May821 ./

@fishfish5771 数据质量是关键

最近看了 @PrismaXai 在 @decasonic Web3 Investor Day 的分享，我最有感的一句话是： Physical AI 真正缺的，不是更好的硬件，而是更合适的数据。 现在机器人数据公司越来越多。 有人提供第一视角视频，有人使用动作捕捉，有人制作合成数据，也有人通过夹爪工具收集操作轨迹。

The First 100 Validators 最后冲刺了！7月31日即将截止，现在已经有 7万多人 参加，是 Galxe 上最火的活动之一。完成任务就能分 $2,500 USDC，还能攒 Prisma Points，冲进 Founding Validators！ 另外本周 Validator Watch Wednesday 也来了，社区大神们又分享了不少实战经验：完整打分流程、怎么看 Motion Data、标准为什么重要……新手看了也能快速上手。 现在边学边打分，冲刺最后一波最合适了！你每一次认真打分，都在帮 PrismaX 把数据质量往上提。赶紧动起来，一起冲进 The First 100！ @PrismaXai

简简单单又一餐

@base @axisrobotics @hack_vc @NomadCapital_io @PiCoreTeam @10kventure Axis on Base

@axisrobotics @hack_vc @NomadCapital_io @PiCoreTeam @10kventure based

We’re thrilled to announce a $12M Seed round, led by @hack_vc, with participation from @NomadCapital_io , @PiCoreTeam Ventures , @10kventure and top angel investors. Physical AI has a data problem. Models need more than static datasets—they need diverse data that evolves with them. Axis’s compounding Data Engine is here to fix this gap. Our end-to-end closed-loop workflow unites large-scale simulation, egocentric real-world capture, and human-in-the-loop post-training to unlock scalable production of structured, multi-diverse robotic data — the core missing piece for Physical AI. The capital will accelerate Axis’s mission to build a massively parallel, human-in-the-loop global data engine. We’re just getting started.

@bilixbili 这个图做的好好

The CLARITY bill has entered its final crucial stage. From a personal perspective, let's discuss the impact of the bill on Web3. This bill establishes a regulatory framework for digital assets in the United States and clarifies the jurisdiction of the SEC and CFTC. @get_optimum

@slws1234 哇，好妖娆

今天开车听广播说今年已过三分之二，突然感觉时间流逝飞快，看看日历还有十天就立秋了。 秋，象征着丰收，深藏着喜悦，@SeismicSys 社区的小伙伴们，让我们继续默默耕耘，静待花开，愿大家在即将到来的秋季收获满满！ @nanajiang8888 @xealistt

Optimum is really rewriting the MEV game. In the past, slow networks meant a lot of high-value MEV opportunities were simply lost. Research shows that around 10% of slots experience cross-slot value leakage, with cumulative losses reaching up to 1/6 of total MEV. Now with mump2p + RLNC, propagation is about 6x faster and latency variance is much lower. Validators can more confidently delay block proposals to chase higher bids — without compromising security. Speed is Money. Consistent Speed Captures More. Institutional capital and real-world finance are accelerating on-chain. Network-layer optimization is no longer optional. @get_optimum isn’t just making the network faster — it’s enabling the entire decentralized system to handle true financial-grade efficiency.

@Hz9r8nxlJason 真人？！我都想打人😂

63.58K 人已经加入 The First 100 Validators 🦾 目前，它已成为 Galxe 上最活跃的活动之一。热度和参与度，数据已经说明了一切。🔥 从远程操作真实机械臂，到参与数据验证，越来越多人开始关注 Physical AI 背后的数据生产与质量标准。 完成任务，即有机会瓜分 $2,500 USDC 奖励，并开启你的 PrismaX 体验。 活动将于 7 月 31 日结束： 👉 app.galxe.com/quest/fio5HZ84… X：x.com/PrismaXai Discord：discord.gg/gdNgVCDkr2

