Skip to content

Projecting AI Task Completion to 2030

An interactive visualization showing the exponential growth of AI task completion capabilities from 2019 to 2030, with time on the horizontal axis and task completion horizons projected into the future.

Run the MicroSim Fullscreen

To use this interactive chart on any website simply copy this line of HTML into your web page: 

<iframe src="https://dmccreary.github.io/tracking-ai-course/sims/projecting-ai/main.html" width="100%" height="588"></iframe>

This MicroSim extends the analysis from METR.org's research by projecting AI capabilities forward to 2030 based on the observed exponential growth pattern.

Overview

This MicroSim visualizes how AI task completion capabilities are projected to evolve over time, showing:

  • Historical data from 2019-2025 showing actual AI model performance
  • Time on the horizontal axis (year) for easy temporal understanding
  • Task completion horizons on the vertical axis (in minutes/hours/days)
  • Exponential trend line projecting capabilities to 2030 based on 7-month doubling time

The visualization demonstrates that AI's ability to complete longer tasks is growing exponentially, with capabilities roughly doubling every 7 months according to METR's research.

Key Insights

Exponential Growth Pattern

The data reveals a clear exponential trend with a doubling time of approximately 7 months. This means: - Every 7 months, AI models can reliably complete tasks twice as long - By 2030, if this trend continues, models could work on tasks lasting days or weeks - This growth rate is faster than many other technology trends

What the Projection Means

If the 7-month doubling time continues: - 2026: Models could reliably work on tasks lasting 10-20+ hours - 2028: Tasks lasting multiple days become feasible - 2030: Week-long autonomous task completion may be possible

Important caveat: Exponential trends don't continue forever. Physical limits, diminishing returns, or fundamental barriers could slow this growth.

Interactive Features

Time Axis (Horizontal)

  • X-axis shows years from 2019 to 2030
  • Actual model data points are shown as colored dots
  • Easy to see the temporal progression of AI capabilities

Scale Toggle

  • Linear Scale: Shows absolute differences in capabilities
  • Log Scale: Better visualizes the exponential growth pattern and makes it easier to compare models across the full range

Success Probability Toggle

  • 50% Success: Time horizon where a model has 50% chance of completing a task
  • 80% Success: More conservative metric requiring 80% success probability (typically shorter horizons)

Trend Line Toggle

  • Show/Hide the exponential trend line
  • Trend line extends to 2030 based on the 7-month doubling time
  • Helps visualize where AI capabilities might be heading

Hover Tooltips

Hover over any data point to see: - Model name - Time horizon in minutes, hours, and days - Release date - Success probability level

Color Coding

  • Green points: Frontier models (state-of-the-art)
  • Gray points: Non-frontier models
  • Blue dashed line: Exponential trend projection

Understanding the Projection

The 7-Month Doubling Time

Based on METR's analysis, AI task completion capabilities have been doubling approximately every 7 months. This is calculated from: - Historical model performance from 2019-2025 - Consistent pattern across different model families - Both 50% and 80% success probability metrics

How the Trend Line Works

The projection uses an exponential formula: 

Future Capability = Current Capability × 2^(months_elapsed / 7)

For example, starting from GPT-5's 8,239 minutes (137 hours) at 50% success: - 7 months later (March 2026): ~16,500 minutes (275 hours / 11.5 days) - 14 months later (October 2026): ~33,000 minutes (550 hours / 23 days) - By 2030: Potentially 100,000+ minutes (1,700+ hours / 70+ days)

Limitations and Caveats

This is a projection, not a prediction. Several factors could slow or stop this trend:

  1. Fundamental limits: Tasks requiring real-world feedback may hit physical time constraints
  2. Diminishing returns: As tasks get longer, new bottlenecks may emerge
  3. Measurement challenges: Defining and measuring very long tasks becomes harder
  4. Economic factors: The cost of running models for days/weeks may be prohibitive
  5. Reliability requirements: Longer tasks may require entirely new approaches to ensure robustness

Model Data

Frontier Models (Green)

The most capable models at their time of release: - GPT-5: 8,239 min / 137 hrs (50%) — Latest projected frontier model - Grok-4: 6,605 min / 110 hrs (50%) - o3: 5,531 min / 92 hrs (50%) - Claude 3.7 Sonnet: 3,254 min / 54 hrs (50%) - Earlier models: o1-elicited, Claude 3.5 variants, GPT-4 family, GPT-3.5, GPT-2

Non-Frontier Models (Gray)

Other capable models: - Claude Sonnet 4.5: 6,798 min / 113 hrs (50%) - Claude 4.1 Opus: 6,330 min / 106 hrs (50%) - Plus DeepSeek, Gemini, Qwen, and other model families

Use Cases

This visualization is valuable for:

Strategic Planning

  • Technology roadmapping: Understand when AI might handle multi-day tasks
  • Resource allocation: Plan for infrastructure to support longer-running AI agents
  • Capability planning: Anticipate what tasks AI could automate in coming years

Education & Communication

  • Illustrating exponential growth: Clear visual of how quickly AI capabilities are advancing
  • Setting expectations: Help stakeholders understand the pace of AI development
  • Trend analysis: Compare AI progress to other technological trends

Research & Analysis

  • Identifying patterns: Spot deviations from the exponential trend
  • Model comparison: See how different model families stack up over time
  • Future scenarios: Explore implications of continued exponential growth

