Sam (@Samxweb3og) header
Sam (@Samxweb3og) profile picture

Researcher | Web3 content creator | learner | programmer

Joined January 2023
Gmum Created this little piece of art for @get_optimum A stone bridge inspired by Optimum’s vision of building stronger connections across Ethereum. Sometimes, the biggest ideas start with a simple bridge. Here’s to making Ethereum more connected, one step at a time.
38
2
58
562
GprismaX A dataset can contain thousands of robot demonstrations and still fail to teach a robot the skill it needs. The reason isn't always the model. Sometimes, it's the data itself. Imagine training a robot to pick up and place an object. Some demonstrations show smooth movements and successful placement. Others contain hesitation, inconsistent grips, unnecessary corrections or failed attempts. If these differences aren't understood and handled properly, the dataset may not provide the learning signal the model needs. For robotics, collecting more data is only part of the job. Making that data useful is equally important. A good robotics dataset needs attention to: πŸ”Ή Trajectory Quality – Smooth, consistent robot movements that demonstrate the intended task. πŸ”Ή Task Diversity – Different objects, scenes and conditions to expose models to meaningful variations. πŸ”Ή Operator Consistency – Demonstrations performed with clear task objectives and repeatable procedures. πŸ”Ή Quality Validation – Reviewing episodes against defined standards before they become part of a dataset. This is where PrismaX puts its focus. Through its Robotics Data platform and VLA Foundry, PrismaX brings together: Structured Data Collection – Tasks, hardware setups and scenarios designed around training requirements. Teleoperation – Skilled operators capturing demonstrations on real robotic hardware. Egocentric Capture – First-person recordings of hands-on human activities. Quality Review – Episodes assessed against defined standards, with unsuitable data filtered out. The goal isn't simply to build a bigger dataset. It's to build a dataset that gives physical AI models something meaningful to learn from. Because in robotics, data quality isn't a finishing step. It's part of the foundation. @PrismaXai #PrismaX #Robotics #PhysicalAI
26
Gmum good morning Huge congratulations to everyone who received a role upgrade this week! πŸŽ‰β€οΈ A special shoutout to @Yuvraj_j2 and @rizuu015 Always great to see familiar faces moving forward and getting the recognition they deserve. And congratulations to all the new Refined members. Your consistency, creativity, and contributions are paying off. Keep creating, keep learning, and keep pushing the Optimum community forward. Well deserved, everyone ♾️ @get_optimum @discordflash @CryptoSundayz @TheBl0ckBoy
4
7
68
Gmum and GM, everyone! β˜€οΈ A fresh morning, a fresh mindset, and new opportunities to build something great. Grab your coffee, enjoy the little moments, and keep moving forward. Wishing everyone a productive day ahead! πŸ’™ Let's make today count with Optimum @get_optimum @discordflash @CryptoSundayz @tgogayi @aqccapital @TheBl0ckBoy @ada_pegasus
5
5
152
good morning CT have a good day happy weekend
9
1
13
168
Creating a prediction market is one thing. Resolving it reliably is another. A market can attract participants, generate trading activity and reach its closing date. But the most important question comes at the end: How do we determine the actual outcome? That's where @ultramint_xyz takes an interesting approach. Built on Base and powered by @fermah_xyz, Ultramint brings market creation, trading and outcome resolution into one connected workflow. Here's how it works: β†’ Create: Turn a real-world event into a YES/NO market with defined resolution criteria and data sources. β†’ Trade: Participants take positions on the outcomes as the market develops. β†’ Resolve: The Universal Resolution Engine evaluates the event using resolution rules, relevant data and AI-assisted evaluation where needed. β†’ Settle: The final outcome is recorded onchain, with settlement and rewards handled according to the market rules. What makes the resolution layer particularly interesting is its three-part approach: Defined Criteria β€” Clear rules and trusted data sources. Automated Evaluation β€” Data sources and rule engines help determine the outcome. AI-Assisted Resolution β€” An evidence-based AI Oracle can help evaluate more complex events using trusted information. Behind this workflow, Fermah provides the execution infrastructure, connecting the resolution process with flexible execution methods. The bigger picture is simple: A prediction market shouldn't stop at trading. It needs a reliable path from an event to its final outcome. That's what makes Ultramint worth exploring. Create the market. Trade the outcome. Trust the resolution. @fermah_xyz @ultramint_xyz @Sipra_ETH @7wealthh
4
7
29
