📢 The Race for Speed Is Moving Onchain Tokenized stock DEX volume has surged 10,163.7% YoY, reaching $48.7B over the past year. QQQb leads with $14.1B, followed by SPYx at $2.6B and NVDA at $2.4B. This growth shows that equity trading is moving onchain, bringing a familiar challenge: speed matters. Traditional finance invests heavily in low-latency infrastructure, but onchain markets require a different approach. This is where Optimum comes in Its data acceleration network is built to make information move more efficiently across decentralized systems. As onchain equities grow, @get_optimum has an opportunity to demonstrate the value of faster data delivery. The market is evolving, and the infrastructure behind it matters more than ever.

📢 PrismaX x Penn GRASP I'm excited to see PrismaX become a capstone partner of the @Penn @GRASPlab at the University of Pennsylvania. Through this collaboration, PrismaX Robotics Data will support research into reinforcement learning and adaptive robot policies, helping researchers explore how robots can learn from experience and respond to real-world challenges. Why does this matter? High-quality robotics data is essential for building smarter, more adaptable systems. By connecting industry and academia, @PrismaXai is helping support the next generation of robotics researchers and advance Physical AI. A great step forward for the future of robotic.

📢 The End of Copy-Paste Networking? The more I explore @get_optimum the more I realize that blockchain scaling isn't just about faster transactions or more powerful nodes. It's also about how intelligently data moves across the network. What if nodes could stop sending the same data over and over? That's where RLNC comes in. Instead of simply forwarding individual packets, nodes can exchange coded combinations that help reconstruct the original information. Why does this matter? -> More efficient data propagation -> Better use of network bandwidth -> Less unnecessary data transmission What I find interesting about Optimum is its focus on improving the networking layer itself, making data distribution smarter rather than simply pushing for more speed. Maybe the future of blockchain scaling isn't about sending more data. It's about making every transmission count.

📢 Teleoperation and Physical AI The more I explore teleoperation, the more I realize it's not just about controlling a robot. During sessions with @PrismaXai I began paying closer attention to how the robot responds to my actions. The smoothness of movement, gripper positioning, speed, camera angle, and timing can completely change the result. Where the value comes in Every movement and correction creates valuable data about human-robot interaction. That's what makes teleoperation so interesting: human experience becomes part of the Physical AI training process. Better demonstrations -> better data -> more capable robots For me, teleoperation is becoming more than just a way to interact with a robot, it is a path to learning through firsthand experience that simultaneously helps train the physical AI systems of the future. Discord: discord.gg/prismaxai Platform: app.prismax.ai

📢 PrismaX at IROS 2026 IROS 2026 brought together researchers, engineers, and teams building the next generation of robotics and Physical AI. @PrismaXai brought its VLA demo to booth #1122, giving visitors a chance to see the model running live and discuss the challenges behind training robots for real-world tasks. A key moment was @castorhat CEO & Co-founder of PrismaX, taking the Tech Talk Stage with “What Makes Good Robotics Data?” The main takeaway was simple: Good robotics data starts with a real problem. Teams first identify what their model still can't do, then define the task, environment, embodiment, and specific data needed to help the robot learn it. Instead of collecting data solely to increase volume, PrismaX focuses on obtaining high-quality, targeted demonstration examples that meet the models' actual needs. The IROS conference once again demonstrated that progress in robotics is not limited to creating larger models. It starts with better data.

📢 Optimum is touching down in Singapore A big week is ahead at TOKEN2049 Singapore. It will be particularly interesting to see the discussion featuring Muriel Medard, CEO & Co-Founder of Optimum, together with Kent Lin, Eli Laipson, and David Song. Data Availability, blockchain infrastructure, and new ways to make data transmission faster and more efficient are exactly the topics worth watching right now. For @get_optimum, TOKEN2049 is a great opportunity to meet builders, researchers, and partners and showcase how RLNC and Data Acceleration can reshape how blockchain networks handle data. If you’re attending, check out the Optimum events and join the conversation.

