Intelligence with Everyone: RL @ MiniMax, with Olive Song, from AIE NYC & Inference by Turing Post

Olive Song from MiniMax shares how her team trains the M series frontier open-weight models using reinforcement learning, tight product feedback loops, and systematic environment perturbations. This crossover episode weaves together her AI Engineer Conference talk and an in-depth interview from the Inference podcast. Listeners will learn about interleaved thinking for long-horizon agentic tasks, fighting reward hacking, and why they moved RL training to FP32 precision. Olive also offers a candid look at debugging real-world LLM failures and how MiniMax uses AI agents to track the fast-moving AI landscape. Nathan uses Granola to uncover blind spots in conversations and AI research. Try it at ⁠granola.ai/tcr⁠ with code TCR — and if you’re already using it, test his blind spot recipe here: ⁠https://bit.ly/granolablindspot⁠ LINKS: Conference Talk (AI Engineer, Dec 2025) – https://www.youtube.com/watch?v=lY1iFbDPRlwInterview (Turing Post, Jan 2026) – https://www.youtube.com/watch?v=GkUMqWeHn40 Sponsors: Claude: Claude is the AI collaborator that understands your entire workflow, from drafting and research to coding and complex problem-solving. Start tackling bigger problems with Claude and unlock Claude Pro’s full capabilities at https://claude.ai/tcr Tasklet: Tasklet is an AI agent that automates your work 24/7; just describe what you want in plain English and it gets the job done. Try it for free and use code COGREV for 50% off your first month at https://tasklet.ai CHAPTERS: (00:00) About the Episode (04:15) Minimax M2 presentation (Part 1) (17:59) Sponsors: Claude | Tasklet (21:22) Minimax M2 presentation (Part 2) (21:26) Research life and culture (26:27) Alignment, safety and feedback (32:01) Long-horizon coding agents (35:57) Open models and evaluation (43:29) M2.2 and researcher goals (48:16) Continual learning and AGI (52:58) Closing musical summary (55:49) Outro PRODUCED BY: https://aipodcast.ing SOCIAL LINKS: Website: https://www.cognitiverevolution.ai Twitter (Podcast): https://x.com/cogrev_podcast Twitter (Nathan): https://x.com/labenz LinkedIn: https://linkedin.com/in/nathanlabenz/ Youtube: https://youtube.com/@CognitiveRevolutionPodcast Apple: https://podcasts.apple.com/de/podcast/the-cognitive-revolution-ai-builders-researchers-and/id1669813431 Spotify: https://open.spotify.com/show/6yHyok3M3BjqzR0VB5MSyk

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Nathan Labenz?

  Hello and welcome back to the Cognitive Revolution. The presenting sponsor of today's episode is Granola, the AI notepad that helps you get the doing done. Whether it's identifying to-do items after a call, turning a brainstorming session into a product spec, or looking back at multiple calls to identify cultural trends at your company, Granola takes your raw meeting notes and makes them awesome. Right now, Granola is featuring AI recipes from AI thought leaders, including several past guests of this show. My own contribution is a blind-spot-finder recipe that looks back at recent conversations and attempts to identify things that I am totally missing. This was immediately useful in the context of contingency planning for my son's cancer

• 0:43Nathan Labenz?

  treatment, and the more data Granola collects as I continue to use it, the more valuable it becomes for suggesting AI topic areas that I really ought to explore. See the link in our show notes to try my blind-spot-finder recipe and experience for yourself how Granola puts your meetings to work. Now, today, I'm excited to share a special combined crossover episode featuring Olive Song, a senior researcher specializing in reinforcement learning and model evaluation at the Chinese AI company Minimax, creators of the M series of models, the most recent of which, M2.5, currently tops the open router usage leaderboard. To give you the most complete picture possible, we're combining two sources.

• 1:28Nathan Labenz?

  First, a presentation Olive recently gave at the AI Engineer Conference in New York, where she had previously lived for six years, and second, an interview with Kasania Seh from her podcast, Inference by Turing Post. Together, they provide an excellent overview of Minimax's goals as a company, the capabilities they're prioritizing in their models, the techniques they're using to get there, and the day-to-day ups and downs of training frontier LOMs. Highlights include how Minimax's strategy of building both models and user-facing applications in-house creates tight feedback loops that enable their cross-functional research and engineering teams to identify and address model weaknesses as quickly as possible. An overview of how interleaved thinking, which allows the model to take an action,

• 2:13Nathan Labenz?

  get feedback from the environment, and pause to think again before continuing, improves performance on long-horizon agentic tasks.

• 2:21Nathan Labenz?

  A description of the perturbation pipeline they use to systematically vary the model's training environment in order to encourage robust generalization. Olive's perspective on the constant battle she and teammates are fighting against reward hacking. A window into the tedious debugging that is sometimes required to diagnose training issues, and how they realized that they needed to run reinforcement learning at FP32 precision. And finally, how the team at Minimax is using AI agents to keep up with the daily flood of AI news. While Olive recognizes that Minimax's models, like all open source models in the world today, can't quite match the performance of top American models, I think there is still a lot of value in the details she shares about their approach to reinforcement learning

• 3:05Nathan Labenz?

  and how they structure their team and work. And in any case, I always appreciate the opportunity to hear directly from Chinese AI researchers, who, just like their American counterparts, are figuring things out step by step as they go, even as major questions about issues such as the governance of increasingly powerful open source models remain fundamentally unanswered. With that, I want to thank Swix, the creator of the AI Engineer Event Series, which I absolutely recommend attending if you can, and Ksenia, the creator of TuringPost, which has what I find to be some of the very best topic selection of any AI newsletter, for allowing me to create and post this combined episode. And I hope you enjoy this window into the development of some of the best open weight

• 3:49Nathan Labenz?

  models in the world with Olive Song of Minimax.

