The Race to Unlock Superintelligence

“Ihave never been congratulated so much for winning one game,” said a smiling Lee Sedol, at the packed press conference amid enthusiastic applause and camera flashes. The great 9-dan Go player had just defeated DeepMind’s AlphaGo in game four with an unexpected wedge move that his adversary critically underestimated. Two days later, AlphaGo would prevail, winning four of the tournament’s five games. It wasn’t Lee’s move 78 that captured the world’s imagination, but AlphaGo’s move 37 two games earlier. Move 37 was so unconventional that Lee and much of the audience thought it was a mistake before realizing how “beautiful” it was. It had a 1-in-10,000 likelihood of being played by a human and was a landmark demonstration of AI’s ability to be superhumanly creative. 

During the tense final minutes of game five, the AlphaGo team crammed into the VIP suite at the Seoul Four Seasons Hotel that had become its command center. The team lead, David Silver, was flanked on one side by DeepMind founder Demis Hassabis and on the other by Ioannis Antonoglou, the research engineer who accelerated the neural networks that made AlphaGo tick. The moment when Lee finally resigned marked the end of a two year journey that pushed the science of reinforcement learning (RL) to new heights, and the team reveled in the accomplishment.

For Ioannis, it was a vindication of what had seemed a risky career move, joining the AI startup in London in 2012 as employee #25 and researcher #6. “It was the only place in the world where people were seriously considering building AGI,” he recalls now, and he was amazed at what RL had achieved in such a short time. He wasn’t the only one who was amazed. In Chicago, a freshly minted PhD in quantum physics named Misha Laskin read the AlphaGo paper and abruptly changed the course of his life. Here was something as beautiful as theoretical physics but with real world impact. Misha realized, “this was a real glimpse into what it will be like to live among superintelligence.”

AlphaGo and beyond

Ioannis’s first collaborator at DeepMind was Vlad Mnih, the researcher credited with building the first deep reinforcement learning agent, DQN (Deep Q Network). He asked Ioannis to write a clean implementation of DQN as a library that DeepMind could use to advance deep reinforcement learning research. “Ioannis’s attention to detail and intellectual rigour stood out right away,” says Vlad, “he would always ask why we were going with particular design choices, why we weren’t incorporating things like LSTMs sooner, etc.” 

A year later, when David Silver started working on AlphaGo, DQN had demonstrated how to use deep reinforcement learning in game playing. AlphaGo built on this foundation with many sophisticated innovations including using Monte Carlo Tree Search, a powerful search algorithm that allowed it to explore possible future moves by simulating game outcomes. AlphaGo also featured a dual network architecture with a policy network to predict the next move and a value network to evaluate board positions. 

This was a monster of a system by contemporary standards and Silver knew that maximizing throughput and minimizing latency was mission critical to train the system and then deploy it for realtime game play. He tapped Ioannis to write kernels on GPUs that could speed up the neural networks they were training. DeepMind was acquired by Google in 2014 and Ioannis quickly moved on to implementing the system on Google’s first TPUs (tensor processing units). These first chips were extremely temperamental and hard to work with but Ioannis figured out how to tame them. “If I couldn’t get it to work,” Ioannis says now, “then we wouldn’t be able to achieve the things that we achieved with AlphaGo.”

After the triumphant victory in Seoul, Ioannis continued to work on expanding the capabilities of these game-playing agents with a small core team. “One of the first papers that we wrote was AlphaGo Zero,” he remembers. “We showed that you can actually get the same performance—actually much better performance than AlphaGo—starting from completely random behavior all the way to superintelligent behavior.” This tabla rasa approach combined with self-play was important to demonstrate that RL can scale to solve new problems.

Then the team built AlphaZero, which added superhuman-level play of chess and shoji starting from random play, with no domain knowledge except the game rules. The zero in this case also meant that the algorithm can learn to play any zero sum board game. Three years later they introduced MuZero, which added Atari games to its arsenal while no longer supplying game rules—just raw pixels. Using reinforcement learning on problems without access to a perfect simulator of the world, as you have in Go or chess, was a longstanding challenge that MuZero solved. “You can actually have a system that both learns the world dynamics,” says Ioannis, “and uses this learned model of the world to actually plan into the future.”

