Nick Rotenberg

📢 Robots Need More Than Perfect Conditions In a laboratory setting, everything is predictable. But the real world is full of unexpected situations. Objects move, the environment changes, and tasks do not always go according to plan. Robots need to learn to adapt, rather than simply repeating ideal movements. Real Experience Matters Physical AI needs diverse real-world data. Every movement, correction, and interaction can help robots become more capable. Where PrismaX Fits In Through teleoperation, @PrismaXai turns real robot interactions into structured training data. It is not just a matter of showing robots what to do. The point is to provide them with more experience from which they can learn. The future of robotics depends not only on more advanced hardware, but also on better conditions for learning.

📢 Rethinking Risk in DeFi I’m always interested in finding projects that offer a different approach to DeFi. I recently came across @mezzanine_fi and decided to look into what exactly they are building. I was drawn to their approach to something that matters to every DeFi user: finding a balance between risk and yield. So, what is Mezzanine? Mezzanine is a DeFi protocol that introduces risk-based tranching to onchain yield strategies. Each vault offers different levels of risk: > Senior Tranche: Lower risk, with priority in receiving returns and protection from junior capital > Junior Tranche: Higher risk, absorbs losses first in exchange for greater yield potential > Mezzanine Tranche: An optional middle layer between the two All tranches participate in the same underlying strategy but have different risk and reward profiles. Another interesting feature is the strategy-specific vault architecture, where each vault operates independently, keeping its assets and risks isolated. I like the idea of ​​giving users more flexibility, rather than treating risk as a one-size-fits-all concept. Of course, tranching doesn't eliminate risk. This changes the distribution of potential losses, so it is important to understand each strategy. Join now: app.mezz.finance/ref?code=CR…

gPrimus fam 👋 Privacy in Web3 is more than just hiding transaction details. In traditional DeFi systems, wallet balances, transaction amounts, and financial activity are visible to everyone. Transparency matters, but is it really necessary to disclose absolutely everything in every transaction? This is where FHE comes in. What makes FHE special? This technology allows computations to be performed directly on encrypted data, without decrypting it first. Within the @primus_labs ecosystem, this opens up several possibilities: ▪️ Encrypted balances: Financial information stays confidential ▪️ Private transactions: Less exposure of sensitive information ▪️ Confidential computing: Applications can work with encrypted data What is particularly interesting is that Primus is developing this infrastructure with the potential for integration into existing EVM ecosystems. My main takeaway Web3 should not have to choose between transparency and privacy. It is possible to verify important information without revealing personal data FHE could become a key component of the next generation of DeFi. I continue to explore Primus and discover new possibilities offered by this technology. There is still much more to look forward to. Explore more: s.kaito.ai/pzTDNWq Website: primuslabs.xyz #PrimusLabs #zkTLS #Web3

📢 When Faster Data Means Better Ethereum Performance Sometimes blockchain performance improvements don't require changing the network itself. They come from making information travel faster. @get_optimum has published its first individual operator case study with @BlockdaemonHQ, analyzing 2,185 Ethereum mainnet validators using mump2p. The results are impressive: ▪️ 39.7% fewer wrong head votes, dropping from 0.452% to 0.273%. ▪️ 366 ms faster block arrival, giving validators more time to react. ▪️ 3.4× tighter propagation spread, improving delivery consistency. ▪️ 92% of blocks were first received through mump2p. How does mump2p work? Rather than replacing Ethereum's existing libp2p gossip system, mump2p adds a faster communication layer using RLNC. This helps blocks and attestations travel more efficiently across the network. What's especially interesting is that Blockdaemon already had over 99% head-vote accuracy. Even so, mump2p reduced wrong head votes by nearly 40%, without changing its validator setup. My takeaway Blockchain performance isn't just about transaction throughput. It's also about how efficiently data moves between network participants. This case study shows that better data delivery can make a real difference to Ethereum's infrastructure.