• 4:16Speaker 01

  Hi, hi, everyone. I'm Olive. It's my great honor here today to present on our new model, Minimax M2. I actually lived in New York City for six years, so it feels great to come back, but with a different role. I currently study reinforcement learning and model evaluation at Minimax. Let me just get a quick sense of the room. Who here has heard or have tried of Minimax before? Oh, a couple of there, yeah. Not everybody, but I guess, yeah. But here's the value, right, of me standing here today. So we are a global company that works on both foundation models and applications.

• 4:57Speaker 01

  We develop multimodality models, including text, vision language models, our video generation model, HaLo, and speech generation, music generation stuff. And we also have many applications, including agents and stuff in-house. So that's the specific thing that's different from the other labs or other companies. So we both develop foundation models and applications. So we have research and developers sitting side by side, working on things. So our difference would be that we have firsthand experience from our in-house developers into developing models

• 5:41Speaker 01

  that developers would really need in the community. And here I want to introduce our MiniMax M2, which is an open weight model, very small, with only 10 billion active parameters, that was designed specifically for coding workplace agentic tasks. It's very cost efficient.

• 6:05Speaker 01

  Let me just go over the benchmark performance, because people care about it. So we are ranked very top in both intelligence benchmarks and also agentic benchmarks. I think we're on the top of the open source models. But then, numbers don't tell everything, because sometimes you get those super high-numbered models, you plug into them into your environment, and they suck, right? So we really care about the dynamics in the community. And in our first week, we had the most downloads. And also, we climbed up to top three token usage on OpenRouter.

• 6:46Speaker 01

  So we're very glad that people in the community are really loving our model into their development cycle. So today, what I want to share is how we actually shape these main model characteristics that made M2 so good in your coding experience. And I'm gonna present to you the training behind it that supports each one of them. From coding experience to long horizon state tracking tasks, to robust generalization to different scaffolds, to multi-agent scalability. So first, let's talk about code experience, which we supported with scaled environments

• 7:30Speaker 01

  and scaled experts. So developers need a model that can actually work in the language they use and across the workflow that they deal with every day. So which means that we need to utilize the real data from the internet and then scale the number of environments so that the model, when during training, for example, during reinforcement learning, it can actually react to the environment, it can actually target verifiable coding goals and to learn from it. So that's why we scaled both the number of environments and also our infrastructure so that we can perform those training very efficiently.

• 8:15Speaker 01

  So with data construction and reinforcement learning, we were able to train the model so that it's very strong, it's full-stack, multilingual. And what I want to mention here is that besides scaling environment that everybody talks about, we actually scaled something called expert developers as reward models. So as I mentioned before, we have a ton of super expert developers in-house that could give us feedback to our model's performance. So they participated closely into the model development and training cycle, including

• 8:54Speaker 01

  problem definition, for example, bug fixing, for example, repo refactoring and stuff like that. And also, they identify the model behaviors that developers enjoy. And they identify what's reliable and what developers would trust. And they give precise reward and evaluation to the model's behaviors to the final deliverables. So that it is a model that developers really want to work with. And that can add efficiency to the developers.

• 9:29Speaker 01

  So with that, we were able to lead in many languages in real use.

• 9:35Speaker 01

  And the second characteristic that Minimax M2 has is it performs good in those long horizon tasks. Those long tasks that require interacting with complex environments, that requiring using multiple tools with reasoning. And we supported that with the interlead thinking pattern and reinforcement learning.

• 10:01Speaker 01

  So what is interlead thinking? So with a normal reasoning model that can use tools, it normally works like this. You have the tools information given to it, you have the system prompts, you have user prompts. And then the model would think, and then it calls tools. It can be a couple of tools at the same time. And then they get the tool response from the environment. And then it performs a final thinking and deliver a final content. But here's the truth, right? In real world, the environments are often noisy and dynamic. You can't really perform this one test just by once.

• 10:42Speaker 01

  You can get tool errors, for example, you can get unexpected results from the environment and stuff like that. So what we did is that we imagined how humans interact with the world. We look at something, we get feedbacks. And then we think about it, we think if the feedback is good or not. And then we make other actions, make other decisions. And that's why we did the same thing with our M2 model. So if we look at this chart over a diagram on the right. So instead of just stopping after one round of tool calling, it actually thinks again.

• 11:23Speaker 01

  And reacts to the, reacts to the environments to see if the information is enough for it to get what it wants. So basically, we call the interleaved thinking, or people call it interleaved thinking because it interleaves thinking with tool calling. A couple of time, it can be, you know, tens to a hundred turns of tool calling within just one user interaction turn.

