When it comes to AI nowadays, everyone is already familiar with the upstream and downstream. When hyping up storage, we know to look at optical modules, materials, and equipment; when hyping up computing power, we know to look at NVIDIA, power supply, and heat dissipation.

However, when it comes to robots, most people's understanding still remains at those humanoid robots that can take a few steps on the Spring Festival Gala, thinking it's cool. But beyond that, many people can't say much about the upstream and downstream of the robot industry or what the "sage leaf" of the robotics industry is.

Actually, the robotics industry also has its own supply chain and is also branching out into more and more specific tracks. Some work on hardware, some on models, but there is another link that is rarely noticed by ordinary people yet may affect the upper limit of robot capability: data. More precisely, it is the data that provides "hands-on experience" for robots.

The rise of large models has a very simple premise: a massive amount of text, images, code, and videos has already been deposited on the Internet. The initial challenge for model companies was how to ingest this data, expand computing power, and train larger models.

Robots are different.

Robots do not have a natural internet corpus. Trajectories that can be directly used for robot control learning typically include observations, actions, and robot states; depending on the task, they may also include object states, contact information, success conditions, task semantics, and control frequencies. Real-world data collection is slow, expensive, dangerous, and highly dependent on specific robots.

And this is the background of Axis Robotics's founding. On July 27, this Physical AI data engine company announced the completion of a $12 million seed round of financing, led by Hack VC, with participation from Nomad Capital, Pi Network Ventures, 10K Ventures, and several angel investors. This investment is not betting on another team building robots but on a team dedicated to providing "experience" to robots.

In this context, BlockBeats interviewed Axis Robotics founder Chris. Chris previously served as COO of Chainbase, co-led Theia project, the first native encrypted base model on Hugging Face (an AI model community); prior to that, Chris was a venture capitalist at Vertex Ventures and a management consultant at Bain & Company.

From Data Business to Robot Data Gap

BlockBeats Question: Before starting Axis, what were your main entrepreneurial and professional experiences? How did you get involved in the direction of robots, Physical AI, and robot data? Was there a specific opportunity that made you realize there was an entrepreneurial opportunity here?

Chris: Before joining Axis, my previous startup was focused on data infrastructure, similar to Databricks, helping companies organize their data and unlock its value. The biggest realization from that experience was that while data may not be glamorous, it is often the key variable that determines a company's ceiling, deciding how far a company can ultimately go.

The beginning of 2024 was a crucial time for me. AI was trending, with everyone discussing models, parameters, and computing power. However, I increasingly felt that the industry was shifting from being "model-driven" to "data-driven." While models are certainly important, the factor that often determines whether a model can advance further is the data.

What really intrigued me was Surge AI. Established in 2020, the company surpassed $1 billion in revenue in just four years, outperforming even Scale AI and with minimal funding. I kept pondering: why could a data-focused company grow so rapidly? Have we underestimated the value of data in the age of intelligence?

During that time, I frequently engaged in discussions with friends from Nanyang Technological University, UC Berkeley, and NVIDIA, listening to their insights on the evolving landscape at the intersection of academia and industry. A clear consensus was gradually formed: while models iterate quickly, data serves as the foundation, being the most challenging, least standardized, and most easily overlooked layer. From the pretraining of large language models to the high-quality data required post-Agent emergence, data is not becoming less critical; quite the opposite—it is becoming increasingly complex.

Consequently, I began to wonder if Physical AI would follow a similar trajectory. The physical world is far more intricate than the textual world. Factors such as environmental changes, sensor noise, various edge cases, and diverse user habits amplify the data requirements exponentially. The volume of data needed for Physical AI and the error-correcting scenarios during post-training may be a hundredfold or more compared to today's AI.

However, if you were to ask who was working on Physical AI data at the time, hardly any names could be readily mentioned. In a race that could potentially birth a hundred-billion-dollar company in the future, there was no dominant player with a stronghold. For entrepreneurs, this signal was already crystal clear: it was worth diving in wholeheartedly.

Therefore, starting from the summer of 2025, I engaged in in-depth discussions with friends from NTU and UCB about the technical roadmap and product structure, and Axis Robotics truly took off at that stage.

BlockBeats Question: When discussing the robotics industry, external focus is often on hardware and models. Why does the Axis team believe that data is the "suzaku leaf" most likely to limit the scalability and implementation of Physical AI in this wave?

