Harrison Chase identifies the pattern that makes long-horizon agents practically useful today: “If you can find these framings where they run for a long period of time but produce like a first draft of something, those to me are the killer applications of long horizon agents right now.”
The key insight is that agents aren’t reliable enough for end-to-end autonomous execution, but they can do enormous amounts of preparatory work. The human reviews a rich artifact rather than creating from scratch.
Examples Chase cites:
- Coding: PRs are reviewed, not pushed directly to production (unless vibe coding)
- AI SRE: Traversal produces an incident report that a human reviews
- Research/report generation: “You don’t send it out to all of your followers right away. You look at it, you edit it.”
- Customer support: When first-line AI fails, a long-horizon agent runs in the background producing a comprehensive handoff report for the human agent
This is the same pattern behind Autonomous coding loops need small stories and fast feedback to work but at a higher level: the entire agent run is one “story” that produces a reviewable artifact. The human isn’t giving step-by-step feedback — they’re reviewing the finished first draft.
Connects to Verification is the single highest-leverage practice for agent-assisted coding — the first-draft pattern IS a verification pattern: agent generates, human verifies. Also connects to Autopilots capture the work budget — six dollars in services for every one in software — the first-draft pattern captures the work budget (report creation, research, analysis) while keeping humans in the verification loop.