Transformers in Toronto

Although Misha was captivated by DeepMind’s research, it took him a while to find his way there. He caught the entrepreneurial bug and went through Y Combinator with an AI startup, but the frontier of AI research kept calling—he dove into a postdoc in Pieter Abbeel’s lab at Berkeley to hone his AI research chops. Like a good reinforcement learning agent, he maximized learning from this environment of other brilliant researchers. Entering as a former physicist had a side benefit: “Because I didn’t know anything, I had this kind of clean slate. And I think that enabled me to try some pretty simple things that other people thought were too obvious to try.” He adds that, “It was surprising to me how doing the simple thing correctly was hard. And oftentimes when people try something simple and it didn’t work, it’s because they didn’t try it hard enough.”  

Over the course of the next two-and-a-half years, Misha followed the interests he developed through self study, improving RL methods in the DeepMind Control Suite and Atari Games environments along with the OpenAI Gym. Specifically, he worked on data augmentation strategies and extracting high-level features from raw pixels (CURL). The most cited paper he co-authored at Berkeley was Decision Transformer, which reframed RL as a sequence modeling problem, leveraging the simplicity and scalability of the Transformer architecture.

In 2021, Misha interviewed with Vlad, who was starting a new general agents team for DeepMind in Toronto. “I was immediately struck by Misha’s enthusiasm and belief in the power of reinforcement learning and unsupervised approaches, so it was really exciting when he decided to join the team,” recalls Vlad. “Misha was very productive from the start. He got up to speed with Google infrastructure and started producing research results in a matter of weeks, the same amount of time it took him to find all the good coffee shops in Toronto.” 

Vlad also appreciated the clean slate Misha brought with him from his academic research, “he liked to start on new problems with very simple approaches. It’s easy to overcomplicate things in research, but Misha seemed to have a knack for finding the simplest approach that will work on a problem.”

Vlad and Misha would go on to collaborate on a paper titled, In-context reinforcement learning with algorithm distillation after working together on a DeepMind hackathon project. “Misha and I decided to train a Transformer to mimic a reinforcement learning algorithm to see if it would exhibit self-improvement on tasks it wasn’t trained on,” says Vlad. “We had a lot of fun pair programming and getting the first prototype to work. This was an early example of in-context self-improvement, which has become a very active area of research in the LLM literature. We got really excited when we discovered that the Transformer learned a much more efficient RL algorithm than the one it was trained to mimic.”

Both Ioannis and Misha were pushing hard on how to make deep RL agents more capable and general through increased scale and algorithmic improvements. But on November 30, 2022, AI history took an abrupt turn. “Everything changed when ChatGPT came out,” Ioannis remembers. “I think that was a big moment for everyone in the industry to see that these systems are so powerful and they can be a starting point for something that’s even more intelligent.”

The Race to Gemini

ChatGPT’s success exposed that the value of deep reinforcement learning had been hiding in plain sight: Reinforcement learning with human feedback (RLHF) was a crucial part of its training, aligning the model’s outputs with human preferences and values. Like the Transformer itself, OpenAI didn’t invent deep RL, but was the first to deploy it at this scale.

ChatGPT’s ramp to 100 million users in two months sent shockwaves through the tech industry—but particularly at Google. Not only had OpenAI capitalized on the Transformer architecture, they had added an RL step in post-training to make a runaway consumer product. At the beginning of 2023, Google’s CEO Sundar Pichai called a “code red,” and enlisted the help of founders Larry Page and Sergey Brin to help the company catch up to OpenAI. A few months later, the Google Brain team was merged with DeepMind to become Google DeepMind under the leadership of Demis Hassabis.

