注册并分享邀请链接,可获得视频播放与邀请奖励。

与「spa」相关的搜索结果

spa 贴吧
一个关键词就是一个贴吧,路径全站唯一。
创建贴吧
用户
未找到
包含 spa 的内容
SpaceXSI Principal Engineer @poteto 在 AI Engineer Code Summit 2026 @aiDotEngineer 分享「All Code Is Slop (Except Mine)」 她真正想论证的是软件工程师的工作变化:在大幅提高产出的同时,维持甚至提高代码质量。 有人把这条路叫做软件工厂,她更愿意把它说成信任!
显示更多
行情不好不行学学英语吧,省的亏钱,学好了还能找份好工作继续补仓。 外网最火的学外语方法叫可理解输入,带火它的是教西班牙语的 Dreaming Spanish,他们自己的路线图写着,听满 50 个小时,认识 300 个常用词就学会了😂
显示更多
看到 SpaceX 星舰被机械臂接住的视频,还是感叹 SpaceX 的强大! 想起之前某个公司老板,很崇拜「臭鼬工厂 Skunk Works」,对黑鸟、U2、F117 等等如数家珍,还把各种臭鼬工厂的飞机和口号挂满公司,经常挂在嘴边的两句话: 我们要以臭鼬工厂的做事标准来要求自己! 臭鼬工厂做的才是开创性的,SpaceX 不过是大力出奇迹,就是做足动力,把火箭推上去。
显示更多
SpaceX catches Starship for the first time , this is so peak engineering bruh
Not every serious trader wants to spend the entire day staring at six monitors. That's one reason I find @Hypesium worth paying attention to. For years, sophisticated trading has been associated with complicated desktop terminals, multiple browser tabs, and dashboards filled with information. But Web3 participation doesn't always happen behind a desk. Sometimes you're commuting, traveling, working, or simply away from your computer when something important happens in the market. That creates an interesting design challenge. How much of a trading workflow can become mobile-friendly without sacrificing the information needed to make responsible decisions? Hypesium approaches this through a Telegram-based trading interface focused on the Elysium ecosystem, connecting users with trading activity across multiple launchpads. I think the significance of that approach goes beyond convenience. Telegram is already familiar to many crypto participants. Bringing trading interactions into that environment can reduce the need to constantly switch between applications. However, mobile accessibility introduces its own limitations. A smaller screen provides less room for detailed research. Notifications can encourage impulsive reactions. And the convenience of executing a transaction quickly doesn't necessarily mean the decision has been properly evaluated. That's why I believe mobile-first trading should be designed around clarity rather than simply maximizing speed. Users need enough context to understand what they're doing, even when they aren't sitting in front of a full trading setup. For me, @Hypesium represents an interesting example of how specialized trading infrastructure can fit into the everyday habits of crypto users. The opportunity isn't to turn every spare minute into another trading session. It's to make market access more flexible while preserving thoughtful decision-making. Because accessibility should expand a trader's options, not increase the pressure to act. Do you think Telegram-based trading could eventually become more practical than traditional desktop terminals for everyday Web3 users?
显示更多
0
20
25
20
转发到社区
@onlystreetsfun: launchpad turning real-life moments into coins on Solana. @rarelistio: creator campaigns that pay in USDT. The moment is the asset. OnlyStreets has announced no token name, no supply, no roadmap, and no launch date. The launchpad is still coming soon on Solana. What it has announced is the asset itself. A real-life moment, caught with a fresh photo taken directly in the app. No gallery uploads are allowed. Most projects sell tokenomics first and a product later. OnlyStreets sells a moment first, and the coin follows. That is why the coin cannot be copied. A moment only happened once, on one street, at one second. The Rarelist campaign runs November 11 to 17. Rank follows RARS, not post count. Only qualified activity earns RARS, and spam does not count. Rarelist's last batch paid $5,500 across three campaigns. Its Season 1 holds $50,000. Own the moment. The coin will follow.