Physical AI 当前最大的挑战，不是有没有数据，而是有没有真正干净、标准、可复用、能支撑模型泛化的高质量数据。这句话精准点出了整个行业的核心痛点。 过去几年，机器人数据的数量确实在快速增长——开源数据集、商业数据提供商、遥操作平台不断涌现。但真正能有效训练 Foundation Model 的数据却极度稀缺。 市面上大量数据存在以下问题： 不干净：充满犹豫、挣扎、重复尝试、动作卡顿、视野缺失等问题。模型一旦学到这些坏习惯，就会在真实部署中复制错误。 不标准：不同硬件、不同采集方式、不同传感器配置的数据格式混乱，难以统一处理。 不可复用：数据高度绑定特定机器人或特定场景，换一个具身或环境就基本失效。 难以支撑泛化：大多数数据集中在少数简单任务和受控场景，模型在真实复杂环境（厨房、仓库、家庭）中表现急剧下降。 这就造成了一个矛盾现象： 数据量越来越大，但真正有用的数据却依然稀缺。 很多团队花大量预算买数据，最终发现模型效果提升有限，甚至出现性能退化。 因此，Physical AI 的竞争本质已经从谁能收集更多数据，转向了谁能持续生产、验证、积累真正高质量、标准化的训练级数据。 这也是 @PrismaXai 把重点放在数据标准、验证体系（Verify Quality）和数据飞轮上的原因——只有解决质量问题，才能真正让机器人从实验室走向真实世界。

模型再强，如果成本居高不下、高度依赖云端、难以真正落地，其价值也会大打折扣。 在顶级模型持续迭代、价格战愈演愈烈的背景下，单纯追逐最新最贵的模型，往往会陷入性能虽好，却用不起、控不住的困境。 Gradient 通过 Parallax 所坚持的路径——本地分布式、硬件中立、主权可控——恰好提供了一个清晰的替代方案： 不盲目追赶最新最贵的模型，而是高效利用现有与开源模型，在本地构建真正可控、可扩展的 AI 能力，把生产力牢牢掌握在自己手中。 @Gradient_HQ

@Darkwjq 你最近输出也很猛啊

感谢 baised.♾ 开发的超有趣的应用 Optimum 的可爱吉祥物直接从电脑屏幕里走出来 陪我过夏天啦~ 太萌了！ @baisedit @ada_pegasus @get_optimum

@Darkwjq 带我们起飞

@may60535 @get_optimum 小超人来了

In the era of institutional on-chain adoption, @get_optimum is becoming critical infrastructure. Traditional Wall Street is losing to its own limitations, while the New Wall Street is being born on-chain. As stablecoins, RWA, and institutional-grade DeFi accelerate their migration on-chain, blockchain is about to face the test of trillions of dollars in transaction volume. The traditional network layer will become the biggest bottleneck due to high latency and variance. Optimum provides the answer with RLNC + mump2p: 1）Approximately 6x faster block propagation with extremely low latency variance. 2) The more nodes and the wider their global distribution, the faster and more stable the network becomes. 3) Enables validators to safely delay block proposals and capture more MEV. 4) Builds a Latency Marketplace, making speed a fair and accessible resource. Optimum is not just another Layer 2. It is a project that reconstructs the foundational capabilities of blockchain from the network layer. It turns decentralization from a performance trade-off into a true competitive advantage. Speed is Money. Consistent Speed is Power. In the new era of infrastructure competition, Optimum is helping blockchain evolve from the experimental stage into a mature global financial system — providing reliable infrastructure for the massive on-chain migration of institutional capital.

@metanoetix 速度真快

Many retail ETH staking investors, like myself, likely share this feeling: after staking, they basically stop paying attention, assuming the APR (Average Per Transaction) is fixed and completely ignoring the impact of node latency. Surprisingly, latency can actually widen the gap in returns over time! Currently, based on historical slot time data and actual MEV (Mean Electricity) bidding tests by the Optimum team, we can see the logic behind RLNC's latency optimization increasing returns. This is easy for the average person to understand. The core key point that most of us overlook is that Ethereum's tiny proposal latency of 50-150ms directly determines your actual staking APR and MEV bonus cap. 1️⃣ A real increase in staking APR of 0.66%-1.97%: While the latency difference of tens of milliseconds may not change the base block reward out of thin air, it precisely captures every available slot. Over the long term, this allows retail investors to consistently achieve annualized returns far exceeding the market average. The cumulative effect over time will amplify this difference significantly. 2️⃣ Maximize MEV Bonus Returns: Many ordinary users are unaware that a large portion of validators' excess returns comes from MEV arbitrage bonuses. Ordinary nodes, due to latency redundancy, miss numerous high-quality bidding opportunities. Optimizing latency by 50-150ms significantly improves node block proposal response speed, allowing them to reliably capture all the previously lost low-priced MEV profits. Predictions suggest an additional 150-190 ETH per week. 3️⃣ 13%-16% Increase in Block Price Value Per Week: This is the most direct data. With the same staking amount and the same on-chain environment, simply optimizing latency parameters directly increases the overall block price value by over 10%. For ordinary users, this represents pure incremental returns with zero additional capital and zero high risk. In conclusion, none of the above are theoretical models from a laboratory; they are practical conclusions drawn by the Optimum team based on a full analysis of Ethereum's historical slot times and real MEV price tracking. These conclusions apply to large staking service providers, MEV search agencies, and ordinary retail stakers like ourselves! As ordinary stakers, we shouldn't just focus on the rise and fall of ETH price and miss out on these risk-free, compounding, and stable ways to increase core returns. When Optimum launches on the ETH mainnet, it will be an excellent choice for ordinary stakers like myself, something I've been eagerly anticipating. @get_optimum

