Curie vs autoresearch
AI-driven scientific experimentation compared
Curie and autoresearch both use AI agents to run experiments, but they target different research workflows. Curie supports broader scientific and machine-learning investigations using a user's data and code, while autoresearch focuses on repeated language-model training experiments on one NVIDIA GPU.
Curie: Automate Scientific Experiments with AI Agents
Curie is a Python framework that uses AI agents to plan, run, analyze, and report on scientific experiments. It suits researchers who want to automate exploratory or machine-learning work, but its inactive development and lack of recent releases warrant caution.
autoresearch: Let AI Agents Run LLM Training Experiments
autoresearch gives an AI coding agent a small, single-GPU language-model training setup to modify and test autonomously. It suits researchers with supported NVIDIA hardware who want to explore training changes through repeated, time-bounded experiments.
| Curie | autoresearch | |
|---|---|---|
| Language | Python | Python |
| License | Apache-2.0 | — |
| Stars | 370 | 97.2k |
| Forks | 32 | 13.5k |
| Last analyzed | Oct 9, 2026 | Oct 3, 2026 |
Key differences
- Curie covers hypothesis formulation, experiment execution, analysis, and reporting across scientific discovery, system analysis, and machine-learning tasks; autoresearch focuses on modifying and testing language-model training.
- Curie accepts user-provided datasets and starter code; autoresearch centers on an agent-editable `train.py`, with data preparation and runtime utilities kept separate.
- Curie runs experiments in Docker and may take several minutes depending on the task and budget; autoresearch uses a fixed five-minute training budget per experiment.
- Curie is intended for researchers and ML practitioners seeking a broad automated first pass; autoresearch suits users exploring training changes on supported NVIDIA hardware.
- Both projects use Python and are not archived. Curie uses an Apache-2.0 license; autoresearch's license is not included in the supplied facts.
- Curie has 370 stars and 32 forks; autoresearch has 97.2k stars and 13.5k forks. These counts alone do not establish suitability or dependability.
Choose Curie if you…
- need an agent to investigate questions using your own dataset or starter code.
- want experiment analysis, reports, and recorded artifacts across scientific or machine-learning tasks.
- need verification modules intended to support methodical, reproducible procedures.
Choose autoresearch if you…
- want an agent to iteratively modify language-model training code and evaluate changes.
- have a single supported NVIDIA GPU and want experiments with a consistent five-minute budget.
- want a compact training setup with an instruction file guiding the agent's research work.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.