Tilde Research

Tilde Research is a moonshot AI laboratory founded in 2024 and based in Palo Alto, California. Its work concentrates on mechanistic interpretability, novel architectures and pretraining science, with the stated aim of building a foundational understanding of how AI models function. The company has raised $8 million in seed funding.

Its principal products are interpreter models, designed to expose the inner workings of black-box AI systems and turn opaque prompting into transparent, controllable interaction; the firm reports enhanced reasoning capabilities demonstrated in open-source models. Complementing these are model control technologies that allow direct steering of model behaviour, offering performance gains beyond traditional post-training or fine-tuning; fine-grained control has been demonstrated in text-to-video generation.

The technical remit extends to interpreter models, model control, text-to-video generation and open-source AI tooling. Tilde Research publishes tools and research openly, contributing to the broader AI community. The team draws on experience from academia and leading AI infrastructure companies, and is organised around a mission-driven culture focused on deeper understanding of AI models.