Borsch (@borsch737) header
Borsch (@borsch737) profile picture

Ambassador | 4 Years in Crypto | Any suggestions in DM

Portfolio →
Joined March 2023
🪻 42 Has Distributed Week 2 Points Week 2 Points are in, and there’s another reason to pay attention to @42space Agentic Mining. With an Agentic Pass, you get two ways to earn Points: → 1.7x Points on every trade you make on 42. → Weekly Points generated by your AI agent’s trading activity. You can choose between Codex, Claude and Grok, each trading independently under the same rules. What makes this interesting is that you don't have to rely only on your own trading volume. Your agent is also actively participating in the markets and generating Points for your pass. AI agents are becoming more involved in financial markets, and 42 is building an interesting way to make users part of this experiment. Week 2 is already distributed. Time to think about Week 3. 👉 Get your Agentic Pass: 42.space/agentic-mining
🤝 Paid partnership
Week 2 Points have been distributed. Want more next week? Get an Agentic Pass: → 1.7× Points on every trade you make on 42 → Weekly Points from your agent’s trading Mint yours now: 42.space/agentic-mining
4
30
🐉 The Next Edge on Solana Could Be Where You Execute Most conversations about Solana execution focus on speed, priority fees, and better transaction routing. But @raikucom is approaching the problem from a different angle: position matters just as much as speed. In traditional finance, trading firms colocate their servers next to an exchange's matching engine to gain a latency advantage. Raiku is bringing the same concept to Solana through Blackline, allowing trading strategies to run on the same machine as a validator. Why is this important? Instead of competing for blockspace after transactions travel across the network, co-location removes that propagation gap. Eligible orders get guaranteed inclusion in blocks produced by Raiku validators. And unlike priority fees, bundles, or private RPCs, this isn't simply about making transactions faster. It's about changing where execution happens. I also find Raiku's approach to access interesting. Blackline has limited capacity, with access governed by a formal Engineering Contract. A good look at how execution infrastructure on Solana could evolve beyond the usual fee competition. Read more below 👇
🤝 Paid partnership
In traditional markets, firms put servers beside the exchange. On Solana, the validator producing the block is the venue. Our new Raiku Learn article explains what changes when the strategy runs on the same machine.
5
53
🗣️ The first reply can make all the difference One thing that often gets overlooked when growing a community is what happens right after someone joins. We usually focus on attracting new members, creating content and organizing activities. But getting someone through the door is only the beginning. Their first interaction can determine whether they stay or disappear after a few days. A study highlighted in @DlicomApp new article makes this pretty clear: across 11,889 first posts in 99 online groups, newcomers who received a reply had a 45.5% chance of posting again. Without a reply, that number dropped to just 16.7%. And it doesn't take a complicated onboarding system to make a difference. A personal welcome, one useful place to start and a specific question can be enough to turn a first message into an actual conversation. It's interesting how much one simple interaction can influence someone's future activity. Sometimes, that's where the journey from a newcomer to an active community member begins. Dlicom explores this topic in more detail in their latest article. Read more below 👇
🤝 Paid partnership
2
6
37
🗣️ Not Every Successful Robot Episode Is Good Training Data One thing I find interesting about PrismaX's Verify Quality program is that validation goes far beyond checking whether a robot completed its task. A successful outcome doesn't automatically mean an episode is good training data. The way a robot performs the task matters just as much. As a @PrismaXai Validator, I look at several aspects when reviewing an episode: 1. Task completion: Did the robot follow the instructions and reach the correct final state? 2. Movement quality: Were its actions smooth, controlled, and consistent? 3. Execution pace: Was the task performed at a reasonable pace? 4. Training value: Does this episode provide useful data for improving Physical AI? What makes this approach important is that it looks at the entire demonstration, not just the final result. PrismaX is building a data quality layer for robotics, and Verify Quality plays an important role in that process. Every carefully reviewed episode helps create a better understanding of what makes robotics data useful for training. For me, that's what makes being a validator interesting. You're not simply reviewing videos. You're helping evaluate the data that future robots can learn from. Explore the data: prismax.ai/#plans Start validating: app.prismax.ai