The code red brought the AlphaGo lineage of research to an end for Ioannis, who was tapped to lead the RLHF effort for Google’s new large language model (LLM) Gemini. Misha was looking to join the Gemini team as well and Ioannis asked him to lead on reward models. Over the next two years they worked together to launch Gemini 1 and 1.5. Misha was initially intimidated by Ioannis, but soon came to see his colleague in a new light. “What was interesting about the RLHF team that he led was that he was extremely people oriented and worked very hard on ensuring that the people were self-motivated. He would deeply understand what drove them and would figure out the direction to steer them to ensure that what they were doing was impactful for Gemini. I think that’s a superpower that he has as a leader.”

By December of 2023, Google had announced Gemini and when v1.0 was released in February of 2024 it was comparable or better on most benchmarks to OpenAI’s GPT-4. This was a huge accomplishment but the team didn’t let up. A couple of months later it announced Gemini 1.5, three months before OpenAI released its next major upgrade, GPT-4o. Gemini 1.5 improved compute efficiency through a Mixture of Experts architecture, added multimodal capabilities and a million token context window.

For Misha, the sudden prominence of LLMs was a revelation. “We were trying to solve the general agents problem, but we were looking at it completely the wrong way,” he says. “Because we thought that generality wasn’t solved.” For all their limitations and confabulations, LLMs are the most general systems ever invented. “They’re not very reliable,” Misha explains. “They can’t do any work for you autonomously. You can’t ask them to do something in the way that you would ask a coworker, but they’ll answer almost any question that you have.”

Combining that broad intelligence with the ability to take specific actions reliably was still an unsolved problem. “In AlphaGo, the reward is clear. You win the game or you lose. And in these systems where it’s clear, reinforcement learning works like magic,” says Misha. In most real life situations, however, the reward is quite fuzzy. How do you know if a complex task is complete? How do you know if a human would like a model’s response? This is where his work at DeepMind on open ended reward functions for general agents come in. “It just so happened that that building block that I was focused on outside of language models was one of the core building blocks for training them,” Misha explains.

Misha and Ioannis had achieved their goal of becoming, in Misha’s words, “the best craftsmen of training language models.” They understood these systems inside and out, and yet they didn’t feel the trajectory was moving fast enough towards their burning research question: how to build superintelligent systems capable of acting autonomously?

LLMs have succeeded at achieving a very wide range of capabilities, but their competence is shallow. They have, as Misha puts it, solved the breadth problem but not the depth problem. "It became very clear that these two streams of research are the two building blocks for building a superintelligence," he says. "You get generality from the language models and you get capability from reinforcement learning." And indeed, Ioannis adds, "all the necessary components to build a superintelligent agent that can have an impact in the real world are already there."

Their bet was that the returns to pre-training scale were compressing while the RL curve was just beginning to climb, letting an agent get better and better at specific tasks. "Things that it gets right, it should just do more of. And the things that it gets wrong, it should do less of. Essentially, this is the model learning from its mistakes," explains Ioannis. "And this really simple mechanism, if scaled correctly, gives rise to superintelligent behavior." Misha puts a finer point on it: "reinforcement learning is the most scalable form of AI. There's basically no ceiling that we know of.”

When they left DeepMind to form Reflection in March of 2024, the strategy was at odds with where most of the industry was heading. Today every major lab is working on reasoning models that augment LLMs with RL, but this wasn't an obvious direction at the time. More unusual still, they decided to do their own training. Most of the labs forming around them were assembling products on top of a closed model's API, with no ability to influence how the underlying model behaved. Reflection set out to build its own large-scale LLM and reinforcement learning platform instead, and run its RL work on open base weights it controlled.

The turn

The “DeepSeek R1 moment” that rattled markets and AI circles in January 2025 wasn’t really a moment, but the start of a progression. DeepSeek proved a small team could reach the frontier on a modest budget. Alibaba's Qwen family made open weights ubiquitous. Then Moonshot's Kimi models took the lead on raw capability. By June 2026, Z.ai’s GLM-5.2 beat GPT-5.5 on multiple long-horizon coding benchmarks at roughly one-sixth the cost. Meanwhile, Meta redirected toward a more closed approach, in line with the hyperscalers. The best open models in the world were now Chinese. The West had largely stopped competing for the category.