📢 A Fresh Look at PrismaX I took the time to explore the new @PrismaXai website, and I really like how much clearer the entire ecosystem seems now. The new design brings together Robotics Data, VLA Foundry, contribution, plans, events and resources in one place. What caught my attention most is the Robotics Data section. There are already 120K+ episodes, covering teleoperation and egocentric data across different tasks, environments and robot platforms. I also liked how clearly PrismaX now explains the different ways to contribute: -> Teleoperation Data -> Egocentric Data -> Custom Data Collection And the scale is impressive: ▪️ 1,700+ deployed scenarios ▪️ 5,000 hours/month capacity ▪️ $1M+ paid to contributors For me, the most important improvement is the simplicity and clarity. I’ve been following PrismaX for a long time, so it’s great to see the project bring all these elements together on a single site. Feels like PrismaX is entering a new chapter. Explore the new website: prismax.ai @vivianrobotics @MaxC16134

📢 What Would a Blockchain Look Like With RAM? We keep making blockchains faster at execution and consensus. But what about the layer that moves and accesses all that data? Think about a computer. A powerful CPU means little if it constantly has to wait for data. The same applies to Web3. -> Faster data propagation -> Lower latency -> More efficient data access This is where @get_optimum DeRAM comes in. DeRAM is a decentralized Random Access Memory layer designed to give applications real-time read/write access to blockchain state. Together with RLNC, mump2p and Flexnodes, Optimum is building infrastructure for faster data propagation, access and storage. Maybe the next step in blockchain performance isn't just more compute. Maybe Web3 needs RAM.

📢 PrismaX Office Hours Recap On Friday, I attended @PrismaXai Office Hours #10 and wanted to share a few moments that stood out to me. Robotics Data Catalog PrismaX has already opened the Robotics Data Catalog in limited release. There are now more than 120K episodes and 92,000 hours of data. Teleoperation and egocentric episodes go through review under the same standard, and the catalog continues to grow every month. Custom Data Collection Another important update is custom data collection. Teams can now define their own parameters: ▪️ Task ▪️ Environment ▪️ Embodiment ▪️ Difficulty This makes it possible to collect data not in a fixed format, but rather in a way that takes into account the specific needs of researchers, developers, and robotics engineers. Why data matters The number of humanoid installations grew from around 2,000 in 2024 to 15,000 in 2025, while Barclays projects 60,000 next year. But more robots also means a greater need for high-quality data to train Physical AI. Human validation Even as PrismaX scales, human involvement in the process remains. Episodes undergo a multi-stage review, during which parameters such as smoothness, variety, and format are evaluated. My main takeaway is simple: in robotics, what matters is not just the volume of data, but also its quality, consistency, and verification. Right now, I’m especially interested in following the Robotics Data Catalog and IROS 2026 @vivianrobotics @MaxC16134

📢 PrismaX Introduces Robotics Data @PrismaXai has introduced Robotics Data, a new solution for teams developing Physical AI systems. It is particularly noteworthy that this is not merely a collection of robot demonstrations. The main idea is usable, diverse, and proven data that can actually help models learn. What’s included? ▪️ Teleoperation ▪️ Egocentric capture ▪️ Custom data collection ▪️ Human review Teams can explore the existing data bank or request datasets built around specific tasks, robots, and environments. Pricing Plans PrismaX also introduced several ways to access Robotics Data: -> Discover Subscription access to the growing robotics data bank. -> Researcher Extended access for teams that need to explore and use available demonstrations more deeply. -> Enterprise A customized option for teams that need custom data collection, specific tasks, robots, and environments. This approach makes it possible to start with existing data and subsequently move on to creating datasets tailored to specific needs. Data is more than a recording The demonstration contains much more than just the robot's final movement. Camera placement, object position, operator decisions, the environment, and task design all affect what a model can learn. That’s why PrismaX connects data collection with training and evaluation. A New Step for PrismaX Robotics Data is now in limited release, starting with the first group of teams building the next generation of Physical AI. For me, the most interesting part is simple: It’s not just about collecting more data. It’s about creating a higher-quality experience that robots can learn from. Explore the data and request access: prismax.ai/#plans @vivianrobotics @MaxC16134

