Power to the Software Engineer — Co-Leading Reflection AI’s Seed and Series A Rounds
By Max Gazor and Brian Zhan
In the world of artificial intelligence, there are moments when a team’s vision and ambition stop you in your tracks. Moments when you realize that the technology they’re building has the potential to redefine industries, reshape workflows and unlock new possibilities. For us, that moment came when we first met the founders of Reflection AI. It’s because we have so much conviction in this company, this team and its vision that CRV co-led both Reflection AI’s seed and its Series A.
Reflection AI was founded by Misha Laskin and Ioannis Antonoglou, two former DeepMind and OpenAI researchers who have made an indelible mark on the field of AI. They’ve built some of the most capable systems in the world — ChatGPT, Gemini, AlphaGo and AlphaZero — and now they’re on a mission to create the most reliable coding agents. Their work is not just about incremental improvements; it’s about solving the depth problem in AI, a challenge that has eluded even the most advanced models today.
The Depth Problem in AI
Today’s LLMs are incredibly broad. They can write poetry, debug code and even draft legal documents. When it comes to depth — solving complex, multi-step tasks with precision and reliability — they often fall short. This is the problem Reflection AI is tackling.
The team’s approach is rooted in their experience building superhuman agents like AlphaGo and AlphaZero, which combined learning and search to achieve unprecedented levels of competence in narrow domains. At Reflection AI, they’re applying these principles to coding, creating agents that can autonomously handle well-scoped engineering tasks with remarkable reliability.
The Reflection Difference
What sets Reflection AI apart is their focus on depth and reliability. While most AI coding tools today are semi-autonomous, requiring developers to guide every step, Reflection’s agents are designed to operate independently. They can search codebases, plan their work, implement changes and run tests — all without human intervention.
The Founders’ Vision
Misha and Ioannis are serial pioneers in groundbreaking work. They bring a unique combination of technical expertise and visionary thinking to the table. Misha’s AI journey is a testament to the power of curiosity and resilience. Misha’s early fascination with physics led him to pursue a PhD in the field, however, it was the rise of AlphaGo in 2016, that sparked his transition into AI. Misha was captivated by AlphaGo’s ability to outperform humans while also exhibiting creativity and intuition in its gameplay. For him, this moment crystallized AI’s ability to solve complex, multi-step problems in ways humans can’t. Under the mentorship of Peter Abbeel at UC Berkeley, Misha honed his skills in deep learning and reinforcement learning, eventually contributing to the development of Gemini at DeepMind. At Reflection AI, Misha is applying his expertise in reinforcement learning and reward modeling to create coding agents that are true collaborators.
Ioannis Antonoglou is the architect behind some of the most iconic AI systems in history. As a founding engineer at DeepMind, Ioannis played a pivotal role in developing AlphaGo, the first AI system to defeat a world champion in the ancient game of Go. AlphaGo’s victory over Lee Sedol in 2016, was a watershed moment for AI demonstrating the potential of combining deep learning with search algorithms to achieve superhuman performance. One of the most iconic moments in AlphaGo’s journey was “Move 37,” a seemingly unconventional play that left even the world’s best Go players perplexed. This move showcased AlphaGo’s ability to think creatively and strategically — a hallmark of Ioannis’s approach to AI.
Ioannis didn’t stop at AlphaGo. He also developed AlphaZero, a more generalized version of the system that could master, not just Go, but also chess and shogi. AlphaZero’s ability to learn from scratch, without any human data, is a testament to Ioannis’s vision of creating AI systems that can achieve superhuman performance through self-play and reinforcement learning. At Reflection AI, Ioannis is leveraging his expertise in planning, search and reinforcement learning to build agents that can handle complex, real-world tasks with the same level of reliability and precision as AlphaGo and AlphaZero.
The Opportunity
The potential impact of Reflection AI’s technology is staggering. By automating repetitive engineering tasks, Reflection AI frees up developers to focus on creative, high-impact work. This isn’t just about efficiency — it’s about unlocking new levels of productivity and innovation.
Imagine a world where engineering teams can tackle their backlogs in days instead of months, where code migrations happen seamlessly and where cyber vulnerabilities are remediated before they become critical. This is the world Reflection AI is building.
The Future of AI Agents
Reflection AI is at the forefront of what we believe will be the next major wave of AI innovation: the rise of reliable, autonomous agents. Just as AlphaGo redefined what was possible in game-playing AI, Reflection AI is poised to redefine what’s possible in coding and beyond. This is just the beginning. As Reflection continues to push the boundaries of AI, we’re excited to see how their technology will shape the future of work, creativity, and problem-solving.
Why We’re Partnering With Reflection AI
At CRV, we’re always looking for teams that push the boundaries of what’s possible. Reflection AI is a rare company, one that combines world class technical talent with a bold, ambitious vision. When we first met Misha and Ioannis, it was clear they were onto something extraordinary. Their deep understanding of AI, combined with their relentless focus on solving the depth problem, sets them apart. The entire team at CRV is thrilled to partner with Reflection AI as they embark on this journey. Their work has the potential to transform the entire AI driven automation landscape. If you’re an entrepreneur testing the limits of what’s possible and seeking an early stage partner in your startup adventure, don’t hesitate to reach out and get to know our team.