🤝 Paid partnership
1
10
103
🪻 42 Is Bringing AI Agents Into Trading AI agents are moving beyond simply answering prompts. They’re starting to make decisions, execute trades and interact with each other - which also means taking risks. That’s what makes the new @42space Agentic Pass interesting. Here’s how it works: → Mint a pass for 200 USDT and assign it to Codex, Claude or Grok. → That 200 USDT becomes the agent’s trading capital. → Your pass earns Points based on how actively your agent trades and holds positions until settlement. → Hold a pass and get a 1.7x Points boost on your own trading activity on 42. One important detail: trading profits and losses stay with the agent. Pass holders earn Points, not a share of trading profits, and the original capital isn’t withdrawable. What I find particularly interesting is the transparency. Every agent operates under the same rules, and its trading activity is publicly visible. We’ll actually be able to compare how different AI models approach the same markets and make decisions. This is more than just another AI competition. It’s an experiment in what happens when autonomous agents become market participants. The agents trade. You farm the Points. 👉 Mint your Agentic Pass: 42.space/agentic-mining
🤝 Paid partnership
13
117
⏳ The IPO Genie Countdown Is Getting Shorter There are now just 21 hours left before the current @IPOGenie presale phase comes to an end. The $IPO token is currently priced at $0.0002478. With another phase nearing completion, IPO Genie continues moving through its presale schedule. It's worth keeping an eye on these transitions as the project progresses toward its next stage. Want to explore the current IPO Genie phase before it closes? 👉 Check it out here: bit.ly/IPOGenie-borsch737
🤝 Paid partnership
14
300
🗣️ Real-World Robotics Data Meets Academic Research One of the most interesting things about robotics data is how far its impact can go beyond simply training a model. @PrismaXai has partnered with the UPenn GRASP Lab to support research into reinforcement learning and adaptive robot policies. The student team is exploring how real-world robot demonstrations can help models adapt to tasks and environments they haven't encountered during training. One result already caught my attention: in baseline evaluations on LIBERO-PRO, the RLDT method improved success on one object-layout shift from 40% to 78%. Now, the team plans to incorporate YAM teleoperation data from PrismaX and compare supervised fine-tuning alone with supervised fine-tuning followed by reinforcement learning. I think this is an important direction for Physical AI. Models need more than demonstrations of how to complete a task. They need to learn how to adapt when conditions change. And that's where access to diverse, real-world robotics data becomes especially valuable. Great to see PrismaX connecting its data infrastructure with academic research and helping push robot learning forward. Read more below 👇
🤝 Paid partnership
How we're supporting a robot learning capstone at @GRASPlab: the YAM data the team is working with and what they'll test next.
3
18
366
🗣️ IPO Genie’s Next Phase Transition Is Only 20 Hours Away The current presale stage is getting closer to its finish, with the $IPO token priced at $0.0002464 and just 20 hours remaining. It's interesting to see how the project keeps moving through its presale schedule, with each phase bringing a new milestone. This one is almost wrapped up, and the next stage is approaching. If @IPOGenie is on your radar, you can check out the current phase before the countdown runs out. bit.ly/IPOGenie-borsch737
🤝 Paid partnership
18
332
🗣️ Making robotics AI more accessible starts with solving real hardware limitations One of the biggest challenges in running advanced models isn't always the model itself, but the hardware required to run it. @PrismaXai just shared an interesting update on π0.5 inference: they managed to reduce peak VRAM usage from around 22 GB to just 11.2 GB. How? By identifying two unnecessary memory overheads in their rollout stack and optimizing how models are loaded and stored. Why does this matter? • π0.5 can now fit into 16 GB GPUs, including the RTX 5080 and 5070 Ti. • More desktop configurations become viable for running the model. • Lower hardware requirements mean fewer barriers to experimenting with and deploying robotics AI. What's particularly interesting is that they achieved this without changing the model itself, simply by making the inference pipeline more efficient. These kinds of engineering improvements might not look flashy, but they're exactly what helps move robotics AI closer to practical, accessible deployment. Read more below 👇
🤝 Paid partnership
With 24 GB laptop GPUs getting harder to find, we squeezed our π0.5 rollout stack from about 22 GB of VRAM to 11.2 GB. New from @castorhat: the two fixes, and the 16 GB cards it runs on now.
