AIShore
archivedAutonomous sprint orchestration for Claude Code
Write a backlog with commander's intent; AI implements, validates, and merges — item by item, branch by branch, hands-off. Retired once Claude Code's own orchestration caught up.
What it was
AIShore drained a backlog of work through Claude Code autonomously: aishore run done and it implemented, validated, and merged item by item, branch by branch. Every backlog item carried a commander’s intent field — when steps or acceptance criteria were ambiguous, the agent followed intent instead of guessing.
Its core discipline was core before features: a Core Gate verified that the application boots and its primary path works before any feature item became pickable. If a sprint broke the core, AIShore auto-generated a heal item and pushed it to the front of the queue. Working code that’s reachable beats tested code that’s isolated.
Why it’s retired
Claude Code evolved fast enough to absorb the job. Subagents, task tracking, plan mode, and native multi-agent workflows now cover most of what AIShore bolted on from the outside. The right move when the platform eats your tool is to stop maintaining the tool.
What carried forward
The ideas outlived the code: intent as the north star, core-gating feature work, prove-it-runs validation. They shaped how I run agent-built projects today — including Sterling and the Markdown editor, both built largely by autonomous agents.
The repository stays up at its final working snapshot.