Customization Guide

Modifying the Data

To update with new models, edit the modelData object in script.js:

const modelData = {
    frontier: [
        {
            name: 'Model Name',
            horizon50: 100.0,  // 50% success time in minutes
            horizon80: 20.0,   // 80% success time in minutes
            date: '2025-MM-DD' // Release date (YYYY-MM-DD format)
        }
    ]
};

Adjusting the Trend Line

To change the doubling time assumption, modify doublingTimeMonths in script.js:

const doublingTimeMonths = 7;  // Change this value

Extending the Time Range

To project beyond 2030, update the endDate in the calculateTrendLine() function:

const endDate = new Date('2035-12-31');  // Extend to 2035

And update the X-axis maximum:

max: new Date('2035-12-31').getTime()

Technical Details

  • Library: Chart.js 4.4.0 with date-fns adapter for time series
  • Chart Type: Scatter plot with optional trend line overlay
  • Data Points: 30 AI models from 2019-2025
  • Projection: Exponential trend to 2030 based on 7-month doubling time
  • Time Range: 2019-2030 (12 years)
  • Data Source: METR Research (March 2025)
  • Interactive Controls: Scale toggle (linear/log), Success probability (50%/80%), Trend line (show/hide)
  • Browser Compatibility: All modern browsers supporting Chart.js and ES6
  • Responsive: Adapts to container width with mobile optimizations

Interpretation Guide

Reading the Chart

  1. X-axis (Horizontal): Shows time progression from 2019 to 2030
  2. Y-axis (Vertical): Task completion time horizon (minutes/hours/days)
  3. Actual data points: Show real model performance (2019-2025)
  4. Dashed trend line: Projects future capabilities based on historical growth
  5. Higher on chart: Models can work on longer tasks

Using Different Views

Linear scale is best for: - Understanding absolute capability differences - Seeing the scale of recent improvements - Comparing models released close together in time

Log scale is best for: - Visualizing the exponential growth pattern - Comparing models across the full time range - Identifying consistent doubling patterns

Comparing Success Probabilities

  • 50% horizon: More optimistic, shows what's theoretically possible
  • 80% horizon: More conservative, shows reliable capability
  • The gap between them indicates model consistency/reliability

Real-World Implications

What These Time Horizons Mean

Current capabilities (2025): - 100-500 minutes: Complex analysis, report generation, multi-step research - 1,000-2,000 minutes (16-33 hours): Overnight project work, extensive code development - 8,000+ minutes (130+ hours / 5+ days): Multi-day research projects, complex system design

Projected capabilities (2030): - If trends continue: 50,000-100,000+ minutes (35-70+ days) - Potential applications: Month-long research initiatives, complex product development, extensive autonomous projects - New challenges: Reliability, cost, real-world interaction needs

Questions to Consider

  1. Will the trend continue? What might cause it to slow or stop?
  2. What bottlenecks emerge as tasks stretch to days or weeks?
  3. How do we measure success for such extended tasks?
  4. What safeguards are needed for AI working autonomously for weeks?
  5. What economic value comes from AI handling month-long tasks?

References

About This MicroSim

This MicroSim reimagines METR's task completion horizon chart with a temporal perspective, making it easy to see how AI capabilities have evolved and where they might be heading. By placing time on the horizontal axis and extending the exponential trend to 2030, it helps viewers understand both the historical trajectory and potential future of AI task completion capabilities.

The 7-month doubling time, if sustained, would represent one of the fastest capability growth rates in any technology domain. Understanding this trajectory is crucial for strategic planning, risk assessment, and opportunity identification in an AI-enabled future.

Self-Assessment Quiz

Test your understanding of AI capability projections.

Question 1: What is the approximate doubling time for AI task completion capabilities according to METR research?

  1. 2 years
  2. 12 months
  3. 7 months
  4. 1 month
Answer

C) 7 months - The research shows AI task completion capabilities have been doubling approximately every 7 months, one of the fastest capability growth rates observed in technology.

Question 2: If AI can handle 5-hour tasks in late 2025, what duration might be possible by early 2027 if the trend continues?

  1. 6 hours
  2. 10-20 hours
  3. The same 5 hours
  4. Less than 1 hour
Answer

B) 10-20 hours - With 7-month doubling, capabilities would roughly double twice in 14 months, going from 5 hours to approximately 10 hours to 20 hours by early 2027.

Question 3: Why is the projection visualized with time on the horizontal axis?

  1. It saves space on the chart
  2. It makes the temporal progression and future projections easier to understand
  3. Time axes are required by law
  4. It makes the chart more colorful
Answer

B) It makes the temporal progression and future projections easier to understand - Placing time horizontally allows viewers to see historical progress and future projections in an intuitive timeline format.

Question 4: What is an important caveat about exponential AI capability projections?

  1. Exponential trends always continue forever
  2. Projections are predictions, not certainties, and may be limited by physical, economic, or other barriers
  3. The projections are guaranteed to be accurate
  4. Only positive outcomes are possible
Answer

B) Projections are predictions, not certainties, and may be limited by physical, economic, or other barriers - Exponential trends don't continue indefinitely; diminishing returns, measurement challenges, and fundamental limits could slow or stop the trend.

Question 5: How should organizations use these projections for strategic planning?

  1. Ignore them because they might be wrong
  2. Plan for a range of scenarios and prepare for capabilities arriving faster than intuition suggests
  3. Wait until 2030 to start planning
  4. Assume the projections are impossible
Answer

B) Plan for a range of scenarios and prepare for capabilities arriving faster than intuition suggests - Organizations should use projections to consider multiple futures and develop flexible strategies that can adapt as AI capabilities evolve.