• 11:52Speaker 01

  So it helps adaptation to environment noise. For example, just like what I mentioned, the environment is, is not stable all the time. And then something is subbotonal, and then it can choose to use other tools or do other decisions. It can focus on long horizon has, can automate your workflow using, for example, Gmails, Notions, Terminal, all at the same time. You just need to maybe make one model call without minimal, with minimal human intervention, it can do it all by itself. And here's a cool illustration on the right, because it's in New York City, I feel the vibe of, you know, trading and marketing.

• 12:36Speaker 01

  So you can see that there was some, there was some perturbations in the stock market, I think last week. And then our model was able to keep it stable. So just like I said, there's like environment noise, there's new information, there's like, yeah, news, it looks like there's like other trading policies and stuff like that, but our model was able to, to perform pretty stably in these kind of environments.

• 13:07Speaker 01

  And the third characteristic is our robust generalization to many agent's schedules, which was supported by our perturbations in the data pipeline.

• 13:19Speaker 01

  So we want our agent to generalize. But what is agent generalization? At first, we thought it was just too scaling. We trained the model with enough tools, various tools, kind of new tools. We invent tools. And then it would just perform good on unseen tools. Well, that was kind of the truth. It worked at first. But then we soon realized that if we perturb the environment a little bit, for example, we change another agent's scaffold, then it doesn't generalize. So what is agent generalization? Well, we conclude that it's adaptation to perturbations across the model's entire operational space.

• 14:01Speaker 01

  If we think back, what's the model's operational space that we talked about? It can be tool information. It can be system prompts. It can be user prompts. They can all be different. They can be the chat template. They can be the environment. They can be the tool response. So what we did is that we designed and maintained perturbation pipelines of our data so that our model can actually generalize to a lot of agent's scaffolds.

• 14:35Speaker 01

  And the fourth characteristic that I want to mention is the multi-agent scalability, which is very possible with M2 because it's very small and cost-effective.

• 14:49Speaker 01

  I have a couple of videos here. This is M2 powered by our own Minimax agent app. We actually have the QR code downside, so if you want it, you can just scan and try it. So it's like an agent app we developed. And here, we can see different copies of M2, right? It can do research. It can write the research results and analyze it and put it in a report. It can put it in some kind of front-end illustration, and they can work in parallel. So because it is so small and so cost-effective, it can really support those long-run agentic tasks and tasks that maybe require some kind of parallelism.

• 15:37Speaker 01

  So what's next, right? For Minimax M2, from what I've introduced, we gathered environments, algorithms, data, expert values, model architecture inference, evaluation, all these stuff to build a model that was, you know, fast, that was intelligent, that could use tools, that generalizes. What's next?

• 16:03Speaker 01

  For M2.1 and M3, or in the future, we think of better coding, maybe memory or context management, proactive, AI for workplace, vertical experts. And because we have those great audio generation and video generation models, maybe we can integrate them. But all our mission is that we're committed to bring all these resources, whatever is on the screen and maybe more, yeah, and values, and put them all together to develop models for the community to use. So we really need feedback from the community, if possible, because we want to build this together.

• 16:47Speaker 01

  And, you know, this is kind of a race that everyone needs to participate. And then we are committed to share it with the community.

• 17:00Speaker 01

  And that's all the insights for today. We really hope, again, we really hope you to try the model because it's pretty good. And then you can contact us up there. You can try the models by scanning the URL code. Yeah, basically, that's it. Thank you all for listening.

• 17:35Speaker 01

  During reinforcement learning, the model tries its best to hack a lot of things. The current open models can achieve that level of understanding. It is a solvable problem and we are working on it. Engineering is very, very, very important. I didn't know that during school.

• 17:54Speaker 02

  Hey, we'll continue our interview in a moment after a word from our sponsors.

• 18:02Speaker 03

  Hello, everyone. Today, I have the pleasure of talking to Olive Song, a senior researcher at Minimax. And recently, they've been launching very interesting open weight models specialized in different areas. And Olive is currently working on Minimax on the new version, Minimax 2.2. Thank you for taking the time at 9:00 PM on Sunday night. Does everyone works like this at the company? I'm really impressed.

• 18:29Speaker 01

  I think different people work on different schedules. We do have people who work even overnight, but they sleep at daytime. So like we have a very flexible schedule, you know, goes with your experiments. For example, if the experiment runs for all day, the person can take a break. And then if, you know, there are a lot of analysis to do, maybe because we are very curious about the results and we're very passionate, right? We can't really wait a very long time. So yeah, everyone has their own schedule.

• 18:56Speaker 03

  That's telling about the success of the models. I think that's influenced that. You specialize in reinforcement learning and model evaluation, as far as I understand, which are two of the least forgiving parts of model development. And you also have more constraints than big American AI labs. What does a good day look like for you and what does a bad one look like?

• 19:17Speaker 01

  I can share something about our recent weeks. So there's not a whole good day or a whole bad day. We were joking that during one day, we have good results in the morning and then sometimes it becomes bad results at night. You know, sometimes we call it, we have like ICU in the morning and then KTV at night. So typically, a good time would be usually receiving some good results. Or like, even if running into new problems is a good time. So if we, for example, during reinforcement learning, right? We can see the model doing a lot of different stuff to achieve the results. And sometimes we just discover like new model behaviors. And that's really exciting, even though it might not be safe. Or like, it might not be expected. It's kind of exciting. So I, I call it a good time.

