estimated effort: 1 hour

💡 Consider Later
│ Refactor authentication service (task-789)
│
│ Reason: No deadline, medium priority, requires focused time.
│         Better done after current sprint.
│
│ Estimated effort: 1-2 days

🗑️ Consider Deleting
│ Reorganize component file structure (task-234)
│
│ Reason: Created 3 months ago, low priority, no progress.
│         Ask: Does this still matter?
AI explains its reasoning. You’re not blindly following recommendations—you understand why they make sense.

​
Using AI Triage

​
Daily Planning

Start your day with AI recommendations:
1

Run prioritization

stardust tasks prioritize
AI analyzes all your tasks and surfaces top recommendations.
2

Review recommendations

Look at the Start Now section. These are tasks with:
  • High urgency or impact
  • Clear ownership (you)
  • No blocking dependencies
  • Context you already have
3

Pick one task

Don’t try to do everything. Pick the top recommendation and start:
stardust task update task-456 --status in_progress
4

Delete the noise

Review the Consider Deleting section. If tasks have been sitting for weeks with no progress, question whether they matter:
stardust task delete task-234
The best work is the work you don’t do.

​
When Overwhelmed

AI helps when you have too much on your plate: 
# Get AI recommendations for your work
stardust tasks prioritize --owner @me

# Focus on critical items
stardust tasks prioritize --priority critical,high

# See what's blocking other work
stardust tasks prioritize --show-blockers
Example output: 
⚠️ You have 12 tasks in progress

AI Recommendation: Finish something before starting more.

Top 3 closest to done:
1. Update API documentation (80% complete)
2. Fix dashboard loading spinner (90% complete)
3. Review authentication PR (awaiting final approval)

Consider pausing these:
- Refactor cache layer (can wait)
- Explore new UI framework (not urgent)
Limit work in progress: AI will warn you when you’re overcommitted. Trust the recommendation and finish tasks before starting new ones.

​
Team Triage

Managers can use AI to understand team health: 
# Get team-wide priorities
stardust tasks prioritize --team engineering

# Find blocked work
stardust tasks prioritize --status blocked

# Identify at-risk tasks
stardust tasks prioritize --risk-level high
Example output: 
Team Health Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

⚠️ High-Risk Items (3)
│
│ Authentication service migration (alice@example.com)
│ → Stuck in progress for 14 days, blocking 7 tasks
│ → Recommendation: Break into smaller pieces or add help
│
│ Database schema update (bob@example.com)
│ → Blocked on DevOps for 5 days
│ → Recommendation: Escalate blocker or find workaround

💡 Quick Wins Available (5)
│ Tasks that are low-effort, high-impact, and ready to ship

🎯 Top Priority (2)
│ Critical items that need immediate attention

​
Advanced Features

​
Custom Scoring

Adjust how AI weighs different factors: 
# Emphasize urgency over impact
stardust tasks prioritize --weight-urgency 2.0

# Focus on quick wins
stardust tasks prioritize --prefer-quick-wins

# Minimize context switching
stardust tasks prioritize --group-by-area

​
Time-Based Recommendations

Get recommendations based on available time: 
# What can I do in the next hour?
stardust tasks prioritize --time-available 1h

# What can I finish before end of day?
stardust tasks prioritize --time-available 4h

# What's realistic for this week?
stardust tasks prioritize --time-available 40h

​
Historical Analysis

Learn from past work patterns: 
# See how accurate AI predictions have been
stardust analyze accuracy

# Identify chronic bottlenecks
stardust analyze bottlenecks

# Find patterns in completed work
stardust analyze velocity
Example output: 
Velocity Analysis
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Tasks completed last 30 days: 47
Average time to complete: 2.3 days
Median time to complete: 1 day

Fastest completions:
- Bug fixes: 4 hours average
- Documentation: 1 day average
- Code reviews: 2 hours average

Slowest completions:
- Refactoring: 5 days average
- Migrations: 7 days average
- Cross-team features: 12 days average

💡 Insight: You complete 3x more tasks when working
   on a single area vs. context switching daily.

​
Trust But Verify

AI recommendations are powerful, but you remain accountable:

Override When Needed

AI doesn’t know everything. If you have context it doesn’t, trust your judgment.

Provide Feedback

Mark recommendations as helpful or not. AI learns from feedback.

