# Long-Horizon Task

> A goal that requires many dependent steps, decisions, tool calls, and course corrections before completion.

- Category: Agents & Autonomy
- Stage: Rising
- Also called: Long-running task, Extended task
- Canonical: https://www.joelcsaji.com/ai-dictionary/long-horizon-task

## Plain English

It is work that cannot be solved in one neat prompt, such as migrating an app or researching a market.

## Example

An agent spends hours modernizing a codebase, running tests, fixing regressions, and documenting the result.

## Why it matters

Long-horizon ability is becoming a major benchmark for whether agents can handle valuable real-world work.

## Related

- [Agent Harness](https://www.joelcsaji.com/ai-dictionary/agent-harness)
- [Long-Running Agent](https://www.joelcsaji.com/ai-dictionary/long-running-agent)
- [Compaction](https://www.joelcsaji.com/ai-dictionary/compaction)

## Sources

- [Anthropic: Effective Harnesses](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents)
- [Stanford HAI: 2026 AI Index](https://hai.stanford.edu/ai-index/2026-ai-index-report)