• 20:01Speaker 01

  A bad time would be, see, really, isn't a bad time. Except for, you know, looking at the, you know, just finding out the bad results. The moment itself is bad, but then trying to figure out the problem and breaking it down is pretty good time.

• 20:17Speaker 03

  What were the recent model behavior that you didn't expect?

• 20:20Speaker 01

  So the model, during reinforcement learning, the model tries its best to hack a lot of things, right? For example, it uses patches a lot. And sometimes it might not be very safe behaviors as our expert developers say. Cause sometimes the expert developers have their own expectations on how the model works. But then it doesn't go that way, you know, if we don't constrain it. So we do a lot of alignment to solve that issue.

• 20:45Speaker 03

  You just launched Minimax Her, and that goes like all over the Twitter. How do you come up with those ideas? Because, you know, role playing is sort of, like, is it an alignment question? Is it not? Like, how do you, how do you do that?

• 20:59Speaker 01

  Frankly speaking, I'm not the expert person on that part. We have a whole team on um role playing, her stuff. I'm not an expert, but we do have a lot of discussions. Um, we do believe that role playing or like, you know, accompanying human or like human interactions is very important in the life with AI or like how it would change our social life in the future. And it absolutely represents some ability that's very superior. Cause that's like human-like, you know, it has emotions, where you understand your emotions. It's not just, you know, working out some exams. That's absolutely another side of the AI capability.

• 21:37Speaker 03

  What is called AI with everyone, right? In Minimax?

• 21:39Speaker 01

  Yeah, it's intelligence with everyone.

• 21:42Speaker 03

  Intelligence, intelligence with everyone. Uh, what does it mean for you?

• 21:45Speaker 01

  For me personally, I feel like it's more like how it changes my life and it enables me to do more work. And then how it can connect me better to different people. Cause for example, before I wouldn't be able to understand a lot of very professional, for example, right? Very professional coding problems or like optimization problems. And now I am able to do that with A I and so that I can communicate with more people and exchange more ideas. That's one side. And on the other side, it generally helps my daily life, right? So it helps with my work, my daily routine, my self-care, changes life for me and I hope that it changes life for everybody. Obviously in the good way.

• 22:24Speaker 03

  Can you tell me a little bit how day-to-day work organized in your lab? I remember from your talk at A I engineer that it's a very interconnected between developers and researchers. I would love to hear more about that.

• 22:35Speaker 01

  Absolutely. We sit around every day, so we share our experiment results. For example, as I just said, during experiments, for example, reinforcement learning experiments, we see some scores going up high. We look at the model's behaviors and we look at the model behaviors with the developers in that area as well. We sit together and then they will spot the issue right away. And then we we are able to come up with new ideas to fix it or like building more data on it.

• 23:04Speaker 03

  If we can go into details, like your current work on the current model, uh, the current version, what are the biggest problems you're trying to solve comparing to the previous version?

• 23:14Speaker 01

  One important thing we focus on right now and also in the future is human alignment. Cause we are focusing on coding models for two point one, two point two and the M two series, right? And then what we realize is that for it to become very productive in our daily work or like for it be productive and safe at the same time, we have to do a lot of alignment on it. So the model can't just grow on its own and then do some dangerous behaviors just to achieve the final goal. So for us, the important thing would be how we define human alignment, how we define expert expectation, how we. We actually train the model to be more aligned with our expectations.

• 23:56Speaker 03

  So I want to go like on some real details here. And you're an expert here, so correct me if I'm wrong, but I saw that there was a recent interest in details like keeping the LM hat in F P thirty two during reinforcement learning training. Why do small decisions like this end up mattering more than just like a clever new algorithm?

• 24:17Speaker 01

  It all ends up being closer to the theoretical algorithm. So we have the theoretical reinforcement learning algorithm. And then, but when we implement it, it could be a little bit off. That creates a little bit gap to the theoretical extreme of this algorithm. So that's how we think and approach to this problem is that we try to, you know, scale to the theoretical extreme. And for example, the precision part is one thing that we found that would, you know, kind of like prevent us from being close to that extreme and that that's how we solve it. That was a very funny story, actually, for when we discovered that. I kind of talked about it when we published Minimax M one.

• 24:59Speaker 01

  So during our experiments, we found that the accuracy didn't go up. We we looked at like layer by layer. We looked at the log profs layer by layer and found it out. Theoretically speaking, it has to work, right? And then there has to be some gap between the theoretical and how we approached it. So we thought about the gap and we'll analyze it layer by layer and then eventually found.

• 25:22Speaker 03

  Is there anything like this happening now?

• 25:24Speaker 01

  Definitely, yeah, every single day, every single day and in every different groups. I can't actually disclose something that we haven't have the concrete conclusion. Cause we want our conclusions or like anything public to be very concrete and we understand it very deeply. So if we have, you know, breakthroughs, we'll definitely publish it later. But I'll have to say, we do encounter these problems every day and we think, I think it's called the first principle, right? So we think from the very fundamental part of the problem and then approach it.

• 25:54Speaker 03

  The models that you launch are open weights. And from your perspective and from the alignment perspective, what do builders actually gain from open weights? And what responsibility do they have to take on that like you don't have to take responsibility for this?

• 26:09Speaker 01

  Again, I'm actually not an expert in building developments or like building things with models. I feel like because it's open weight, people can have like free use with it. For example, they can deploy it on by themselves or they can even fine tune it with the ways. And then have, you know, all the data on their their properties is very safe.

