Reflection AI, a startup founded by former Google DeepMind researchers, has launched with $130 million in early-stage funding to advance AI toward superintelligence. The company secured its funding over two rounds, beginning with a $25 million seed investment led by Sequoia Capital and CRV, followed by a $105 million Series A round co-led by CRV and Lightspeed Venture Partners. Additional investors include Nvidia’s venture capital arm, LinkedIn co-founder Reid Hoffman, and Scale AI CEO Alexandr Wang.
With a valuation of $555 million, Reflection AI aims to develop autonomous AI agents capable of performing complex computing tasks. The company’s first initiative focuses on building an AI-powered programming assistant designed to automate coding workflows, identify vulnerabilities, optimize memory usage, and enhance software reliability. Co-founders Misha Laskin and Ioannis Antonoglou emphasized that their team has contributed to major AI breakthroughs, including AlphaGo, AlphaZero, and Google’s Gemini language models.
Laskin, who led training workflows for Gemini, and Antonoglou, who specialized in optimizing large language model outputs, are now applying their expertise to developing autonomous AI agents. Reflection AI is leveraging reinforcement learning and large language models (LLMs) to streamline dataset creation, eliminating the need for manual annotations. The company is also exploring alternatives to the Transformer neural network architecture to push the boundaries of AI efficiency and scalability.
Early job postings indicate that Reflection AI plans to train its models using tens of thousands of GPUs while working on vLLM-like platforms to optimize memory efficiency for non-LLM models. As the company expands its research into next-generation AI architectures, it positions itself as a key player in the race toward superintelligent autonomous systems.
