Reflection AI

Today we're sharing the next phase of Reflection. We're building frontier open intelligence accessible to all. We've assembled an extraordinary AI team, built a frontier LLM training stack, and raised $2 billion. Technological and scientific progress is driven by values of openness and collaboration. The internet, Linux, and the protocols and standards that underpin modern computing are all open. This isn't a coincidence. Open software is what gets forked, customized, and embedded into systems worldwide. It's what universities teach, what startups build on, what enterprises deploy. Open science enables others to learn from the results, be inspired by them, interrogate them, and build upon them in order to push the frontier of human knowledge and scientific advancement. AI got to where it is today through scaling ideas (e.g. self-attention, next token prediction, reinforcement learning) that were shared and published openly. Now AI is becoming the technology layer that everything else runs on top of. The systems that accelerate scientific research, enhance education, optimize energy usage, supercharge medical diagnoses, and run supply chains will all be built on AI infrastructure. But the frontier is currently concentrated in closed labs. If this continues, a handful of entities will control the capital, compute, and talent required to build AI, creating a runaway dynamic that locks everyone else out. There's a narrow window to change this trajectory. We need to build open models so capable that they become the obvious choice for users and developers worldwide, ensuring the foundation of intelligence remains open and accessible rather than controlled by a few. Over the last year, we've been preparing for this mission. We’ve assembled a team who have pioneered breakthroughs including PaLM, Gemini, AlphaGo, AlphaCode, AlphaProof, and contributed to ChatGPT and Character AI, among many others. We built something once thought possible only inside the world’s top labs: a large-scale LLM and reinforcement learning platform capable of training massive Mixture-of-Experts (MoEs) models at frontier scale. We saw the effectiveness of our approach first-hand when we applied it to the critical domain of autonomous coding. With this milestone unlocked, we're now bringing these methods to general agentic reasoning. We've raised significant capital and identified a scalable commercial model that aligns with our open intelligence strategy, ensuring we can continue building and releasing frontier models sustainably. We are now scaling up to build open models that bring together large-scale pretraining and advanced reinforcement learning from the ground up. There is a window of opportunity today to build frontier open intelligence, but it is closing and this may be the last. If this mission resonates, join us.

Thank you Forbes for naming us one of the Next Billion-Dollar Startups for 2025. We're hiring for a number of roles across the company. Head to our careers page to apply to join the team.

Our CEO, Misha Laskin joined Matt Turck on the MAD Podcast this week to talk about building enterprise superintelligence. Misha breaks down why today's agentic search still feels like "exploring a jungle with a flashlight" and how we're solving it with Asimov - the best-in-class code research agent that captures engineering team's collective knowledge. He also covers: - Why solving autonomous coding is the fastest on-ramp to superintelligence - How focusing on organizational superintelligence creates more practical value than chasing general superintelligence - His journey from theoretical physics to leading RLHF for Gemini at DeepMind Check out the full conversation below:

Introducing Asimov: the best-in-class code research agent, built for teams and organizations. Engineers spend 70% of their time understanding code, not writing it. That’s why we built Asimov. In blind tests, Asimov's answers to complex questions were preferred 60 - 80% of the time. Asimov works because… 1] Asimov builds a single source of truth for eng knowledge. Asimov looks at more than just code. It pulls knowledge from your codebase, your team’s messages, your project management tools, and more. 2] Asimov captures team-wide tribal knowledge with memories. Asimov learns from expert feedback and captures tribal knowledge stored in engineers' minds (e.g. "asimov, remember X works in Y way"). Once an update is made it benefits the entire team. 3] Asimov is designed to ingest a lot of context. Today agent designs fall into two categories: RAG or agentic search. Both struggle with large codebases. Asimov uses a new multi-agent design (a big reasoner with small retrievers) to ingest large codebases. Sign up for the waitlist: https://lnkd.in/eeuxFHyK.

Our GTM team is growing with two new roles both out of our NYC office: 1. Solutions Architect 2. Marketing Lead-Growth If you are aligned with our mission to build autonomous coding agents for enterprise customers consider joining us: https://lnkd.in/e3gGYyFK

Solving autonomous coding is the root-node problem that will lead to general superintelligence. Today, we are building tools that will transform engineers into architects who manage an AI workforce. Thank you Sequoia Capital for sharing our story. https://lnkd.in/guSfuDGC

"At the end of 2024, we started seeing these really powerful reasoning models come out, which are driven again by this technology called reinforcement learning… and that’s not going to stop anytime soon."- Reflection Co-Founder Misha Laskin on NYSE More detail on the Reflection approach at: https://lnkd.in/eWtKA9Y3 https://lnkd.in/eEmdA_Zz

For a long time, superintelligence seemed beyond the grasp of existing technology, much like finding a unified theory of physics has remained elusive. We are now within striking distance of AI systems that can assist with most computer-based work across disciplines. These systems will help scientists make discoveries and programmers build sophisticated software—all at least ten times faster than today. We have entered the final act of this research arc. Currently, the primary way of interfacing with AI on a computer is through chat or copilot experiences, similar to cruise control. The ultimate stage will be like an autonomous vehicle: a system that takes direction from a user and handles most of the work independently. We value speed, craftsmanship, intensity, and kindness. If our mission and approach inspire you, consider joining us.

"We think it's really important to not just do research in the wild, but to be really closely coupled to product and customers. A motto we have in the company is "The only eval that matters is a real world eval." We think that by being really close to customers we will be able to see through the mission (superintelligence) better."- Misha Laskin (CEO and Co-Founder)

Today we are announcing Reflection. Our team pioneered major advances in RL and LLMs, including AlphaGo and Gemini. At Reflection, we're building superintelligent autonomous systems. We believe that solving autonomous coding will enable superintelligence more broadly. Our company is defined by three things: 1) A team behind some of the most capable RL and LLM systems ever created - the two building blocks for superintelligence. 2) A focus on building the best autonomous coding systems in the world. Rather than doing many things, we do one thing really well. 3) Equal emphasis on research and product. Superintelligence cannot be built in a vacuum. We look for colleagues who have high internal drive, integrity, and a deep interest in the problems we’re pursuing. If that’s exciting to you, join us. We’re excited to partner with Sequoia Capital, Lightspeed, and CRV and want to thank Shirin Ghaffary for covering the story. Read more here: https://lnkd.in/e4Szy9bR