# The Coming Loop

> **TL;DR.** The Coming Loop is Armin Ronacher's term for the outer harness system that orchestrates AI coding agents — deciding whether work is complete, injecting follow-up messages, or routing the task to another machine.

- **Category:** AI / Developer Tools / Philosophy
- **Stage:** nascent
- **Age:** 2 days
- **Origin date:** 2026-06-23
- **First detected:** 2026-06-24
- **Canonical URL:** https://earlyterms.com/term/coming-loop
- **Sources:** 7 primary URLs

## Definition

The Coming Loop is Armin Ronacher's term for the outer harness system that orchestrates AI coding agents — deciding whether work is complete, injecting follow-up messages, or routing the task to another machine. It sits above the inner agent loop (model calls tool, reads result, repeats) and above the agent harness.

Ronacher coined the term in a [June 23, 2026 essay](https://lucumr.pocoo.org/2026/6/23/the-coming-loop/) that reached 405 points and 278 comments on Hacker News within 24 hours. The essay frames the harness-level loop as a new mode of software production and warns about code quality loss, human comprehension gaps, and long-term dependency on machine participation.

## Example

A harness queues a backlog of GitHub issues; a Claude Code agent picks one up, works it, and stops. The harness evaluates the diff — if tests fail or the scope is incomplete, it starts a fresh session with modified context rather than declaring the task done. That outer decide-and-continue cycle is the coming loop.

## Analogy

The agent loop is one factory worker; the coming loop is the shift supervisor deciding what ships.

## Why it's emerging now

Armin Ronacher's June 23 essay named the harness-level loop pattern just as Boris Cherny (Anthropic), Addy Osmani (Google), and Peter Steinberger (OpenClaw) were independently pushing the same idea — that designing the loop around your agents is now more important than prompting the agents. The essay's HN front-page traction (405 points) signals practitioner recognition of a real architectural inflection.

## Related terms

- *child:* agent-loop
- *parent:* agent-harness
- *related:* context-engineering
- *related:* managed-agents
- *parent:* loop engineering
- *alias:* harness-level loop
- *related:* coding-agents
- *related:* long-running-agents
- *related:* agent-teams
- *related:* openclaw

## Sources

1. [The Coming Loop — Armin Ronacher (origin essay)](https://lucumr.pocoo.org/2026/6/23/the-coming-loop/)
2. [HN discussion: The Coming Loop (405 points, 278 comments)](https://news.ycombinator.com/item?id=48643180)
3. [The loop is already here — practitioner reply to Ronacher](https://pocoo.vaked.dev/posts/2026-06-24-the-loop-is-already-here)
4. [The AI world is getting loopy — TechCrunch (Jun 22, 2026)](https://techcrunch.com/2026/06/22/the-ai-world-is-getting-loopy/)
5. [Loop Engineering — Addy Osmani (Jun 7, 2026)](https://addyosmani.com/blog/loop-engineering/)
6. [The Anthropic leader who built Claude Code says he just writes loops now — The New Stack](https://thenewstack.io/loop-engineering/)
7. [Loop Engineering — Cobus Greyling on Medium (Jun 2026)](https://cobusgreyling.medium.com/loop-engineering-62926dd6991c)

---
_Generated by EarlyTerms · https://earlyterms.com/term/coming-loop_