Teaching a robot to pick up a cup sounds simple. Until the cup moves, the surface changes, or the robot has to adjust its grip. That's the reality of robotics. A task that looks easy to humans can require thousands of small decisions from a machine. And this is exactly why real-world robotics data matters. A robot needs to learn: Object Interaction – How objects behave when touched or moved. Force Control – How much force to apply while gripping. Motion Adaptation – How to adjust movements when something goes wrong. Environmental Adaptation – How to perform the same task in different environments. Simulation helps researchers test and develop robotic systems, but real-world demonstrations provide valuable experience that controlled environments may not capture. This is where PrismaX comes in. Through its Robotics Data platform, PrismaX focuses on making robotics demonstrations available to Physical AI teams. Its approach includes: πŸ”Ή Custom Data Collection – Demonstrations designed around specific tasks, robots and environments. πŸ”Ή Off-the-Shelf Robotics Data – Existing demonstration datasets for teams developing robotic systems. πŸ”Ή Dexterity & Precision Tasks – Data collection for challenging operations that require careful manipulation. The bigger opportunity is to make real-world robotic experience easier to collect and access. Because a robot doesn't just need to see the world. It needs to learn how to interact with it. @PrismaXai #PhysicalAI #Robotics #PrismaX
15
20
248
What makes Fermah Kernel interesting? Smart contracts can execute transactions. But they don't automatically coordinate everything that needs to happen before the next transaction. Think about a DeFi liquidation. A position crosses its liquidation threshold. Someone still needs to detect that condition, initiate the required computation, coordinate execution and deliver the result to the protocol. This is the gap @fermah_xyz is working to address with Fermah Kernel. Kernel is designed to turn conditions and inputs into coordinated, executable workflows. What makes its approach particularly interesting is that every step doesn't have to use the same execution method. A workflow can choose between: β†’ ZK proofs when cryptographic verification matters. β†’ TEEs when speed and flexible computation are more important. β†’ Hybrid execution when different steps have different requirements. The idea is simple: choose the right execution path for each step, while keeping the complete workflow connected. And this isn't just an architectural concept. Ultramint, built on Fermah Kernel, brings automated market resolution into a real application. Fermah Kernel is moving the conversation beyond transaction execution towards autonomous protocol operations. That's the part worth watching. @7wealthh @fermah_xyz
3
7
39
Proudly wearing my @get_optimum shirt at Web3 Week Asia @W3WOfficial and spreading the word about Optimum πŸ”₯
65
3
1
140
1,771
A robot can recognize an object, understand a command, and still fail to pick it up. Why? Because understanding what needs to be done and knowing how to physically do it are two very different challenges. Consider a simple task: picking up a small object. A robot must coordinate its vision, estimate the object's position, control its fingers, adjust its grip, and apply the right amount of force. This requires something more than intelligence. It requires high-quality real-world robotics data. This is where @PrismaXai comes in. PrismaX is building infrastructure to help solve this challenge: πŸ”Ή Custom Robotics Demonstrations: Collecting task-specific demonstrations for complex operations requiring dexterity and precision. πŸ”Ή Off-the-Shelf Robotics Data: Making existing demonstration data available for robotics development. πŸ”Ή Human-in-the-Loop Data Collection: Enabling real-world robot demonstrations through human operators and teleoperation. πŸ”Ή VLA Training Data: Supporting the data foundation needed to develop and improve Vision-Language-Action models. The bigger picture? PrismaX aims to bridge the gap between AI models that understand instructions and robots that can execute them in the physical world. Because smarter robots need more than powerful models. They need real-world experience. #PrismaX #PhysicalAI #Robotics
10
1
13
76
AI is moving beyond screens and into the physical world. But there's one major challenge: How do we teach robots to understand and perform real-world tasks? That's where @PrismaXai comes in. PrismaX is building the service layer for Physical AI, connecting robotics data, human demonstrations, and robot operators to help advance intelligent machines. Here's what makes it interesting: πŸ”Ή Robotics Data: Large-scale, diverse, real-world demonstrations for Physical AI builders. πŸ”Ή VLA Foundry: A network where operators capture real-world experiences and contribute to robot learning. πŸ”Ή Teleoperation: Humans remotely operate robots, generating demonstrations that AI models can learn from. πŸ”Ή Egocentric Data: First-person recordings of everyday tasks that provide another learning signal for embodied AI. The bigger picture? Making real-world experience more accessible to the teams building the next generation of intelligent robots. We're just getting started. Let's explore what @PrismaXai is building.
