Rodrigo Laneth

Rodrigo Laneth is a Brazilian security researcher, software developer, and infrastructure operator. They founded return moe and serve as the studio’s lead researcher, developing its characters, software, and applied artificial intelligence research.¹²

Work at return moe

Laneth’s work at return moe includes language-model evaluation, mechanistic interpretability, purpose-specific AI agents, and experiments in generative media.²³⁴ They created and originally authored Soraya, and have developed chatbot systems used to instantiate the character. Their writing distinguishes Soraya’s fictional identity from the changing prompts, models, and software used to make her interactive.⁵

They introduced Miru Tracer under the Miru name as a workbench for inspecting and intervening in language-model generation. Their other return moe research includes the Chinese Political Neutrality Benchmark and the Watermark Removal as a Denoising Task experiment.³⁴⁶

Visual representation

The infobox portrait depicts an anime-style original character that Laneth generally uses to represent themself on social media. The character has long, straight dark charcoal hair with loose bangs, large blue-gray eyes, pale skin, and a restrained smile. She wears a black skirt and blazer against an abstract blue background.²

Broader work

Laneth is also affiliated with Miralium Research and works in information security and Internet infrastructure. Their wider biography, security research, infrastructure work, and projects outside return moe are documented on their official website.¹⁷

References

Footnotes

1. Rodrigo Laneth’s official website. ↩ ↩²

2. return moe official website. ↩ ↩² ↩³

3. Miru Tracer v0.2.0: from token probabilities to model internals, return moe blog. ↩ ↩²

4. Announcing the Chinese Political Neutrality Benchmark, return moe blog. ↩ ↩²

5. Echoes in the Latent Space: Existence, Identity, and Future, return moe blog. ↩

6. Watermark removal as a denoising task, return moe blog, 21 December 2025. ↩

7. Miralium Research official website. ↩