📢 Your Blockchain Is Fast. Your Network Isn't We often measure blockchain performance by TPS, execution speed, or block time. But there is another question: How fast does the data actually move? A block can be created quickly, but it still needs to reach validators across the network. Every millisecond of delay can affect propagation, validator performance, and the efficiency of the whole network. This is where @get_optimum focuses. With mump2p and RLNC, Optimum is building a smarter data propagation layer designed to move blockchain data faster while using less bandwidth. Because scaling isn't only about processing more. It's about moving more data, faster. Your blockchain might be fast. Your network should be too.

📢 Why Validation Matters for Physical AI The more I validate robotics datasets on @PrismaXai the more I realize how much can hide behind a simple task. During the check, I focus on more than just the final result. It is important to understand exactly how the robot performed the task and whether this episode can provide useful data for further learning. As a validator, I focus on a few key things: -> Did the robot follow the instructions correctly? -> Were its movements smooth and controlled? -> Did the robot stay fully visible in the frame? -> Were the camera feeds clear and synchronized? -> Can this episode serve as a useful example for training? For me, validation isn't just about giving a Pass or Fail. The goal is to ensure the reliability, high quality, and utility of data for training robots to perform tasks in the real world. Sometimes, even the smallest details in a single episode can play a crucial role. Discord: discord.gg/prismaxai Platform: app.prismax.ai @vivianrobotics @MaxC16134

📢 WHEN ROBOTS MAKE MISTAKES The more I study the field of Physical AI, the more convinced I become of one thing: A robot does not get a second chance. An LLM can give you a wrong answer. A robot can turn a wrong prediction into a real physical action. A small position error, a wrong movement, or a poorly learned behavior can damage the robot or something around it. That makes the quality of training data much more important. Data Shapes Behavior Robots learn from real actions. ▪️ What they see ▪️ How they move ▪️ How a task is completed What happens when something goes wrong. Thus, the question is not just about the volume of data we collect. What matters is the behavior that data trains. This is one of the reasons I find @PrismaXai interesting. Through teleoperation, real-world actions can be collected, structured, and turned into training experience for Physical AI. -> Better data -> Better behavior -> Safer robots The future of robotics will depend not only on improvements in hardware. It will also depend on what we teach robots. @vivianrobotics @MaxC16134

📢 Twinstake Joins mump2p Testing @get_optimum has added @twinstake_io as a new mump2p testing partner. As an institutional staking provider, Twinstake runs high-performance Ethereum validator infrastructure and has reported MEV APR around 35% above the network average over the past 12 months. Now its infrastructure will be tested with mump2p, using RLNC to help Ethereum data propagate faster and with less unnecessary bandwidth. Another real operator. Another real-world test. It’ll be interesting to see what faster propagation can unlock for validator performance.

📢 PrismaX at IROS 2026 IROS 2026 is getting closer, and @PrismaXai will be in Pittsburgh at Booth 1122. Interestingly, PrismaX is not merely aimed at collecting additional data in the field of robotics, it focuses on ensuring that this data is genuinely useful for training. From Task to Data Teams can define the task, environment, robot, and difficulty, with data collected through teleoperation and egocentric capture. Every episode then goes through a quality check: -> Smoothness -> Diversity -> Format -> Task completion If the data doesn't meet the standard, it doesn't make the training set. Learning the Process PrismaX also works with sequences of interconnected tasks, helping models master holistic processes rather than individual, isolated actions. If you're at IROS, Booth 1122 is where to find PrismaX. CEO @castorhat will also speak on the Tech Talk Stage on Tuesday the 29th. @vivianrobotics @MaxC16134