Reflection found itself in a cumbersome  position. A company whose customers would turn out to be governments and regulated enterprises could not credibly ship a Chinese base model. "We found ourselves in a position where this core piece of technology – that we believed needed to exist, and that we were going to rely on – no longer existed," Misha says. Filling that gap meant going a layer deeper than the company had ever gone: pre-training a frontier model from scratch, on tens of trillions of tokens, rather than post-training someone else's. 

In October of 2025 Reflection raised $2 billion to do exactly that, arguing that the U.S. had to foster an open model ecosystem akin to the mounting one in China––and offer organizations, globally, a path to their own, sovereign AI. 

Rent or own

The distinction Misha draws isn't really technical. It's about control. "In the housing market there's a rental market and there's a buying market," he explains. Closed models are the rental market: a landlord meters your tokens and can throttle you, threshold you, or raise the rent, because the asset was never really yours. Open weights are the ownership market. You run the model on your own infrastructure, and you can, as he puts it, build your garden however you want.

That didn't matter much when AI was a chat-dominant market. It matters enormously now that AI agents do real work. "If you're a large bank, you don't want to rent something that is going to be the single most important IP asset you have," says Misha. The loudest version of the argument comes from Palantir CEO Alex Karp, who has spent much of 2026 telling enterprises that what technical buyers want is control of their compute, models, data and alpha. "They want to know they own the means of production."

Reflection's second case for ownership is security—self-hosting is the only arrangement where proprietary data never leaves the environment, and open weights are the only ones an organization can test without asking permission. Owning the weights is what lets an organization post-train against its own domain and end up with a system competitors can't buy off the shelf. 

The third is cost. Deploying open-weight models creates predictable economics at scale, eliminating the per-token API margin (and risk) of closed models. This also gives enterprises pricing visibility that they can use to manage inference costs, optimizing them as agentic workloads grow.

And then there are the sovereignty and national security concerns: DeepSeek's V4 release reportedly slipped by months while its inference was re-engineered for Huawei's Ascend chips. SemiAnalysis found the model was co-designed in part for Ascend, and it ran on Huawei silicon from day zero. Adopt the model and you have quietly adopted a chip roadmap, a software stack and a set of assumptions about who can revoke what. Misha calls open models Trojan horses for the infrastructure that rides along with them, up to and including the risk that a state actor could train deceptive behavior into the weights themselves. As countries wake up to the need for sovereign cyber AI defense, open weights are an increasingly critical ingredient: you can run the model in your own environment, freeze a version, and test it as adversarially as you like—none of which is available to a renter.

This has become Reflection’s new mission: building open intelligence, accessible to all. The company has committed to releasing its model weights, publishing the underlying research and open-sourcing the deployment software, while building the AI factories and custom systems that put its models into production. It joined NVIDIA's Nemotron Coalition, signed a memorandum with Shinsegae to build a 250-megawatt AI factory serving Korean enterprises and agencies, and partnered with the Department of Energy as a model provider for the Genesis Mission across the 17 national labs. "You want to be in a world where there are multiple competing parties," Misha says, "where countries have choice, where businesses have choice."

AI at its current stage is like the early steam engines before the discovery of thermodynamics: Lack of theory didn't stop inventors from producing new engines. "I think deeply understanding why a model works from a theoretical sense will certainly be a very useful thing," says Misha. "Whenever a thing has been deeply understood theoretically, at least in physics, it spurred a new era of empirical innovations because scientists knew where to search. But you don't need to wait for that to build reliable systems today."

Richard Feynman was an early hero of Misha's, inspiring him to study physics. In his lecture on the conservation of energy, Feynman said, "It is important to realize that in physics today, we have no knowledge of what energy is." The same is true of AI today, and the study of intelligence. During a Nobel Prize interview in Stockholm, DeepMind founder Demis Hassabis summed up where we are in the quest for superintelligence: "I think the science of AI is about trying to explore and understand what intelligence is, and the best expression of understanding something is actually trying to build it."