Self-improving AI / SICA: A Self-Improving Coding Agent
SICA: A Self-Improving Coding Agent
“We demonstrate that an LLM coding agent, equipped with basic coding tools, can autonomously edit itself, and thereby improve its performance on benchmark tasks.”
SICA: A Self-Improving Coding Agent is one of the projects on the self-improving AI map. Below is what it changes about itself, whether the change lasts, and whether it improves the improver, each with the line from its own source.
Where it sits
Columns: what it changes, from the surface to the deepest change. Rows: whether the change outlasts the run. A number is how many projects sit in that square. See it on the 3D map
- What it changes
- Tools, harness, its own code“is capable of editing its own codebase to improve itself with respect to its cost, speed and benchmark performance”
- Does the change last
- Becomes the next version“Note that since we are not performing any weight updates, this numerical score is only used to pick the next meta agent, as well as the base agent for the next iteration.”
- Does it improve the improver
- improves the task agent; no evidence the improver got better“SICA takes the best performing agent from the archive up until this point as the meta-agent, which is instructed to look through the archive, identify an improvement, and implement it.”
- Who sets the goals
- people set tasks and scoring“We define the ‘best’ performing agent by a utility function, which incorporates common desiderata for a good agent”
- How it was tested
- own benchmark numbers“We find performance gains from 17% to 53% on a random subset of SWE Bench Verified”
The agent edits its own Python codebase, for example adding diff-based file editing and an AST based symbol locator, and the best agent so far becomes the meta-agent for the next round. Model weights are not updated. The authors set the benchmarks and the utility weights; the 50 SWE Bench Verified questions behind the 17% to 53% figure are part of the benchmark set used to pick agents, and AIME 2024 and GPQA Diamond showed less improvement. No experiment measures whether the agent got better at improving itself.
Does it count as RSI?
It depends on the definition. Under the four used on What counts as RSI, by what its own source says:
- Loose (Eliezer Yudkowsky, Intelligence Explosion Microeconomics (2013)): counts
- Durable (Ren et al., self-improvement survey (2026)): counts
- Strict (Roman Yampolskiy, From Seed AI to Technological Singularity (2015)): does not count
- Autonomous (Anthropic, When AI builds itself): does not count
How it got here
Earlier work in this catalog that SICA cites, oldest first. Under each: where that work sits on the map, then where SICA sits (changes in bold), and what SICA's own paper says about it. Its whole family on the citation map
2023 Promptbreeder sets itself apart
Prompts, memory → Tools, own code · gone when the run ends → becomes the next version · improves the task agent
“OPRO ([Yang et al., 2024](https://arxiv.org/html/2504.15228v1#bib.bib37)) and Promptbreeder ([Fernando et al., 2023](https://arxiv.org/html/2504.15228v1#bib.bib13)) focus on optimizing prompts through language.”
“Since our agent can edit its entire codebase, this includes the ability to tune its own prompts to optimise the behaviour of any part of the agent.”
2024 ADAS builds on it
Prompts, memory → Tools, own code · gone when the run ends → becomes the next version · does not improve the improver → improves the task agent
“The main running loop of SICA resembles the ADAS loop [Hu et al. (2024)](https://arxiv.org/html/2504.15228v1#bib.bib16).”
“Note that this is different from ADAS, in that ADAS has a fixed meta-agent, so does not select the meta-agent from the archive (the archive in ADAS is an archive of target-agents, not meta-agents).”
2024 Gödel Agent sets itself apart
Tools, own code · gone when the run ends → becomes the next version · improves the task agent
“First, Gödel Agent ([Yin et al., 2024](https://arxiv.org/html/2504.15228v1#bib.bib40)) has specific tools (such as `action_adjust_logic` and `action_read_logic`) that allow modification of small parts of the agent as it is running.”
“Thus, it is not a general-purpose coding agent, as it is traditionally understood.”
2025 SICA
Tools, own code · becomes the next version · improves the task agent
Also cites STaR 2022, Reflexion 2023, Self-Refine 2023, Voyager 2023, STOP 2023
Cited by Darwin Gödel Machine 2025, Hyperagents 2026, Ren, self-improvement 2026, Weco AIDE² 2026, Fibona Agent 2026, LM2LM 2026, RSI Arena 2026, Pu, recursive self-improvement 2026
In the same square
Darwin Gödel Machine, Dream-RSI, Hyperagents (DGM-H), Infinity Machine, RSI-Opt, Skill∞, STOP: Self-Taught Optimizer, VISTA, Weco AIDE²