2
5
33
A closer look at the architecture behind @fermah_xyz. For a blockchain protocol, executing a transaction is only one part of the process. Before that transaction reaches the chain, several things need to happen: an event must be detected, the right execution path must be selected, and the action must be scheduled. Fermah approaches this through four operational layers of Protocol Agency Infrastructure. 1. Initiation The workflow begins by detecting an event or condition that requires an action. For example, identifying when a DeFi position becomes eligible for liquidation. 2. Routing Once an action is initiated, it needs to reach the appropriate execution environment or service. Routing connects the task to the right place. 3. Scheduling Not every action needs to execute at the same moment. Scheduling determines when a task should run, taking its timing and execution requirements into account. 4. Execution The final step is carrying out the action and completing the required operation onchain. These layers are connected. If initiation fails, there is nothing to route. If routing or scheduling breaks down, execution can be delayed. That's the architectural challenge Fermah is working on: moving beyond isolated transactions towards coordinated, composable execution flows. Because decentralised execution doesn't begin when a transaction reaches the blockchain. It begins much earlier. @fermah_xyz @7wealthh @fermah_xyz @7wealthh
1
2
20
Gmum Finally, the wait is over ❀️ After 5 months, I finally got the Refined role at @get_optimum Honestly, feels really good to see the effort pay off. Thanks to the Optimum team and everyone who supported me throughout. Also, thank you to everyone who congratulated me. Means a lot @get_optimum @discordflash @CryptoSundayz @tgogayi @aqccapital @TheBl0ckBoy
12
26
303
Gmum New week, new role upgrades! πŸŽ‰β€οΈ Big congratulations to @HanumD3e for becoming Optimized and @0xTynar & @akashranjan77 for earning the Chronicler role Also, congrats to all the new Refined members. Well deserved, everyone Keep creating and growing. More wins for the Optimum family @get_optimum @discordflash @CryptoSundayz @tgogayi @aqccapital @TheBl0ckBoy @ada_pegasus
13
2
23
218
It's DEEPAWALI, not Diwali. Just because some people can't pronounce the full word doesn't mean we have to shorten it. And to everyone saying "stop bursting crackers" every five minutes, STFU and let people celebrate. You don't have to celebrate the way we do, but you don't get to dictate how everyone else celebrates either. A little respect for our traditions and a little less unnecessary preaching. Happy Deepawali .
18
1
53
417
Wishing the entire Optimum community a very Happy Diwali! πŸͺ”βœ¨ May this festival of lights bring fresh ideas, new opportunities, and brighter beginnings. Here's to building stronger connections, creating meaningful innovations, and moving forward together. Wishing everyone a joyful and prosperous Diwali! 🧑 @get_optimum @discordflash @CryptoSundayz @tgogayi @aqccapital @TheBl0ckBoy @ada_pegasus
16
1
22
160
Gmum Everyone wants the Refind role 😭 And honestly, creating content deserves appreciation. But there's a little difference between simply posting and creating something that actually adds value. Admins aren't asking for more posts. They're looking for creativity, consistency, and quality. So keep creating, keep improving, and give the team something worth noticing. The role is nice, but quality content speaks louder πŸ‘€ No pressure, creators. Just keep pushing πŸ˜‚ @get_optimum @discordflash @CryptoSundayz @aqccapital @tgogayi @TheBl0ckBoy @ada_pegasus
7
13
96
A new mind joins the mission. πŸš€ A big welcome to Joanne Fuller, joining @get_optimum as the new Senior Data Scientist! With a PhD in Cryptography and experience at Consensys' R&D team, Joanne brings a strong background in formal verification, Ethereum, and deep infrastructure research. Her expertise will help Optimum dive deeper into network performance, understand the numbers behind it, and measure the real impact of better data propagation. Excited to see the insights and contributions she'll bring to the team. Welcome to Optimum, Joanne πŸŽ‰ @get_optimum @discordflash @CryptoSundayz @aqccapital @tgogayi @TheBl0ckBoy @ada_pegasus
Introducing Joanne Fuller, Optimum's new Senior Data Scientist! She comes from an academic background, earning her PhD in cryptography from Queensland University of Technology in Australia. From there she took her talents to Consensys' R&D team, working on both formal verification and The Merge. Now she brings her passion for numbers and deep Ethereum infrastructure knowledge to our data team, analyzing Optimum Network performance and quantifying the impact it has for our users.
10
1
15
229
Introducing Joanne Fuller, Optimum's new Senior Data Scientist! She comes from an academic background, earning her PhD in cryptography from Queensland University of Technology in Australia. From there she took her talents to Consensys' R&D team, working on both formal verification and The Merge. Now she brings her passion for numbers and deep Ethereum infrastructure knowledge to our data team, analyzing Optimum Network performance and quantifying the impact it has for our users.
127
34
79
319
29,288
How mump2p Changes the Way Ethereum Data Travels Ethereum nodes constantly exchange data to keep the network in sync. But here's the challenge: when a block is divided into pieces and shared across a peer-to-peer network, delays or missing pieces can slow down the entire process. This is where mump2p takes a different approach. Built by @get_optimum, mump2p uses Random Linear Network Coding (RLNC) to make data propagation more efficient. Here's the basic idea: πŸ”Ή 1. Split the data A block is divided into smaller pieces, called symbols. πŸ”Ή 2. Create coded packets Instead of simply forwarding individual pieces, nodes send mathematical combinations of those symbols. πŸ”Ή 3. Share across the network Peers can exchange these coded packets and even recode them before forwarding. πŸ”Ή 4. Reconstruct the original data Once a receiver has enough linearly independent packets, it can recover the original data. Why is this interesting? Traditional forwarding focuses on delivering specific pieces. RLNC allows useful combinations of data to travel through the network, giving peers more flexibility when exchanging information. And that's the core idea behind mump2p: improving how Ethereum data moves between nodes, without changing Ethereum's consensus rules. It's not just about processing blocks faster. It's also about making sure the data reaches the network efficiently. Better data movement. Stronger network propagation. @get_optimum @discordflash @CryptoSundayz @aqccapital @tgogayi @TheBl0ckBoy @ada_pegasus
22
27
181