Mind Lab builds Mint Recursive to help companies train their own AI

Mind Lab's Mint Recursive platform addresses a critical shift in AI investment. Attention is moving from applications to the underlying infrastructure for large models. Investors now prioritize systems for post-training and inference. Mind Lab’s Macaron-V1.1, a 752-billion-parameter model, serves as a proof point. It was post-trained on Mint Recursive, showcasing the platform’s ability to handle complex model optimization. This reflects a broader industry need for robust, scalable AI infrastructure.

The platform's focus on LoRA (low-rank adaptation) and continual learning offers a direct benefit for Chinese enterprises. Mind Lab’s approach enables companies to develop domain-specific models, covering the 70-80% of use cases general models miss. This reduces reliance on generic models and allows companies to 'own their own intelligence,' as Fireworks AI CEO Lin Qiao puts it. This will drive more localized and specialized AI applications across China.

The thing to watch is how quickly Mint Recursive's self-improvement process matures. While humans are still involved, the platform’s ability to record failed interactions and convert them into new training data is key. This could accelerate model development cycles for enterprises. The true test will be its ability to automate more of this feedback loop in the next 12 to 18 months, reducing human intervention and operational costs.