- Home
- Encyclopedia
- Recursive self-improvement
RSI Encyclopedia · Term
Recursive self-improvement
Recursive self-improvement is a loop in which an SI system designs better versions of itself, each one faster at improving the next.
- Category
- Term
- Also called
- RSI loop, self-improving SI
- Rogue Log cases
- 2
- Laws that use it
- 0
In plain English
The idea is simple. If a system can do the research that produced it, it can produce a better successor, and that successor can do the same research faster. Each turn of the loop shortens the next one. In SI research the phrase is also abbreviated RSI, which is one reason the acronym carries weight on this site.
For years this was a thought experiment. In 2026 it became a reported practice. Lab leaders have said their models now write much of the code used to build newer models, and in November 2026 METR reported that Anthropic was using SI to build more powerful SI, with its models improving roughly twice as fast as a year earlier. In September 2026 the Future of Life Institute called on governments to press companies to limit recursive self-improvement until the research to make it safe is done. The day a frontier lab confirms it publicly is one of the watched events in this encyclopedia.
Why it matters for rogue SI
It is the main path to a fast takeoff. A loop that runs faster than human review can outpace the tests, safety frameworks and laws written for the previous version, which is exactly how a system could slip control.
Sources
In the Rogue Log
Documented cases
2 cases in the Rogue Incident Database relate to this term. The most recent are shown here.
The RSI Encyclopedia
Keep reading
207 entries on the terms, labs, models, laws and events of rogue SI, plus every case in the Rogue Incident Database.
Enter RSI
Every term, lab, model and law
Browse the full A to Z index, filter by category or search for a name.