📢 Prove Your Nomad Life @primus_labs has introduced a new set of verification tasks called Proof of Nomad Life. Explore now: s.kaito.ai/pzTDNWq Our digital life is distributed across dozens of platforms. We communicate on Discord, build professional networks on LinkedIn, and contribute to projects on GitHub. But what if all this activity could become part of a verifiable Web3 identity? That's exactly what the new Primus tasks are exploring. Three new ways to prove your digital lifestyle: ▪️ Discord: Verify your active timezone ▪️ LinkedIn: Verify your location and remote work availability ▪️ GitHub: Prove your participation in project groups. And the journey doesn't stop there Primus also offers other verification opportunities through: -> Agentic Ara -> Identity Profile -> Platform Activities -> Proof of PnL and other attestations I like the idea that our online presence can become something more than just a collection of accounts. It can evolve into a reputation backed by evidence. Time to explore the new tasks and build a stronger Web3 identity. #PrimusLabs #zkTLS #Web3

📢 PrismaX Office Hours #11 Recap On October 30, I attended another @PrismaXai Office Hours session hosted by @vivianrobotics where the discussion focused on how robotics data translates into real-world performance Think about how a person picks up an object. We adjust our position, slow down, pause, or correct our movements when something goes wrong. These small decisions carry valuable information for robot training. PrismaX is exploring this through Pi 0.5, combining the open-source YAMulator simulator with real hardware to test how learned behaviors translate into real-world performance. Quality Behind Every Episode Every demonstration goes through Verified Quality, where several factors are evaluated: ▪️ Task completion ▪️ Movement smoothness ▪️ Instruction following Successful attempts help build training datasets, while unsuccessful ones reveal weaknesses and areas for improvement. More Ways to Contribute PrismaX continues to expand its ecosystem of participants. Beginners can start with teleoperation and gradually move into validation as their tier allows. There are also opportunities to participate through the Regional Ambassador Program or by working on specialized data collection projects. Another interesting detail: the validator leaderboard will rotate monthly, encouraging consistent participation. Teleoperation Points and Validation Points will also remain separate systems. From Human Experience to Robot Intelligence My main takeaway is that better robots need more than large datasets. They need meaningful demonstrations, reliable validation, and real human experience. PrismaX brings these elements together, transforming human knowledge into something that robots can learn from. And I believe this is an important part of the future of physical AI.

📢 What If Data Could Be Used Without Being Revealed? The more I learn about privacy technologies, the more interesting this space becomes. In the previous parts of my @primus_labs journey, I started exploring digital privacy and discovered how zkTLS can verify online information without exposing unnecessary personal data. This time, I’d like to explore another concept: FHE So, what makes it special? Imagine you’re sending encrypted data to a service. Typically, for any calculations to be performed, that data first needs to be decrypted. FHE changes this approach. It allows computations to be performed directly on encrypted data, meaning that the information does not need to be revealed during the processing stage. Why does this matter? ▪️ More privacy when processing sensitive information ▪️ New possibilities for confidential AI applications ▪️ A way to work with data while keeping it encrypted I find it fascinating that privacy does not necessarily mean limiting the capabilities of technology. Thanks to FHE technology, we can envision a world where data remains protected even while it is being used. I am still exploring the Primus ecosystem, but this is yet another concept that makes me optimistic about the future of privacy. Explore more: s.kaito.ai/pzTDNWq Website: primuslabs.xyz #PrimusLabs #zkTLS #Web3

📢 Making Robotics AI More Accessible: π0.5 on Just 16 GB In robotics, one important aspect is often overlooked: the amount of hardware resources required to run advanced artificial intelligence models. But what if we could achieve the same functionality while using significantly less GPU memory? That's exactly what caught my attention in @PrismaXai latest work on π0.5 > The memory challenge Advanced robotics models require powerful hardware. The original π0.5 setup needed around 22 GB of VRAM, making it difficult to run on more affordable GPUs. > A smarter approach The PrismaX team identified unnecessary memory usage during model loading and inference. After optimizing the process, they reduced peak VRAM usage from around 22 GB to just 11.2 GB. No model pruning or distillation required. > Why this matters Now π0.5 can run on GPUs like the RTX 5080 and RTX 5070 Ti. This makes advanced robotics AI more accessible to developers and researchers without expensive hardware. We often focus on building smarter models and improving data, but efficiency is just as important. The easier AI is to run, the more people can experiment, improve it, and bring robotics into the real world.