显示更多
0
34
28
16
转发到社区
Diffusion language models can rewrite their own drafts — but rewriting the wrong span just reproduces the same mistake. Remasking lets these models revisit generated text, yet they have never had a good sense of when a revision is worth it. DiVR adds a verifier that predicts whether a partial sequence will still reach the right answer, then decides when to regenerate a block and which alternative to keep. On LLaDA-8B-Base across five reasoning benchmarks, average accuracy rose from 35.3% to 43.4% at 1.9–3.5× the cost of greedy decoding — while search baselines needed 3.8–7.4× the compute for worse results. The practical payoff: a real accuracy lift without touching or retraining the generator, and the verifier even transfers to other generators like Dream. The catch is still multiplied compute, so latency-sensitive deployments need to do the math first. Would you trade 2–3× compute for an 8-point accuracy gain on your own reasoning tasks? #DiffusionLM# #LLMReasoning#
显示更多
SpaceX is a honestly an artificial superintelligence
0
16
206
11
转发到社区
Cursor 公布安全事件调查结果 1. 发布情况 北京时间 11 月 14 日 09:13,Cursor 官方账号发帖,公布此前一起安全事件的调查经过和补救措施。[1] 博客文章的落款日期为美国时间 11 月 13 日。Cursor 曾在 10 月 29 日通报,正在调查一起安全事件,事件涉及 Cursor VS Code 扩展市场中出现的一个恶意、未经授权的 Anthropic Claude Code 扩展副本。[2] 2. 攻击经过 2.1 从服务商处泄露的密钥 根据博客,这次攻击始于第三方训练和推理服务商 Fireworks。Cursor 曾向 Fireworks 签发一把访问密钥,这把密钥从 Fireworks 一方泄露。攻击者借此复制了一个用于向 Fireworks 共享数据的只读存储桶。[2] 2.2 泄露的数据及后续访问 存储桶中有数量有限的编程智能体运行轨迹,主要来自 2025 年,均为未开启隐私模式、并允许用于模型训练的会话。其中少量轨迹含有未脱敏的凭据,API 凭据因此泄露的客户不到 0.01%。攻击者用这些凭据进入了 Cursor 使用的部分 SaaS 工具,进而访问了以下数据:[2] 部分已归档的非生产源代码仓库,其中一些包含商务联系信息。 有限的客户支持数据和邮件往来。 产品使用指标和联系方式。 部分编程智能体对话的摘要。 用户发给支持团队的部分凭据。 2.3 恶意扩展 10 月 29 日,攻击者利用窃取的凭据,把一个经过恶意篡改的 Claude Code 扩展直接放进了 Cursor VS Code 扩展市场的缓存。这个副本不是 Anthropic 发布的,此事也不属于 Anthropic 产品的漏洞,该恶意软件从未出现在上游来源 Open VSX 中。10 月 29 日请求过特定缓存版本的用户中,最多约 5000 人可能收到了恶意构建,这个版本存在约 11 小时后被移除。涉及的版本是 Anthropic.claude-code 2.1.312 和 2.1.313。Cursor 提示,仅仅卸载扩展可能不足以消除影响,受影响的用户应按安全通告中的步骤处理。[2] 3. 影响范围 3.1 受影响的用户 Cursor 表示,绝大多数客户没有受到影响,公司已经逐一直接联系了受影响的客户。在未开启隐私模式的用户中,编程智能体轨迹遭到泄露的不到 0.05%,没有收到通知就说明自己的数据不在其中。[2] 3.2 未受影响的部分 以下部分均未受到影响:[2] 开启隐私模式的用户的提示词和代码。 目前发送给 Cursor 的数据和实时流量。 生产系统和基础设施。 Cursor 应用及其官方更新。发布流程使用一套攻击者没有接触过的独立代码托管系统,签名密钥也没有泄露。 面向团队的云端智能体、Grok Bot 及其他 SpaceX SI 产品。 扩展市场中的其他扩展,以及从 Cursor 插件市场下载的插件。 客户的代码仓库和正在进行中的推理请求。 Cursor 称,最后一次观测到攻击活动是在 11 月 1 日,当时已被拦截,因此判断事件已经得到控制。[2] 4. 补救措施 Cursor 已经采取以下措施:[2] 把全部推理流量迁出 Fireworks,停用相关密钥,并切断 Fireworks 对所有共享数据集的访问。 按风险高低依次轮换所有已发现的泄露凭据,并预防性地轮换内部系统中所有在用的凭据。 核验并重建 VS Code 扩展市场。 在各服务商那里收紧网络访问限制,让泄露的凭据无法在预定位置以外使用。 在所有代码仓库和其他内部数据源中增加密钥扫描。 预防性地重新部署全部生产基础设施。 聘请外部取证专家协助调查。 如有安全问题,可以发邮件到 security@cursor.com。[2] 注释与来源 [1] Cursor 原帖
[2] Cursor 博客《An update on our recent security incident》
显示更多