• 26:30Speaker 03

  Well, if we talk about alignment, how do you look at that from that perspective? When the model is out there in the wild, before you launch the model, before you publish it, what does tell you that it's safe to publish?

• 26:42Speaker 01

  We have some internal benchmarks in terms of safety and it has different dimensions. Something that's sensitive safety or something that's like alignment safety. We have that as our evaluation and then before launching about one or two weeks before launching, we do scaled up evaluations and we do scaled up alignments on the model. And that's how we assess if the model is safe. But then if it's already open weight in the wild, people actually can do something on it. I guess that's what you are approaching at, right? People can do more things on the model that we can't control. I don't know how we handle that, frankly speaking. There are laws on that, right? There are regulations where like people do agree on some moral standards on that.

• 27:27Speaker 03

  Do you see, do you follow any reinforcement learning failure modes? That haven't showed up on benchmarks, but then become obvious and really genticues? How do you collect feedback for the next versions for improving the reinforcement learning process?

• 27:42Speaker 01

  We collect feedbacks on the model itself first. So when we publish a model outside, many developers use it or like many people use it. We collect it systematically, we analyze these problems. Some of them are fundamental, some of them are, you know, just something that we missed and we can fix it real quick. So there are two parts. First, we do the internal evaluation with the developers and they point out problems and that's how we can fix this part. But they are not enough and more feedbacks will come to us after we officially publish the models and then we collect it. Because uh the way we organize our group is that different people work on different capabilities of a general model. If we collect some things that we think we should improve in the future, different people take their parts, right?

• 28:28Speaker 01

  So they're like, okay, I I think I can solve this issue and I'll solve it next in the next generation. And that's how we, you know, collect feedbacks and then improve the model.

• 28:38Speaker 03

  How did you initially decide to not build one general use model, everything for all, and go more into specialization, like coding?

• 28:46Speaker 01

  I think we are approaching generalized models. It's just that we are putting more emphasis on coding. For example, our model also, you can take it into any general agent scaffold, including our own agent product. And that's for general purpose. We do work on like, you know, researching, report writing, P P T, the stuff like that, that's more general. Personally speaking, I feel like with coding, you can, you know, kind of structure the whole world. Or like you can model a lot of stuff with.

• 29:16Speaker 03

  Yeah, engineer it.

• 29:17Speaker 01

  Yeah, with engineering. So behind it, it's it's scaled up humanity for me. So it itself has a lot of intelligence with it and a lot of work to do. So that's how we view this issue. But we do work on generalized stuff and even more generalized stuff in later versions. For example, we, uh, our model can do some general workplace scenarios in the future. And that's not just coding.

• 29:44Speaker 03

  If we talk about coding and agentic use, it requires long horizon. How do you solve long horizon for agentic use?

• 29:51Speaker 01

  I think defining your goals good and define the model behavior is good. And also, we require great infra, extraordinary infrastructure. For example, for reinforcement learning, right? The very important issue besides algorithm, besides things that people have been working on for a very long time, what's special for agentic stuff is how we define agents. How you define how an agent model would work. First, you need to define the task, you need to define the model's goal, especially in a long horizon task, you need need goals that are actually hard and diverse. And then the second part is that you need environments. You need great engineering environments, scaled up environments, different diverse environments,

• 30:34Speaker 01

  not just coding or like more, for example, workplace, different kind of tools. That's great engineering. And then you need great infrastructure. You need outstanding R L infrastructure to let the model really roll out in a very long horizon. And then with, you know, very efficient, for example, G P U use, very efficient a training rollout in training and stuff. I feel like that's what's different in agentic reinforcement learning as compared to before.

• 31:00Speaker 03

  Are you affected by G P U's constraint? How do you solve the compute problem?

• 31:05Speaker 01

  We do have a team that works on how we utilize the compute the most. You know, that's actually one of the RL scaling issue is to utilize the compute very efficiently. So their purpose would be to minimize the compute use and then while train more, right? Personally speaking for me, I don't really have a GPU constraint. It's that we have a great team who works on it to utilize the compute most while, you know, stabilize the training most.

• 31:32Speaker 03

  But do you have problems that you need to solve with your expertise? How to use more efficiently, or it's just that team?

• 31:38Speaker 01

  We are actually the same team, because we're actually the reinforcement learning team. We view this issue on different perspectives. It can be, you know, implementation, right? It can be implementation. You can view it as a data perspective. You can view it at different perspectives. But our goal is the same.

• 31:54Speaker 03

  We're always, you know, we're always looking forward to new solution that comes from Chinese labs because it's always mind blowing.

• 32:01Speaker 01

  We are actually working on some new agentic reinforcement learning stuff, but it won't really come out with two point two. With the next generation model, we are actually, we are still working on it. I'm not sure what I can share or not, so. I can share it later when I have concrete conclusions, as I said before. I can't really say something that we, you know, don't document yet.

• 32:23Speaker 03

  Will it be available when the model is out?

• 32:26Speaker 01

  That depends on our time. I'm not very confident yet, but we are dedicatedly working on it.

• 32:33Speaker 03

  Yeah, a lot of constraints talking to researchers.

• 32:38Speaker 03

  Well, if we talk about openness, then this whole conversation that I'm having with people right now in this quarter is about open source. I wonder if you can talk about the company's strategy, why the company decided to go and publish open weights of the models? What's the benefits? What's the cons to that?

• 32:56Speaker 01

  So for our team, like for the researchers team, we all, we always wanted to go open source. Because, you know, open source community is fantastic. I learned that from day one when I joined the team, is open source community is fantastic. So as researchers, we did want to, you know, join the open source. But then on the other hand, speaking of the cons, we are a company. That people can care about if this can be, you know, sell money or like if this is a business. So the cons would be if the the ways are open source less, people will use APIs. But then as a researcher, that really isn't my focus that much. So I'm not very confident about the company's strategy. For the tech part, we just believe that we can build better models with the open source community.

• 33:41Speaker 03

  How much do you use open source tools yourself from different other companies?

• 33:44Speaker 01

  A lot. For example, instance, we use, I'm not sure if I'm allowed to say specific open source branches, but then we collaborate with both V I M and S G line. They are they are open source code repositories.

• 33:58Speaker 03

  How do you look at open source stack? Because when we talk open source, sometimes it's perceived as one thing, but actually it's it's multi-layered. How do you look at it?

• 34:07Speaker 01

  Uh, for example, there are a lot of open source agent scaffolds, both coding agents and agent agent scaffolds that we use ourselves to test our models. And then we look at their logics, we look at their code to see how the design specific, both scaffolds and, for example, engines. And then we take what they worked on was, you know, really good. And then we reflect on how we think the problem, how we structure the problem. If we're on the same page and stuff like that. So we learn from each other.

• 34:37Speaker 03

  Do you think team underestimate how much like engineering discipline open models require comparing to using closed APIs? It always requires a lot of like setting up and it's a different compute. And you need to have a talent for that to use it, engineering talent, instead of just like, you know, choosing closed API, tune it on and and use it. Do you have any difficulty to do that? Or just inside the company, the open source stack is like established and and working?

• 35:06Speaker 01

  Personally, I don't have a problem with that. There are other open source models and if they publish, I'll just download it and and, you know, deploy it on a machine and then work with it if I want. Personally, I don't have that issue, but if there are, you know, personal developers out in the world, I understand the problem, especially when they don't have their own compute. And then it will be easier to, you know, connect to a model through, for example, open router and stuff like that.

• 35:31Speaker 03

  Do you use a lot of other open models on the same open router, let's say? Do you play with them?

• 35:36Speaker 01

  Yeah, I play with them. I would play with them day one. If they release at midnight, I play with them in midnight.

• 35:43Speaker 03

  Are you like taking notes?

• 35:45Speaker 01

  I don't actually take notes, but I do have my personal evaluation stack. Uh, a little bit of fun questions that I like to test with every single model to see how they.

• 35:54Speaker 03

  Can you tell me about it? That's super interesting.

• 35:56Speaker 01

  Um, yeah, I've been collecting a bunch of questions since I entered the company on different areas, including uh logical reasoning, including mathematics, proofs, including report writing, including agentic tasks and stuff like that a lot. And some of them are very, you know, I I just like to see how the model reacts to these problems and how they approach it. And different model have different personalities when approached.

• 36:22Speaker 03

  That's true, and you always need to adjust them. If we want to give like sort of a little guide to people who want to evaluate a model themselves, can you give me examples of the questions? Like five questions you need to ask the model to understand how it works, if it works well.

• 36:39Speaker 01

  From the professional evaluation perspective, five questions isn't enough. So if we want to do like, you know, very standard and there is fair comparison among models, uh, we have make it very a confident test. So there has to be a certain number of questions in each domain to see how the model performs. And usually you need to test it multiple times because models are not very stable themselves. If you're testing for fun, use the fun questions. But if we are actually assessing the model's capabilities, we need some sets. And that's, you know, very fair among different models. That's correct because some problems are not correct.

• 37:21Speaker 01

  Some questions, the the answers are not single, for example. Um, and sometimes when we run the test, the the environments are not fixed, for example. For example, the golden habit wouldn't pass and stuff like that. So if we're doing professional evaluation, you have to make make sure the uh evaluations is correct. It's diverse, it's above a certain threshold so that the test is. You know, confident.

• 37:44Speaker 03

  You mentioned characters. How do you work with your model character?

• 37:48Speaker 01

  I don't work on my models' characters. Okay, that's how I think of this issue. A general model should have all characters. Or like, it should be able to perform all characters. It might have default character. If the user wanted to be a different character, it should be. If the model injected into the system prompt, it should be. That's how I view this issue.

• 38:09Speaker 03

  I find it hard to adjust the new models because they're so different in terms of character all the time. I just don't even understand why it happens.

• 38:18Speaker 01

  I think it has to be be something related to the data that the model was trained on. The different patterns the models have been trained on. And also, different people, a different team might have their own, you know, constitution.

• 38:35Speaker 03

  In the system prompt?

• 38:36Speaker 01

  Or like, as the model's default behavior.

• 38:39Speaker 03

  If you look at open models in production today, I don't know if it's if it's a relevant question, but where do they fail first? Open models, specifically like reasoning, tool use, state tracking, evaluation, blind spots, like there are all those risks for open model. Where does it break first?