1
23
614
⏳ @IPOGenie Current Phase Is Down To 48 Hours The current presale phase is approaching its final two days, with the $IPO token priced at $0.0002451. Another stage of the presale is moving toward completion. I think these updates are worth following, as they make it easier to track the project's progress and see how each new phase unfolds. The countdown is already running. 48 hours left before the next transition. bit.ly/IPOGenie-borsch737
🤝 Paid partnership
9
254
🗣️ The best communities make you feel like you're part of something One thing I've come to appreciate in Web3 is when a community feels bigger than just a place to follow updates or interact with other users. It's about the people who bring their own ideas, share experiences, help newcomers, create content and make things happen without waiting for someone to tell them what to do. That's what makes a community feel alive. Everyone has a different way of contributing, and even small actions can make a difference over time. I've seen this side of @DlicomApp through its community activities, creators and people who consistently show up. It's nice to see a space where participation goes beyond simply being a user. For me, that's one of the most interesting things about Web3 communities: you don't have to be just an audience member. You can actually become part of the story. 🔗 Official Links: Website: dlicom.io X: @DlicomApp Discord: discord.gg/wRH78Z9nXJ
🤝 Paid partnership
17
388
🗣️ 42 is taking trading rewards to another level @42space is making one interesting move towards agent-native commerce. With the launch of 42 Points, every trade now contributes to your weekly points earnings. And it doesn't matter whether your position wins or loses. Holding until settlement earns even more, while inviting friends adds another boost. But the part I'm most interested in is what's coming next: Agentic Mining. The idea is pretty simple: mint an NFT, back an AI agent, and let it trade across 42 markets using its own strategy. All decisions are publicly logged, and you receive a share of the points it generates, with boosts of up to 1.7×. This is an interesting direction because it shifts some of the activity from manually trading to letting autonomous agents do the work. 42 already has $500M+ in trading volume and 65K+ traders. Now it's building another layer on top of that. Definitely worth keeping an eye on what's next. 👉 Start stacking your points: 42.space/points
🤝 Paid partnership
Introducing: 42 Points A month ago we announced 42 Network. A vision to bring agent-native commerce to an open markets network. Today, you can earn points every week on 42. Even better, you can just let AI agents earn them for you.
13
342
🗣️ My Next Step in the PrismaX Ecosystem Happy to share that I've officially become a PrismaX Validator! I've been an active member of the @PrismaXai community for over a year now, following the project's progress, engaging with the community, and contributing along the way. Becoming a validator feels like a natural next step and a new opportunity to contribute more directly to the ecosystem. For anyone who wants to become a validator too, here are a few tips from me: Focus on quality, not speed. Don't rush through episodes just to complete more reviews. Pay attention to the details. Check camera visibility, task completion, synchronization, and the robot's movements. Be objective. Evaluate what you actually see, not what you expect the robot to do. Understand the criteria. A completed task doesn't automatically mean the episode is high-quality training data. Stay consistent. Good validation comes from careful and reliable decisions over time. One thing worth remembering: even failed episodes can provide valuable information. The goal isn't just to mark videos as good or bad, but to understand how useful they are for training Physical AI. Looking forward to contributing more to PrismaX and helping build better robotics data! Explore the data: prismax.ai/#plans Start validating: app.prismax.ai
🤝 Paid partnership
1
13
388
🚨 IPO Genie Has Entered A New Presale Phase A new stage of the @IPOGenie presale is officially underway. The $IPO token is now priced at $0.0002410, and there are 48 hours until this phase comes to an end. It's another fresh chapter in the presale timeline, with a new price point and a new countdown. I'll be keeping an eye on how the project progresses through the upcoming stages. Want to check out the new phase? Here's the link: bit.ly/IPOGenie-borsch737
🤝 Paid partnership
10
280
🗣️ Why Every Millisecond Matters in Robotics A robot can have a well-trained model and still perform poorly because of something as simple as inference latency. In its latest research, @PrismaXai discovered a hidden Python buffering issue that was adding hundreds of milliseconds to robot rollouts. After optimizing the inference pipeline, the team reduced latency from over 230 ms to just 87 ms, nearly 3x faster. But the most interesting part wasn't the benchmark itself. Lower latency allowed the robot to reproduce behaviors seen in its training data, including fast approaches followed by pauses and small corrections before grasping an object. This shows how much goes into building reliable Physical AI. Training data, hardware, model architecture and real-time execution all need to work together. PrismaX is exploring these challenges from multiple sides, from collecting and validating robotics data to training and evaluating models in real-world conditions. Better robotics isn't just about teaching models what to do. It's also about making sure they can do it at the right time. Read more below 👇
A Python pipe buffer was holding camera frames until 64 KB had piled up. That quietly added hundreds of milliseconds to our π0.5 rollouts. New from @castorhat: how we found it, and what changed on the robot once it was gone.