@yangdayuya 确实是这样

在 @PrismaXai 体验几天后的真实感受 🦾 这几天我反复在 VLA Foundry 里验证机器人演示数据：打分动作质量、任务匹配度、画面清晰度、犹豫情况，以及整体价值。 最深刻的体会是：PrismaX 不是简单收集数据，而是建立了一个真正的闭环系统——operator 通过远程操作生成演示，我这样的 validator 提供人类判断层，高质量数据再反哺 VLA 模型。 这精准解决了 Physical AI 的核心瓶颈：不是数据越多越好，而是需要真正可靠、高质量的真实世界演示。机器人要理解物理、时机、恢复和精度，坏的演示会毁掉模型，验证过的才能真正落地。 a16z crypto 背书、Solana 生态、实时积分奖励真实贡献——这是目前 AI + DePIN 赛道里，我觉得最脚踏实地的基础设施项目之一。 体验几天后，我越来越相信：下一代智能机器人，就是靠这种“人依然在环中”的方式训练出来的。

@fishfish5771 高质量数据为更好的模型打基础

以前我一直以为，机器人只要拥有更强的 AI 模型，就会越来越聪明。 后来接触 @PrismaXai 后，我发现自己忽略了一件更重要的事情。 AI 不会自己分辨什么是「好的经验」。 机器人每天都会产生大量的数据。 但并不是所有数据，都值得拿来学习。

@Boji_king8ji 又有新项目了啊

最近关注了一个RWA项目——@SimpleChain_RWA 他们刚官宣完成了1500万美元种子轮融资，由家族办公室和机构投资者参与，正在打造一条专注 RWA 的Layer 1链，定位是机构级的RWA操作系统。 核心亮点在于他们不光做链上交易，更强调颗粒度数据和原生合规服务。他们指出，未来RWA要真正规模化，不能只把资产token化，还得解决资产背后的真实状态持续验证问题——比如设备是否在运行、现金流是否真实、所有权和合规情况有没有变化等，让链上代币始终锚定现实世界的经济事实。 简单说，就是在区块链+AI的时代，为机构提供可验证、可审计、全生命周期可追踪的资产基础设施，帮传统资产更好地上链，也让更多真实世界资产（设备、能源、供应链等）能被机构放心投资。 项目方向清晰，有兴趣的朋友可以去关注一下～ task.simplechain.com/?inviteCode=9v…

最好的学习方式，就是向已经在认真做的人学习。 每周三，@PrismaXai 的验证者们都会分享最实用的操作技巧和经验心得，帮助大家更快掌握数据验证的方法。 为什么 Verify Quality 这么重要？ 在物理AI 领域，数据质量直接决定模型的上限。验证者不是简单看视频打分，而是要判断一条演示是否干净、专业、值得被模型学习。通过 Pass/Fail 检查 + 质量评分（速度、平滑度、完成度、多样性等），确保进入训练的数据真正可靠。 这形成了正向飞轮： 操作者生产数据 → 验证者守护质量 → 更好数据集 → 更强模型 → 更好的操作体验。 现在正是好时机： 本周是最后冲刺 The First 100 Founding Validators 的机会！ 成功入选后将获得更多审核权限和认可，成为 PrismaX 数据基础设施的早期建设者。 你的每一次认真打分，都在帮助 PrismaX 打造真正高质量的机器人训练数据，推动物理AI 从能动走向能用、好用。 快来参与，一起变强！

平静

@bilixbili 网络升级很有必要

The MEV-SBC workshop, hosted by Flashbots, is a major annual event in the blockchain field. This year, it was held at Stanford University on July 30th as an adjunct workshop to the Science of Blockchain Conference. @get_optimum