What changed
Mirendil Inc., a startup focused on accelerating AI development, announced it has raised $1 billion in early-stage funding. The Series A round was led by Kleiner Perkins and Clearlake Capital, with participation from over a half-dozen other investors, including Google LLC.
What this means in practice
The funding will support Mirendil’s efforts to improve its platform, Lor, which automates tasks in AI model training by distributing work among hundreds of AI agents. The company claims that Lor can streamline the trial-and-error process of developing large language models (LLMs) by automatically identifying and fixing errors during evaluation.
Mirendil’s automation technology enabled a single researcher to build a coding model, Mirend, without assistance from colleagues. The project, which cost $10 million and took 3.5 months, was completed by agents that wrote over 99% of the code, set up infrastructure, and refined the model’s architecture.
The company also unveiled Fern, a specialized coding model based on Mirendil-Base, which has 100 billion to 500 billion parameters. Fern can process 10 million input tokens per prompt, surpassing the capabilities of Claude Opus 5.5.
Limitations and availability
Mirendil plans to use upgraded versions of Mirendil-Base, with parameters ranging from 500 billion to 10 trillion, to develop more capable LLMs. The company previewed these models late last month, but they are not yet available for public use. The supplied announcement does not specify the exact release timeline or licensing conditions for these models.