Chris: This is a great question. The development of large models gives us a very direct inspiration: once a technical path is validated, the next stage of capability enhancement often comes from the collective expansion of data, computing power, and model scale. From GPT-1 to GPT-3, we have seen that large-scale pre-training can bring about significant leaps in ability.

However, the robotics industry is still in the early stages. Today's hardware and algorithms have actually advanced quite a bit, with various robotic arms and humanoid robots continuously emerging. The model pathways are also gradually converging, such as VLA connecting vision, language, and robot motion; world models learning how the environment may change after an action; and WAM further integrating the prediction of future states with action generation. People are working in these directions.

The issue is that in order for a robot to not only perform in the lab but also work in a real-world environment, achieving a leap similar to the "GPT-1 moment" of language models, a much larger scale of real-world interactive data is required.

For example, today's robots are a bit like someone who has only learned to swim in a textbook, understands the theory but has never been in the water. If you want it to truly swim, you have to let it practice in different pools, different water temperatures, and different waves repeatedly. This "experience in the water" is the data.

However, the scale of publicly available data is still very small.

Open X-Embodiment is one of the representative open datasets, integrating 21 institutions, 22 types of robots, and over 1 million trajectories. While this may sound like a lot, compared to internet-scale data, it doesn't even amount to a fraction. Many other public datasets are still at the level of a few hundred hours.

Therefore, we believe that what is currently holding back the foundational models of robots is not necessarily the model structure itself, but the scale and diversity of the data lagging far behind. Without a sufficient amount of real-world data across scenarios and industries, even the best models and hardware can only run in circles in the lab. At this stage, data is not the icing on the cake but the key variable determining whether Physical AI can cross the inflection point.

BlockBeats Question: Can you use the simplest, most straightforward, and least circuitous way to tell the average person what Axis does?

Chris: In the most straightforward terms, Axis is essentially helping robots continuously accumulate "hands-on experience."

You can think of a robot as a newly hired intern. It is not unintelligent, but it has not done anything yet. To turn it into a seasoned pro, just showing it the instruction manual is not enough. It has to get hands-on experience, make mistakes, be corrected, redo the task, and practice repeatedly. What we do is systematically produce these experiences.

From pre-training preparation, task design, simulation exercises, to real-world data collection, and then data cleaning, error correction after model training, and continuous optimization, we have streamlined the entire process. We have also developed a web-based simulation platform and a mobile data collection tool so that ordinary people can participate. Folding clothes and organizing desks at home can all become learning materials for robots.

In short: Axis is a continuous data training supplier for Physical AI and also a data engine that continuously helps correct its errors, allowing robots to learn faster and make fewer mistakes.

Turning "Hands-On Experience" into Robot Training Fuel

BlockBeats Question: What do you think is the ultimate solution to the current Physical AI data gap? Which is better: first-person data, teleoperation data, or simulation data?

Chris: The biggest challenge facing the industry today is the difficulty of achieving a large-scale dataset, scene diversity, and real-world physical alignment simultaneously. The so-called real-world physical alignment requires that the action patterns in your data match those in the real world; you cannot be proficient in a virtual environment and then be "clumsy" on a real machine.

Let me give you an example. Simulated teleoperation data is like practicing driving on a driving simulator. The barrier to entry is low, no real vehicle is needed, and people from all over the world can generate a large amount of training data by remotely controlling virtual robots through a web interface. We can then mix and match robot forms, objects, scenes, and tasks randomly, quickly achieving diversity. However, the drawback is apparent: no matter how good the simulator is, it is not the real road conditions. The bumps, sudden situations, and irregular road surfaces in reality are challenging to replicate entirely in a simulator. If the model learns in a virtual environment, it may still struggle when put on a real machine, which is what everyone refers to as the "reality gap."

First-person real human data is like equipping a seasoned driver with a dashcam. Ordinary people can use their phones to collect real-life operations at home or in the office. The footage contains visual and verbal information that helps the model understand how humans actually work. However, the downside is that the footage lacks precise joint data of the robot, so it may not be sufficient for direct training of fine movements.

Real-world Remote Operation data means real on-road driving practice. It has the highest trajectory accuracy, closest to real-world physical interaction. Especially when the model itself fails during operation, human intervention to correct the residual trajectory is particularly valuable for model improvement. However, the cost is also the highest, requiring real-world testing and human effort. The capacity is limited and cannot be infinitely expanded.