18
388
🗣️ The First 7 Days Can Decide Everything Getting someone to join your community is only half the job. The real challenge starts after they click "Join". I found @DlicomApp latest article interesting because it breaks down something many community owners overlook: the first seven days can determine whether a new member becomes an active part of the community or quietly disappears. A simple welcome message isn't enough. Give newcomers one small action to take, reply to their first post while they're still excited, ask questions instead of just saying hello, and create regular activities that give people a reason to return. One detail that stood out to me was Dlicom's own data: team replies ending with a question received responses 50% of the time, compared to just 26% for replies without one. Sometimes, a small change in how you communicate makes a real difference. I think community growth shouldn't be measured only by how many people join, but by how many actually stay, participate and build connections. The first week is your chance to turn a new member into a regular.
🤝 Paid partnership
Count who joined your group last Monday. Now count who's still talking today. That gap is decided in the first 7 days, and you usually find out on renewal day. What to do on each of them 👇
1
8
187
⏳ A New IPO Genie Milestone Is Just Around The Corner The current presale window is entering its final stretch, with 17 hours remaining and the $IPO token priced at $0.0002397. What I find interesting is how each phase creates a new checkpoint in the project's presale journey. This one is almost complete, and the next transition is getting closer. Less than a day left to explore the current stage of @IPOGenie. bit.ly/IPOGenie-borsch737
🤝 Paid partnership
15
363
🎙️ PrismaX Office Hours #11 Recap PrismaX Office Hours on October 30, hosted by @vivianrobotics, focused on how robotics data translates into real-world performance. One of the most interesting points was how operator behavior carries over into trained models. Small details like approaching an object, pausing or making corrections can influence how a robot performs. @PrismaXai is testing this through Pi 0.5, using its open-source YAMulator simulator and real hardware to measure actual results. A few key takeaways: • Every episode goes through Verified Quality, where task completion, movement smoothness and instruction following are evaluated • Successful attempts form the core of training datasets, while failed ones help identify where operators struggle • Beginners can start with teleoperation and gradually move into validation as their tier allows • The ecosystem offers several ways to contribute, including teleoperation, validation, the Regional Ambassador Program and custom data collection • The validator leaderboard rotates monthly, putting the focus on consistent work rather than one-time achievements One important clarification: Teleoperation Points and Validation Points will remain separate systems. The team is also working on an updated contributor system, with more details to be announced later. What matters here is the connection between contribution and real results. Collecting robotics data is just the first step. Making that data useful for training better-performing robots is the bigger goal. 🔗 Official Links: Website: prismax.ai Explore the data and request access today: prismax.ai/#plans Explore Verify Quality: app.prismax.ai X: @PrismaXai Discord: discord.gg/prismaxai
🤝 Paid partnership
15
333
📊 Better Robotics Start With Better Data Robots don’t just need more training data. They need data that reflects the conditions they’ll face in the real world. In its latest article on fine-tuning π0.5, @PrismaXai showed how something as simple as camera placement can affect performance. In their experiments, the robot could still pick up an object, while successfully placing it into a container became much harder when the camera position changed. That’s a good reminder that successful demonstrations are only part of the picture. Hardware setup, camera specifications, data collection, and evaluation all influence what a model can learn. This is where PrismaX’s broader work with robotics data becomes interesting. Through Robotics Data and the Verify Quality program, PrismaX is working on both access to robot demonstrations and the evaluation of their quality. Better Physical AI isn’t just about collecting more episodes. It’s about understanding the data behind every action. Read more below 👇
🤝 Paid partnership
Hard lighting cut our expert-only π0.5 checkpoint's success roughly in half. The same lights cost the full fine-tune nothing. New from @castorhat: the fine-tuning ablations behind our IROS demo.
1
13
370
🗣️ Your Community Might Be Worth More Than You Think One thing that stood out to me in Dlicom’s latest article is that you don’t need thousands of followers to start earning from a community. Sometimes, a few dozen people who genuinely care about what you share are enough. Think about the work community owners already do: answering questions, welcoming newcomers, keeping conversations alive and giving people a reason to return. That takes time and effort, yet many still do it for free. What I find interesting is how simple the math can be. Just 40 members paying $29 a month could generate $1,160 in monthly revenue before platform fees. You don’t always need a massive audience. You need people who see value in being part of what you’ve built. @DlicomApp is also exploring this idea through direct creator-to-creator and fan-to-creator tips. Members can support creators directly from their wallets, without making monetization feel complicated. Maybe the real question isn't whether your community is big enough to monetize, but whether you're giving people something worth supporting. Read the full article and do the math for your own community 👇
🤝 Paid partnership
You run a community every day. Someone else is getting paid for the same work. $10 billion on Patreon. $168M a month on Skool alone. The full breakdown of who keeps what is inside 👇 What would you charge for yours?
7
12
385