Aider
An open-source terminal AI pair-programming tool that is developed largely with itself; its release history reports, per release, what share of the new code aider wrote.
Aider, driven by frontier LLMs, writes most of the new code for each aider release under the maintainer's direction. Each release then becomes the tool used to build the next one, and authorship is measured per release with a git-blame count of new source-code lines.
- Aider writes new code for Aider releases
- New code becomes the next release
- Next release builds the following version
- Aider writes new code for Aider releases
- New code becomes the next release
- Next release builds the following version
Why it is a road to recursion
It is a public, per-release measurement of a coding tool writing its own next version, the human-supervised form of a harness building its successor.
Evidence
The release history states that aider wrote 92% of the code in v0.78.0 and v0.82.0, 93% in v0.79.2 and 88% in v0.86.0, and usually about 70-80% of the new code in each release.
Related loops
More self-modifying agents →- Coding agent edits its own code
- Scored child agents enter the archive
- Archived agents make later self modifications
- repeat
- Claude Code writes its own code changes
- Changes become next Claude Code versions
- Release becomes harness for next development round
- repeat
- Outer agent rewrites inner agent code
- Accepted rewrites become the incumbent agent
- Discovered agent becomes the outer loop agent
- repeat