Question Patterns

If AI consistently recommends wrong, investigate why. Maybe priorities are unclear.

Stay Accountable

AI accelerates decisions, but you own outcomes.
AI as accelerator, not autopilot: Stardust helps you make better decisions faster. You’re still the pilot.

​
Privacy & Transparency

AI analyzes:
  • Task metadata (title, priority, due dates, status)
  • Work patterns (completion times, context switches)
  • Dependency relationships
  • Team interactions (who works with whom)
AI does not access:
  • Task descriptions or comments
  • External communications (Slack, email)
  • Personal information beyond work patterns
  • Basic prioritization: Local processing, no external API calls
  • Advanced analysis: Claude 4 Sonnet via Anthropic API
  • Historical patterns: Local database, never shared
All API calls are encrypted and ephemeral. We don’t store or train on your data.
Yes. AI features are optional:
# Disable AI globally
stardust config set ai.enabled false

# Use traditional sorting
stardust tasks list --sort priority,due-date
Stardust works perfectly fine without AI—it just won’t provide smart recommendations.

​
Examples in Action

​
Example 1: Overwhelmed Engineer

Situation: Sarah has 23 tasks in progress, nothing feels urgent, but she’s stressed. AI Analysis: 
stardust tasks prioritize --owner sarah
Recommendations: 
⚠️ 23 tasks in progress is too many

🗑️ Delete These (8)
- Tasks with no progress in 30+ days
- Low priority with no due date
- Vague tasks with unclear scope

⏸️ Pause These (10)
- Can wait until current sprint ends
- Blocked on external dependencies
- Nice-to-haves without urgency

🎯 Focus Here (5)
1. Fix customer-reported auth bug (critical, due today)
2. Complete API migration (blocking 4 tasks)
3. Review security PR (due tomorrow)
4. Ship dashboard feature (90% done)
5. Update deployment docs (quick win)
Outcome: Sarah deletes 8 tasks, pauses 10, focuses on 5. Stress drops. Velocity increases.

​
Example 2: New Team Member

Situation: Alex just joined the team and doesn’t know what matters. AI Analysis: 
stardust tasks prioritize --owner alex --context-level beginner
Recommendations: 
👋 Onboarding Recommendations

Start Here (Low-Risk)
- Fix typo in documentation (good first issue)
- Add test coverage for utils (isolated work)
- Update README with setup steps (no dependencies)

Learn These Areas
- Review authentication codebase (observational)
- Pair with Bob on API refactor (shadow work)

Avoid For Now
- Database migration (high-risk)
- Production deployment (needs context)
- Cross-team coordination (need relationships)

💡 AI Insight: Focus on isolated, low-risk tasks for
   your first 2 weeks. Velocity will come with context.
Outcome: Alex makes progress without fear of breaking things. Builds confidence and context.

​
Example 3: Sprint Planning

Situation: Team needs to plan next sprint but has 50 backlog items. AI Analysis: 
stardust tasks prioritize --team backend --time-available 80h --sprint-mode
Recommendations: 
Sprint Capacity: 80 engineer-hours (4 people × 20h)

Recommended Sprint (70h planned, 10h buffer)

Must-Haves (40h)
- User authentication migration (20h, alice+bob)
- Fix critical performance regression (8h, carol)
- Ship webhook API (12h, dave)

Should-Haves (20h)
- Add rate limiting (8h, alice)
- Update API documentation (6h, bob)
- Dashboard performance improvements (6h, carol)

Nice-to-Haves (defer to next sprint)
- Refactor cache layer (no urgency)
- Explore GraphQL (exploratory)
- UI component library (blocked on design)

⚠️ At Risk
- Authentication migration is complex, may take longer
- Consider splitting if estimates slip
Outcome: Team commits to realistic work, leaves buffer, ships on time.

​
Best Practices

Run Daily

Use AI prioritization every morning as part of your planning routine.

Trust the Data

If AI says you’re overcommitted, you probably are. Act on recommendations.

Provide Context

Update task descriptions when AI misunderstands. It learns from corrections.

Question Everything

If a task keeps appearing in “Consider Deleting”, ask why it exists.

​
Next Steps

Task Management

Learn all task features and capabilities

Common Workflows

See how teams use AI triage in practice

Integrations

Connect AI with Slack and other tools

API Reference

Build custom AI-powered workflows