So the answer is not which one is better, but that each of the three types of data has its own role: simulation scales up the volume, the first-person view helps the model understand common sense in the real world, and real-world remote operation is for precision calibration and closed-loop error correction. Only by integrating the three types of data into the same system is it possible to truly fill this gap.

This is also why we are doing "hybrid data sources and bidirectional closed-loop engine." We use simulation remote operation data and first-person view real human data for dual-supply, combined with Human-Gated DAgger. Its approach is very direct: when the model fails, it is intervened by a human to correct it, and then the correction result is sent back for training. This way, data is collected, cleaned, and enhanced, entering unified visual, language, and motion model training, then to virtual and real-world deployment, failure feedback loop, and data supplementation, forming a cycle that keeps moving forward by itself.

BlockBeats Question: Why did Axis initially choose to start with simulation data and is now beginning to layout first-person view data? Has there been a change in strategic direction?

Chris: This is an extension of the same data strategy. Physical AI must be data-driven, and the core metric of data is not quantity but diversity, as diversity determines generalization ability and robustness. We started with simulation because it is controllable, repeatable, and easy to evaluate. We can design tasks, adjust scenarios and robot morphology, collect data, replay, validate, train, and test. It itself is like a "world model," a virtualization of the real world.

First-person view data can complement the breadth of the real world. Recent research, including DreamDojo, has further shown the industry's potential for large-scale first-person view videos in learning human behavior and world rules. This type of data records how people use tools, handle objects, and complete tasks, behaviors that are scattered across different countries, industries, and life scenarios, making it difficult for a few labs to cover.

Simulation data is easier to produce on a few centralized platforms, but first-person view data is naturally decentralized and requires global contributors to participate. As top model companies like NVIDIA and DeepMind increase their demand for this type of data, the ability to continuously access, process, and validate first-person view data globally will become a critical capability.

More importantly, for Axis, the contributor network, task distribution, and data processing and validation pipelines are reusable. Moving from simulation to first-person perspective data is not a change in direction but a complement to a more complete robot data infrastructure.

BlockBeats Question: Can you walk us through the entire process of embodied intelligence data collection, processing, and training in Axis?

Chris: In Axis, data collection and processing are not fragmented into separate pieces; the entire process is an end-to-end closed loop. The recently released Axis V2 was a crucial upgrade for us: previously, Axis was more like a one-way data collection station; now, we have integrated task generation, data collection, model training, and evaluation optimization into a unified system.

The first step is question generation. We use algorithms in a simulation environment to combine robot body, target object, position, visual conditions, and other variables. For example, by combining 10 actions, 10 objects, and 10 placement methods, the diversity of tasks can be exponentially expanded.

The second step is question solving. Global contributors can complete simulation tasks through a simple web or mobile interface, or they can use a first-person view app on a mobile device to capture real-world operating processes.

The third step is grading and processing, which is also the most critical step. Raw data often contains jitter, invalid actions, and unnatural operations. We first clean, smooth, and resample the data. Then, leveraging RoboVerse, the same batch of data circulates between different simulation environments, migrating simulation data collected on the lightweight web to Isaac Sim for replay and data augmentation. As the data is replayed, the scene materials, lighting, camera angles, and physical parameters are adjusted. It's like taking the same recipe, using different pots, fires, and ingredients to redo it, expanding one original piece of data into many training samples.

After processing, instead of directly feeding it to a large model, the data undergoes success condition checks, anomaly filtering, and format standardization before entering model training. Once a training round is complete, we have the model perform tasks in simulation and evaluation environments, observing where it is prone to failure in certain scenarios and states, and then transforming these weak points into new targeted data collection tasks.

The model first self-executes the task, and once it deviates from the correct path, contributors take over and provide corrective actions. This error-correcting data is reintroduced into training, allowing the model to gradually learn how to deal with situations it was previously unfamiliar with. This flywheel keeps spinning, pushing the data, model, and robot's capabilities higher together.

So, the significance of Version 2 is not just the addition of several features, but transforming Axis from a system that merely "receives data" into a complete system that "helps the model get smarter."

BlockBeats Question: Axis recently released Dataset V1. Why is this dataset important, and what does it specifically demonstrate?