📢 Learning Through Teleoperation The more I practice teleoperation with @PrismaXai the more I realize: it’s not just about controlling the robot. It’s about learning to think along with it. With every session, I pay more attention to: -> Precision of movement -> Camera perspective -> Gripper positioning -> Timing and control I stopped rushing and began observing how the robot reacts to every decision. The more I practice teleoperation, the better I understand what a significant part of the future of robotics could be built upon these small, human-made decisions. And I continue to learn. My next goal is not simply to complete another task. It is to complete it better than the last one. Discord: discord.gg/prismaxai Platform: app.prismax.ai @vivianrobotics @MaxC16134

📢 Optimum x DSRV A Real-World Test for mump2p When I saw @get_optimum announce @dsrvlabs as a new mump2p testing partner, it caught my attention. 300K+ ETH staked. Institutional-scale infrastructure. Strong presence in Korea. Why it matters DSRV runs professional validator infrastructure with a strong focus on uptime. Korea also offers an interesting environment for testing blockchain networking, where geography, peering and latency can affect how quickly data reaches validators. mump2p in the real world Optimum mump2p uses RLNC to improve data propagation without replacing the existing GossipSub setup. Now DSRV gets to test it under real validator conditions. Optimum brings the technology. DSRV brings the infrastructure. Another serious operator. Another real-world test. That's how infrastructure gets proven.

📢 There Are More Robots. But What Are They Learning? Lately, I’ve been thinking less about the number of robots being deployed and more about the experience they create. The number of installed humanoid robots grew from around 2,000 in 2024 to 15,000 in 2025. This figure is expected to reach approximately 60,000 this year. But the number of robots isn't the most interesting part. What matters more is the experience they create Most of them are not yet operating in factories. They are used for research, experiments, and data collection. -> They move -> They see -> They act -> They make mistakes And that experience becomes training data for the next generation of Physical AI. But more data doesn't automatically mean better data. Quality, diversity, different tasks, environments, and operators all matter. That's where PrismaX comes in Through teleoperation, real robot actions become structured training data. For me, @PrismaXai is not just a way to control robots. It is the creation of an internet of physical experience. The number of robots is growing. Now, their experience must scale along with them. Discord: discord.gg/prismaxai Platform: app.prismax.ai @vivianrobotics @MaxC16134

📢 Why Validation Matters for Physical AI More robotics data doesn't automatically mean better AI. A robotics demonstration can look successful while still containing: ▪️ Poor camera quality ▪️ Failed grasps ▪️ Incorrect movements ▪️ Incomplete tasks That’s why validation matters. What Validators Look For It's not just PASS or FAIL. Validators check whether the instructions were followed, the robot stayed visible, the cameras were synchronized, and the execution was smooth and complete. The Bigger Picture Every accurate review helps @PrismaXai separate valuable training data from noise The idea is simple: Better validation -> Better data -> Better Physical AI. Behind every capable robot is not just a powerful model, but high-quality experience to learn from. And that's exactly what PrismaX is helping build. Discord: discord.gg/prismaxai Platform: app.prismax.ai @vivianrobotics @MaxC16134

📢 The Blockchain Speed Limit Nobody Talks About Everyone talks about TPS. But what if TPS isn't the real speed limit? A blockchain can process thousands of transactions per second, yet still feel slow if the data takes too long to reach the network. How fast does the data actually move? A new block is created. Validator -> Network -> Everyone else And this journey happens every time. The faster the information spreads, the faster the network can react. Bigger blocks. More data. More pressure. As blockchains scale, they don't just process more transactions. They move more: ◾️Transactions ◾️ Blobs ◾️ State ◾️ Data Simply adding more bandwidth isn't always the smartest answer. The better question is: Can we make data move smarter? This is where Optimum comes in With technologies like RLNC and mump2p, @get_optimum is working on a smarter way to move blockchain data, making propagation faster and more efficient. Because the next blockchain speed record might not be measured in TPS. It might be measured in milliseconds.