# Weco AI

> Weco is production-grade autoresearch: point it at your code and an evaluation metric, and an automated research loop (AIDE) searches, tests, and ships measurable improvements as auditable, version-controlled code.

Weco AI builds recursively self-improving AI. The `weco` CLI wraps the open-source AIDE agent with orchestration, steerability, and observability so teams can run the autoresearch loop in production — optimizing model accuracy, GPU-kernel latency, inference cost, or any metric you can measure.

## Product & Docs

- [Weco product overview](https://www.weco.ai/product): What Weco does, who it is for, and how the autoresearch loop works.
- [Full documentation](https://docs.weco.ai/): Install, quickstart, CLI reference, and guides.
- [weco CLI (GitHub)](https://github.com/WecoAI/weco-cli): The command-line tool for production-grade autoresearch.
- [AIDE (open source)](https://github.com/WecoAI/aideml): The research agent behind Weco — human-level performance on ML engineering tasks.

## Blog

- [4 Levels of Recursive Self-Improvement](https://www.weco.ai/blog/4-levels-of-recursive-self-improvement): RSI is usually discussed as a binary. We grade it instead: four levels — delegation, net positive, ignition, inflection — each the necessary condition for the next, and precise enough to be falsifiable.
- [How Weco's Autoresearch Improved a Production Fraud Detection Pipeline](https://www.weco.ai/blog/autoresearch-fraud-detection-pipeline): Vardera pointed Weco's autoresearch at their production fraud detection pipeline — F1 up 17.7%, precision up 18.9% — with every change shipping as auditable, version-controlled code.
- [An AI Agent Became the #1 Contributor in OpenAI's Hiring Challenge](https://www.weco.ai/blog/parameter-golf-aiden): Aiden spent 22 days inside OpenAI's Parameter Golf and became the competition's most influential contributor by records, citations, and public signal quality.
- [SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents](https://www.weco.ai/blog/specbench): We built SpecBench to measure the gap between code that passes visible tests and code that actually satisfies the specification. Across 30 systems-level tasks, that gap is real, measurable, and grows with task complexity.
- [AutoResearch vs Classical Hyperparameter Tuning](https://www.weco.ai/blog/autoresearch-vs-classical-hpo): We ran a head-to-head comparison on NanoChat: Optuna vs AutoResearch. The agent wins on sample efficiency, cost efficiency, and generalization.
- [Weco Now Integrates with LangSmith](https://www.weco.ai/blog/weco-langsmith-integration): Point Weco at your LangSmith datasets and evaluators. It handles the search.
- [How to Build a Good Eval](https://www.weco.ai/blog/eval-is-not-a-tool-you-install): Eval is not a tool you install. It's a way of thinking from experimental science. Learn what makes a good benchmark and how to avoid common pitfalls.
- [Evolving a Flappy Bird AI: 100 Solutions Explored, 6.6x Improvement, $12](https://www.weco.ai/blog/flappy-bird-case-study): We gave Weco a simple Flappy Bird AI that scored 2.76 points on average. 100 iterations and $12 later, the AI scored 20.9 - a 6.6x improvement.
- [Weco Is Now in Public Beta](https://www.weco.ai/blog/beyond-trial-and-error-ml-engineering): Hard work scales linearly. Automation scales exponentially.
- [Weco Raises $8M to Build Self-Evolving Software](https://www.weco.ai/blog/seed-announcement): Most AI tools can draft code that runs. None can deliver expert‑grade solutions that push the frontier.
- [AIDE: Human-Level Performance on Data Science Competitions](https://www.weco.ai/blog/technical-report): In the world of data science, Kaggle competitions have become a widely accepted standard...
- [The Future of Machine Learning Research](https://www.weco.ai/blog/intro): Research is a cornerstone in the quest to understand the world and tap into its economic values...

## Examples

- [CUDA kernel optimization](https://github.com/WecoAI/weco-cli/tree/main/examples/cuda/README.md)
- [Hello Kernel World (Colab walkthrough)](https://github.com/WecoAI/weco-cli/blob/main/examples/hello-kernel-world/colab_notebook_walkthrough.ipynb)
- [Prompt optimization](https://github.com/WecoAI/weco-cli/tree/main/examples/prompt/README.md)
- [Model development (Spaceship Titanic)](https://github.com/WecoAI/weco-cli/tree/main/examples/spaceship-titanic/README.md)
- [Triton kernel optimization](https://github.com/WecoAI/weco-cli/tree/main/examples/triton/README.md)

## Full text

- [llms-full.txt](https://www.weco.ai/llms-full.txt): Full text of every Weco AI blog post, concatenated for LLM ingestion.