Chris: The most important aspect of Dataset V1 is not just that it added a batch of data, but that it initially validated something: simulated operations from a large number of ordinary contributors, after task design, success checks, filtering, smoothing, and data augmentation, can generate useful signals for robot pretraining.

V1 includes 207 operational tasks, over 50,000 trajectories, and over 60,000 task and scene variants. In the public LIBERO-Plus evaluation, after continuing pretraining with complete AXIS data, the overall success rate of π0.5 increased from 83.9 to 88.8, a 4.9 percentage point improvement. The RoboCasa control group with the same data volume was at 57.5. This at least indicates that under these evaluation conditions, model improvement depends not only on the quantity of individual data points but also on the diversity of tasks, scenes, perspectives, and perturbation conditions.

This year, we will release Dataset V2, which will further scale up and cover more robot morphologies, tasks, and scenarios.

BlockBeats Question: How does Axis determine the value of a piece of data? What are the general criteria for judgment?

Chris: We don't just look at whether a piece of data has been captured; instead, we continuously ask several questions: Is it clean? Does it bring something new? Can it be useful in the real world? And does it help the model address real limitations.

The first checkpoint is usability. Was the operation completed? Were there any hang-ups, disconnections in the middle? Can the action replay be reproduced? Was there any device lag, operation drift, or objects passing through in a way that doesn't align with physical laws? We will also use this batch of data to train a lightweight model for a quick check-up. If it cannot pass even this first checkpoint, this batch of data will not enter the training pool.

Check the second criterion for any new information. The biggest fear for a robot is encountering questions it has already mastered. We will examine how different this trajectory's scene, objects, and interactions are compared to existing data. If there are rare scene layouts, unique materials, or uncommon interaction methods, the value is high. While a large amount of repetitive "pick up and put down" actions are useful for establishing a baseline, the focus should be on complex tasks where multiple objects obstruct each other, requiring several consecutive steps to complete. Only this type of data can compel the model to generalize.

Proceed to the third criterion: transitioning from simulation to reality. The quantity of simulation data is not as crucial as ensuring it accounts for real-world variations. Will the model cope with changes in lighting? What if the material is altered? Different friction levels? Camera angles askew? The more these variables perturb the data, the less likely the model will falter when deployed on real hardware. If a segment of simulation data always occurs under ideal conditions, it can only serve as fundamental pre-training material and has limited value when transferred to the real world.

Lastly, evaluate the contribution to long-term iteration. When the model fails on real hardware and a human intervenes to correct a small portion of the trajectory, this correction is particularly valuable—it precisely pinpoints the model's weakness. Similarly, new tasks or trajectories from a different robot model can help the model rapidly expand its capabilities. Conversely, if the model has mastered a simple task, further inundating it with similar data yields diminishing returns.

Therefore, high-quality data is more than just being "format-correct." It must be clean, innovative, applicable to the real world, and capable of propelling the model forward. We aim to invest our limited data collection resources in truly intelligence-enhancing data for robots.

BlockBeats Question: Since Axis is a robotics company, why do you need blockchain technology? Axis has chosen the Base chain; could you elaborate on the reasons?

Chris: This is a very incisive and critical question. Our logic is straightforward: Axis is primarily addressing the data bottleneck in Physical AI, and to tackle this challenge, a vast number of global ordinary contributors need to participate. Blockchain technology is not the star of the show; it is more like an efficient foundational tool used for traceability, verification, incentivization, and distribution.

In the era of large language models, training data is often a "black box": Where did the data come from? How was it processed? How is a contributor's value recognized? It's challenging for outsiders to discern. However, Physical AI directly involves a robot's actions in the real world, where data quality impacts safety. Thus, this black box scenario is unacceptable in this context.

We have moved part of the data production process for Physical AI to Base in order to make this process more transparent. Base is a low-cost blockchain network built on Ethereum. Specifically, task IDs, data traceability IDs, user IDs, and their relationships will all be recorded. This way, every action trace and every contributor can be traced, audited, and receive corresponding incentives.

Furthermore, we need a network with broad community coverage and low entry barriers. The Base community is highly globalized, and the Base chain has low usage costs and fast transaction confirmations, making it easy for us to record contributions and distribute rewards.

More importantly, we are not rewarding mere quantity but high-quality contributions. For the robot model, junk data is not only useless but may even be harmful. We use a dual-score and point system to determine whether a piece of data has actually helped the model. The better the contribution, the higher the score, and the greater the reward. The efficiency and low cost of the Base chain perfectly match our need for high-frequency, small-scale incentives for global collaboration.