• 38:59Speaker 01

  I think open models are not very good at adjusting to different environments from what I see right now. We can see, for example, Claude, right? People use Claude in different coding environments and then people think they perform well in all environments where like different tool definitions and stuff. But I don't feel like the current open models can achieve that accuracy or like that level of understanding of the different environments.

• 39:24Speaker 03

  Why? Wh- where is the problem?

• 39:26Speaker 01

  I don't know how Claude does it, but for me, I think it is a solvable problem and we are working on it. We are improving it in two point two, but it's still not as good as, for example, Opus. But for two point five, it might be. We do have some systematic research going on in the area that has shown some results now, but it's still, it's not a concrete conclusion, so I won't say it.

• 39:50Speaker 03

  I'm so curious, but do you think it's the problem of compute? Because they have this infinite amount, they can just throw it at.

• 39:56Speaker 01

  I feel like compute is one side, but how we structure the problem and how we approach it is another side. And that's where we're more confident on that we can solve the issue.

• 40:06Speaker 03

  What can you tell me about M two point two? If it's launched by the time that the interview is out, can you give me some overview?

• 40:14Speaker 01

  Better coding, obviously, and better multilingual coding, obviously, and more stable than before. It has better performance in the area of two point one. In different areas is better, more stabilized, longer horizons and stuff like that. And we are testing it in different environments right now, and we believe that is better than before. So different coding environments, right? Even environments that we haven't seen before, even environments that are totally out of distribution. We see some very promising scores that are higher than two point one.

• 40:48Speaker 03

  I wonder how do you stay updated to everything that happens, which is super hard because the pace is just insane. You said when the models are out, you're playing with them. Do you read research papers? What is your like other interest that helps you kind of cross pollinate with what you do? Can you tell me how you stay up to date and what inspires you?

• 41:10Speaker 01

  There are different articles, different blogs going out every single day and a bunch, you know, all the information. How we deal with it is that we have a internal agent that tracks all the new articles and blogs and papers. And then it dispatches to different subjects and then it summarizes and then it analyzes and researches. So, so we have an internal researcher, if I call it, that does some filtering by itself. And then it gives what is filtered to us. And then we can improve the researchers if we think it doesn't do well. And that's how we, how we filter out a lot of information first.

• 41:49Speaker 01

  And then we play with new code repositories with coding agents so that we can understand it more quickly and then play with it more quickly. So we're pacing up with the, you know, with all the improvements with agents and with our models, for our models.

• 42:06Speaker 03

  That's fascinating. When you became a researcher, when you chose this path, what was you thought you would be doing? And what you actually doing, like, is it close to what you thought?

• 42:17Speaker 01

  That's a really good question. Uh, when I joined the team, I thought I would be reading papers every day because that's what I was doing during school, right? During a labs, we would read papers, come up with ideas, implement ideas, run experiments. If the experiments results are good, we run it as a larger scale. I thought I was, I was about to do that. But then what I realized was that when joining the company and then, you know, working for a couple of months, you already become pretty much top of the area or like of the industry. And then you have to come up with something that's really new. Or like you encounter problems that you just don't know how to, how to solve. It's not like you can read a lot of papers and then building up your thinkings on the papers.

• 43:01Speaker 01

  It's more like you need to really understand the problems from the fundamental and then think of it from the fundamental so that you can find the right solution. And another things would be that engineering is very, very, very important. I didn't know that during school because during school or during labs, it's more toys as compared to companies. It's not that scaled up. But when you really do scale up data, you scale up compute, scale up people, right? You encounter engineering issues that you need to tackle very beautifully. And engineering is very important. That's part two that was different from what I imagined. Pretty much these two, I feel like.

• 43:39Speaker 03

  When you work on the model currently, is it mostly that you're solving problems that you see immediately from your hands-on work? Or is it that the, the company says, oh, we have to achieve, let's say, opus results? How do you set the goals?

• 43:56Speaker 01

  Uh. We have a like a meta goal at the company level. For example, we want to improve the A S capabilities in you, you know, improving productivity, for example, because that's how like people view. So we have a company's mission. As a single researcher in the team, we have our own missions that we set our own goals with.

• 44:17Speaker 03

  What is your goal currently?

• 44:19Speaker 01

  For the next generation, I would be, I really want the model to to be working elegantly with experts. So it's more like better collaboration with experts, with developers. That's my goal as well, but that's maybe like two versions away, I think. I think we're um launching one version about per month or a month and a half, right? For longer horizon, we are, we are definitely working on it. But then for me, for the goal that I set on along that path, that's like a three months away thing. But for the better collaboration thing, that's like a one month or two months away thing.

• 44:53Speaker 03

  I wanted to ask you a little clarification question about interleaving learning. Well, you were talking at A I engineer also that the model doesn't set on like one action and constantly like in the loop of asking more questions and trying things. How do you look at it? Is it a continual learning? Is it part of it? What do we need to solve to have the model continuously doing this learning for like longer and longer horizons?

• 45:18Speaker 01

  That has some overlaps with the fine concept of continual learning. And by overlap, I mean, I think both conceptually and technically. But I don't feel like they are exactly the same or like the things that I talked about at the summit was not at the level of the full continual learning. It's more like on the path to that.

• 45:38Speaker 03

  How do you see being solved? Any ideas?

• 45:41Speaker 01

  We do think that's a different, different problem definition. Or like that's a different way of the model working with people. And we are working on that norm with our own defined question. But if I need to say how we approach it, I would say to, we would approach it through experiments. That's a very interesting question on continual learning and it's still very exploratory, right? That's definitely what we are going at, but then it has different phases or like different stages. We might approach stage one first while exploring more stages later.

• 46:21Speaker 03

  And they're not yet outlined the stages.

• 46:24Speaker 01

  Outlining stages, we do have our internal definitions that I didn't prepare today. I would say first would be to be more stabilized into a long horizon task and what I said at the summit, right? And then the next thing would be optimization.

• 46:39Speaker 03

  If you can repeat it because the people don't know what you said.

• 46:42Speaker 01

  So for example, we see a model, it receives, you know, environment feedbacks in a new environment. It needs to know what to explore and what environments to see because it's a partially observed environment. It needs to know the actions that it takes to receive better information and then better reactions and then perform harder, more complex tasks in the environments. That's like more of stage one, right? That's pretty simple. Basically, all the GPT models can do that to some extent, maybe not perfectly, but to some extent. And that's how we can actually solve it with our current algorithms. But we do see different norms of how a model improves itself in an environment that we don't have a concrete conclusion yet. Maybe in two point five, we will. That will be a different definition than what I said.

• 47:26Speaker 01

  The model itself would be defining its own goal. That's something that that would be different.

• 47:30Speaker 03

  Thank you so much. My last question is about AGI. Do you believe in AGI? And if yes, how does it look to you?

• 47:38Speaker 01

  Okay, that's a very large question. People talk about AGI and ASI every day. Actually, when I was interviewing with Minimax, when I was interviewing with our CEO, I said the same thing. He asked me the same thing, right? And then what I said was that I think people talk about AGI, people have different definitions of AGI, but we can only know the definition of AGI when we achieve it. Or like it is still progressing so fast that the definition even changes every day and people have different comments on it. But what I think is more important is we actually, you know, work towards it, work towards our own definitions of AGI. And as long as we we figure it out, it becomes true. And that's what I said during the interview and that's still my view today.

• 48:21Speaker 01

  The definition will become true when it becomes true.

• 48:24Speaker 03

  When we see it, we know it's AGI.

• 48:26Speaker 01

  Yes, exactly.

• 48:28Speaker 03

  But we're not there yet.

• 48:29Speaker 01

  No, there can still be better AI intelligence, for sure.

• 48:33Speaker 03

  Thank you. One more last question. What was the book that influenced you the most? And it can be a recent book or a book from your childhood.

• 48:41Speaker 01

  Let me just double check the name, though. Something like The Art of Creativity or something that I read um during uh during undergrad. So it's a long time. I don't remember the exact name.

• 48:51Speaker 03

  Yeah, there is a book, The Name Art of Creativity. How did it influence you?

• 48:54Speaker 01

  It opened up how I think of my own mind a lot. And then how I view the world and how I view problem solving. For me now, problem solving is is more of discovery. That's how I would summarize it in one quote.

• 49:06Speaker 03

  Thank you so much. Thank you for your time. Uh that was very interesting.

• 49:10Speaker 01

  Thank you for having me.

• 49:35Speaker 04

  Sunday mat and she's still working on the code. Experiment's running, curiosity overload. Layer by layer, love bros telling the tale. Find the gap between the theory and what the numbers reveal. She said the model tries to hack everything the sees. So we align it, we refine it, put the world at ease. Intelligence with everyone. Scale of humanity, another sun. From the open source, go the organ road. Cognitive revolution, let the story be told. Intelligence with everyone.

• 50:34Speaker 04

  Intelligence with everyone. Look and learn, adapt and turn, cut through the noise like a knife. Fifty-two cars deep, one conversation wide. The environment is noisy, but the model holds the ride. Ten billion strong, but running light as the breeze. Cost-effective, multi-agents doing what you please. Intelligence with everyone. Skilled up humanity. Under the sun. From the open source through the open road. Cognitive revolution, let the story be told. The intelligence with everyone.

• 51:23Speaker 04

  Intelligence with everyone.

• 51:32Speaker 04

  I see you in the morning, KTV at night. Problem solving is discovery in a different light. Scale to the theoretical extreme, push through. The definition becomes true and the work comes due. Engineering is everything.

• 51:52Speaker 04

  First principles, first principles. We build a future together.

• 52:04Speaker 04

  Intelligence with everyone.

• 52:12Speaker 04

  Intelligence with everyone.

• 52:25Nathan Labenz?

  If you're finding value in the show, we'd appreciate it if you'd take a moment to share it with friends, post online, write a review on Apple Podcasts or Spotify, or just leave us a comment on YouTube. Of course, we always welcome your feedback, guest and topic suggestions, and sponsorship inquiries, either via our website, cognitiverevolution.ai, or by DMing me on your favorite social network. The Cognitive Revolution is part of the Turpentine Network, a network of podcasts, which is now part of A16Z, where experts talk technology, business, economics, geopolitics, culture, and more. We're produced by AI Podcasting. If you're looking for podcast production help for everything from the moment you stop recording to the moment your audience starts listening, check them out and see my endorsement at ai podcast dot I N G.

• 53:10Nathan Labenz?

  And thank you to everyone who listens for being part of the Cognitive Revolution.