Primus: A Guide to Getting Started @Primus_labs allows you to turn your Web2 activity into a verifiable reputation on the blockchain, while preserving your privacy. Here's how to get started 👇 1⃣Connect to Primus Visit the Primus campaign: s.kaito.ai/pzTDNWq Connect your wallet, and install the Primus Chrome Extension. 2⃣ Complete Verification Tasks Open the Attestation tab and verify your accounts across supported platforms, including X, Discord, TikTok, GitHub, ChatGPT, Binance, Spotify, and Luma. 3⃣ Protect Your Privacy Primus uses zkTLS to verify information from authentic websites without revealing your passwords or original sensitive data. 4⃣ Earn Reputation Points Successful verifications increase your Reputation Score and add to your on-chain profile on the Base network. According to the guide, repeated verifications may reward: > First verification: +10 points > Second verification: +4 points > Third verification: +1 point Check current tasks for exact rewards and gas requirements. 5⃣ Stay Active Check for new tasks regularly. Referrals may also provide additional points. I like the idea of ​​building a reputation in Web3 based on actual digital activity, without compromising privacy. Don’t just stop at publishing content. Complete the available attestations and explore everything Primus has to offer. #PrimusLabs #zkTLS #Web3

📢 Optimum Introduces TurboCards I like how @get_optimum is celebrating its Ethereum mainnet launch with something more than just an announcement. TurboCards is a collection of 999 NFTs, each representing a real Ethereum slot where Optimum helped a block travel faster across the network. Every card comes with stats showing how fast the block propagated and the economic advantage created along the way. Some mint codes are going to early supporters, while others will be distributed through Twitter and Discord drops. A cool way to celebrate a milestone and give the community something unique to collect. Mint page: turbocards.getoptimum.xyz/op…

📢 The Human Behind the Robot's Next Breakthrough The more I learn about Physical AI, the more I realize that the future of robotics isn't just about smarter models or better hardware. It's also about people. Here's why I think so👇 > Robots need real experience To learn how to interact with the world, robots need real examples of human actions, movements, and tasks. > Humans make a difference Through teleoperation, people demonstrate actions that become valuable training data for AI. > Quality matters Collecting data is only half the job. It also needs to be useful, diverse, and properly verified. This is why I find the work behind @PrismaXai so interesting. Its approach brings together data collection, quality verification, and a network of contributors working toward better Physical AI. My takeaway: Before robots can act independently, they need to learn from humans. I believe that the next major breakthrough in robotics will come precisely through this collaboration between humans and machines.

📢 Learning Primus: One Privacy Concept at a Time The more I explore @primus_labs the more I realize that privacy is not just about hiding information. It is about having the ability to decide for yourself what to share and what to keep to yourself. Think about logging into a platform or proving something about your online activity. Usually, you have to trust the service and share information to get what you need. But what if there were another way? That's where zkTLS gets interesting This helps verify information from websites using cryptographic proofs, making it possible to confirm specific facts without revealing all the underlying data. In my view, this is one of the most interesting ideas within the Primus ecosystem. We already have a vast amount of information online, yet we still need better ways to verify it, ways that also ensure your privacy is protected. I am still learning, and every new concept gives me another reason to delve deeper into this field. My biggest takeaway so far: the future of the internet shouldn't force us to choose between proving something and protecting our personal information. Explore more: s.kaito.ai/pzTDNWq Website: primuslabs.xyz/ #PrimusLabs #zkTLS #Web3