Who Will Foot the Bill for the Robot Data Loop

BlockBeats Question: What is Axis's current commercialization path? In terms of robot data collection, model training, and actual deployment, what products or services do you mainly provide to customers? What feedback have existing customers given on this model?

Chris: Our commercialization path is quite clear, focusing on serving three types of customers and providing end-to-end solutions based on their respective needs.

The first type is robot hardware and embodiment companies, such as Booster Robotics, Feagine Robotics, and AgiBot. They have their own hardware but often require more customized data and directly deployable models to complete their operational capabilities. We can help them with data collection, model training, and deployment starting from task design, and we also provide data solutions for their downstream customers.

The second type is companies that develop vision-language-action models or world models, such as Manycore Tech and SomaStacks. The role of world models is to help machine learning understand how the environment changes. They need large amounts of high-quality operational data to train universal models, and we provide them with high-quality task sets and datasets, as well as collaborate on joint training.

The third type is vertical industry enterprises, such as Lotus and Geely Auto. They are not concerned about having the most advanced models but about how to truly integrate robots into the production line. Therefore, we offer end-to-end automation solutions based on machine learning methods.

We value system capability highly, so our delivery is not just a data dump. We can adjust according to what the customer needs and where the internal capability boundary is: we can provide data and deliver data-driven solutions together. This not only enhances our competitiveness in serving customers but also more directly helps customers improve efficiency and quality.

BlockBeats Question: Who are Axis's partners and customers? When different customers come to Axis, what are their typical needs?

Chris: The types of companies I mentioned earlier are all our representative customers. They all seem to be saying "lack of data," but what they actually lack is not the same.

For example, a hardware manufacturer may have just developed a very good robotic arm. However, once it is deployed at a customer site, as soon as the lighting changes or the items are moved, the robot loses its precision. They come to Axis not just for a pile of messy videos but for structured data generated with multidimensional diversity, cleaning, and semantic annotation. When necessary, we also directly help them with training strategies. The ultimate result to be delivered is to make their hardware more stable and capable of generalizing in unstructured environments.

BlockBeats Question: Since the platform launched, can you share some operational data that has not been publicly disclosed yet, such as the number of registered contributors, task publication quantity, cumulative trajectories, and the activity and contribution of high-quality users?

Chris: It has been 15 weeks since the platform launched, and we have accumulated over 80,000 registered contributors. This number has been cleared of invalid addresses like robot accounts and represents genuine and valid data. We have released a total of 1600 machine learning tasks for pre-training and collected over 2 million data trajectories, which is roughly equivalent to 3500 hours of high-quality data.

But what we value more is user quality and stickiness. Currently, there are approximately 20,000 high-quality contributors, accounting for about 30%. In the past 30 days, 7800 of these users have remained active; in the past 7 days, 5300 have been active, with a weekly and monthly active ratio reaching 68%. In the last 30 days and 7 days, they have contributed roughly 680,000 and 160,000 trajectories, respectively.

This level of activity and engagement shows us that users are interested in the "training is contributing, and contribution can be incentivized" model and indicates that our community is forming a relatively mature and stable network of intelligent robot contributors.

IV. The True Inflection Point for the Industry and Axis's Position

BlockBeats Question: Over the past two years, various humanoid robots, VLAs, and Physical AI have been hot topics, but most ordinary people have not yet truly used robots. How long do you think it will be before robots see large-scale adoption? What will be the true inflection point: a reduction in hardware costs, a leap in model capabilities, or the maturity of data infrastructure?

Chris: We believe that the true inflection point will be the maturity of data infrastructure.

Hardware costs have actually been rapidly decreasing. The advantage of the Chinese supply chain has made the manufacturing cost of robots no longer an insurmountable barrier. But why haven't ordinary people adopted robots on a large scale yet? The core reason is that robots are still not smart enough; they lack sufficient common-sense understanding of the physical laws of the real world and human intent.

It's like having a low-cost car with a good engine but no driver's license and no experience on the road—a vehicle can't be driven well with just hardware and algorithms. Only when we can, like web crawling today, obtain interaction data from the physical world at a low cost and at scale, and turn it into digestible feed for models, will robots truly take off.

We believe this inflection point is getting closer, and what Axis wants to do is to push it further.

BlockBeats Question: In the robotics industry, large model companies and integrated machine companies may both build their own data in the future. Faced with these self-built data teams and Axis's current competitors in the same track, what is Axis's long-term moat? Is it a contributor network, task generation capability, data quality control, customer scenarios, or closed-loop training effectiveness?

Chris: This is a very critical and practical question. Large model companies and integrated machine manufacturers will definitely build their own data teams in the future; this is almost an inevitable result of industry development. However, the real competition lies not in who can or cannot collect their own data, but in who can establish a cross-scenario, scalable, and continuously optimized data infrastructure.

In the short term, everyone is competing based on data coverage and diversity; in the medium term, it's about efficiently turning data into model capabilities; and in the long term, it's about who can truly embed themselves in the customer's intelligent production system and become an irreplaceable part.

Axis's moat will not be just a single point, such as a contributor network, data collection efficiency, or data processing capability. Single-point abilities can all be copied. What we care more about is connecting these capabilities into a continuously strengthening closed loop and gradually embedding them into the customer's training process.

For customers, accessing Axis should be straightforward, with no complicated licensing required to quickly schedule tasks and access data. However, once customers begin using our simulation asset system, distributed collection network, and model correction feedback loop, the training processes on both sides will gradually intertwine. To switch from us, customers are not just changing data providers; they would need to rebuild the simulation infrastructure, reestablish the contributor network, data processing pipelines, and post-training correction processes. Our moat is not closed but rather a structural embedment.

We also highly value the ability to evolve alongside the models. If we were merely selling data, we could easily be replaced at any time. However, if we can continuously identify model deficiencies, design targeted tasks, provide post-training correction data, and effectively help customers improve success rates and robustness, we become part of the model optimization feedback loop. At that point, our relationship with customers transcends being merely a vendor and purchaser; it becomes more like co-builders.

In the next 18 months, the industry will face a significant shortage of large-scale, high-coverage data, where volume and diversity remain central. Looking ahead three years, the industry may require more specialized data for specific domains and scenarios. The assets we truly aim to accumulate are not a specific type of data but rather a programmable task generation system, a schedulable distributed contributor network, and a continuously optimized training feedback loop. It can horizontally expand into more industries and vertically delve into a specific vertical.

In summary, in the short term, we aim for scale; in the medium term, we aim for efficiency, and in the long term, we aim for embedment. Axis's goal is to become part of the Physical AI intelligent production system, rather than a data provider that can be easily replaced.

BlockBeats Question: In the 6–12 months after funding, what are the main tasks and goals for Axis? Is it to continue expanding data scale, validate more real robotic tasks, or onboard more paying customers? Looking back a year from now, what keywords do you hope the outside world will use to describe Axis?

Chris: In the next 6 to 12 months, we will concurrently advance our product, ecosystem, and commercialization.

On the product side, V2 has already transitioned us from a unidirectional collection platform to a complete training loop. The next step is to truly operationalize this system and scale it up. In September, we will further expand the first-person data collection pipeline, having already accumulated tens of thousands of hours of data and are in discussions with cutting-edge model labs in North America. Around October, we plan to release the V2 version of the dataset, covering more robot forms and atomic capabilities. By the end of the year, we aim to release the first large-scale Human-Gated DAgger post-training dataset.

On the ecological front, we will continue to expand our global network, enter the Latin American and European markets. In the next 6 to 12 months, we aim to maintain the stable daily production capacity of first-person data above 500 hours, increase the daily production capacity of simulation data to over 50 hours, and gradually expand the production capacity of DAgger post-training data.

On the commercialization side, our goal is to complete 2 to 3 new paid pilot projects by the end of the year and to be included in the preferred vendor list of foundational model companies early next year.

Looking back a year from now, we hope everyone will use the three words "scale, diversity, closed-loop" to describe Axis. Scale refers to our ability to consistently provide industry-level data supply; diversity means we have truly covered the complexity of the real world; closed-loop indicates that we not only generate data but can continuously optimize around model shortcomings.

More importantly, when the industry mentions Axis, we hope they will say, "This is a system that accelerates the evolution of robot models." We are addressing not only the issue of data quantity but also the efficiency of model evolution. Once a customer integrates with Axis, faster model iterations, higher success rates, and broader scenario coverage represent our value proposition.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia