cy

1.3K posts

cy banner
cy

cy

@Chany2134

Присоединился Şubat 2013
2K Подписки83 Подписчики
cy ретвитнул
知识玩家
知识玩家@BrainQuestor·
🚨 突发:谷歌Gemini现在可以像华尔街分析师一样分析任何股票(免费)。 以下是9个疯狂的Gemini提示词,可替代4000美元/月的彭博终端: 保存以备后用🔖
知识玩家 tweet media
中文
116
198
773
117.8K
cy ретвитнул
郭宇 guoyu.eth
郭宇 guoyu.eth@turingou·
今天正式发布了我的第 13 个 vibe 产品 tuwa.ai 这款产品比较特殊,它是一个电话服务。准确来说,tuwa 是一个 AI 电话网络,连接着世界上超过 100 种不同语言的人们和互联网上的 agent。 任何人都可以不下载 tuwa 而使用它,你只需要拨打免费的转接热线电话 +1 888 886 2968,告诉它你需要打给哪个号码,tuwa 就会帮助你拨打对应的号码,你说自己的母语,对方听到的却是TA的母语,反过来也是一样。兼容任何电话,对方不需要安装应用。固话、手机,世界上任何一个角落,都可以。 tuwa 支持 100 多种语言的实时翻译,你甚至可以在打电话时随便切换语言和对方对话。除此之外,tuwa 还支持语音克隆,每一通电话,都会让你的 AI 语音听起来更像你。 当然,我也为它设计了方便的 web app,如果你想,可以不通过转接电话而使用 web app 拨打,并设置自己喜欢的声音,使用外呼 agent 拨打电话,连接自己的 agent(例如 openclaw 或者 codex/claude code)并让他们自由的呼入与呼出。 外呼电话 agent 是我最喜欢用的 tuwa 功能,只需要交待清楚事情,比如完成餐厅预订,它就会在你希望的时间主动拨打对方的电话,说明来意,达到目的,并记录和翻译所有对话内容。 tuwa 的使用和收费都很简单,每月免费额度,固定套餐,按需付费。 这个产品的命名灵感来自于日语的「通話 tsuwa」最初,我只是想设计一个能帮我预订餐厅的电话服务,但后来,我在 vibe 的过程中慢慢意识到,世界上仍然有很多人无法体验 AI 带来的变化与便利,而电话,是连接他们最简单与自然的方式。我希望 tuwa 能帮助外语普及率低,偏远地区和第三世界国家的人们体会到这一点。
郭宇 guoyu.eth tweet media郭宇 guoyu.eth tweet media
中文
383
476
4K
891.2K
cy ретвитнул
VirtualBacon
VirtualBacon@virtualbacon·
I’m holding $NEAR and $TAO as my core AI altcoins. Both are strong L1 infra plays with 2x+ gains from 2023 bear lows. Holders are happy, price action is solid, and fundamentals keep improving. $TAO’s Dynamic TAO upgrade could be its biggest move yet.
English
84
71
808
102.5K
cy ретвитнул
沐阳
沐阳@yyyole·
简单说说,Google这次的Gemma4。
沐阳@yyyole

google 正式发布了 Gemma 4!! 这是他们迄今为止最强的开源模型家族! 官方的说法是:脱胎于 Gemini 3 的同一套研究成果! 也就是说,你用的是开源版, 底层技术跟他们的旗舰商业模型是同源的! 不愧是大善人! 先说最重要的变化:许可证。 之前的 Gemma 系列用的是 Google 自家的许可协议。 条款繁琐,商用存在灰色地带, 这次 Gemma 4 全系列换成了 Apache 2.0! 商用、修改、分发全部放开!! 更关键的是,微调之后的衍生模型也可以直接商业部署。 不用再研究条款。 对独立开发者和企业来说, 这个变化的意义可能比跑分更实际!! 再说说它的四个版本,和两个定位 Gemma 4 提供四个版本,按部署场景分成两档: 第一个是工作站档。 有 31B Dense 和 26B MoE 两个版本,支持文字和图像输入,上下文窗口高达 256K tokens,一次性可以处理超长文档或整个代码仓库。 第二个是边缘端档。 是 E2B 和 E4B 两个小模型,这里的"E"代表"有效参数",E2B 的有效参数是 23 亿,但实际总参数是 51 亿,这是一种新的架构压缩方式。 再说说跑分。 31B 模型在 Arena AI 开源文本排行榜上拿到了 1452 分,排名全球第三,26B MoE 达到 1441 分,排名第六。 这些数字在不久之前还是闭源旗舰模型的水平 端侧小模型E4B 在 AIME 2026 上拿到 42.5%,LiveCodeBench 达到 52%,这是一个能跑在 T4 显卡上的模型,两个小模型在大多数基准上都超过了上一代 Gemma 3 的 27B! 最有意思的部分是端侧能力,E2B 和 E4B 是 Google 与 Pixel 团队还有高通、联发科联合开发的专为手机、笔记本、树莓派、Jetson Nano 等设备优化,这两个小模型原生支持文字、图像、视频和音频输入,具备语音理解能力,上下文窗口为 128K tokens! 也就是说,现在有了一个可以在手机上离线运行,能看图、听话、还能写代码的本地 AI 智能体,不需要联网,不需要调用任何 API,开箱即用!! 另外,Gemma 4 的函数调用能力是从训练阶段就内置的,而不是靠 prompt 工程"诱导"出来的,专门针对多轮多工具的 Agent 流程进行了优化,全系列原生支持结构化 JSON 输出,直接对接自动化工作流。 31B 模型不做量化的情况下,可以装进单张 80GB 的 H100 显卡,量化版本则可以在消费级 GPU 上运行 目前模型已经可以在 Google AI Studio、Hugging Face、Kaggle、Ollama 等平台体验! Gemma 系列自发布以来累计下载量已超过 4 亿次,社区创建了超过 10 万个模型变体,Google 显然在用开源生态绑住开发者,但现实是Arena AI 开源排行榜的前列正在被中国模型密集占据! Qwen、GLM、Kimi 都在这个参数段打得很激烈就在 Google 主动开放许可证的同时,部分中国 AI 实验室的最新模型反而在收紧开放力度,Gemma 4 的 31B 现在排第三,这个位置能守住多久,还不好说!

中文
10
20
121
13.9K
cy ретвитнул
Bitcoin Meraklısı
Bitcoin Meraklısı@Bitcoinmeraklsi·
#TAO Güncel görünüm. Dostlar $TAO düzeltmesini tamamlıyor gibi. Kısa vade düşen trendi kırabilirse yeni bir yükseliş dalgası başlayacaktır. Ve yükselişin başlaması ile grafikte belirttiğim üst seviyeler tekrardan sıralı hedefleri olacaktır. Yakın takibinizde olsun.
Bitcoin Meraklısı tweet media
Bitcoin Meraklısı@Bitcoinmeraklsi

#TAO Retest geldi. Destek çalıştı. Şimdi kırılım bekleniyor. Dostlar $TAO, yukarı kırdığı Ters OBO formasyonu direncine güzel bir retest yaptı. Bu seviye destek olarak çalıştı ve ilk tepki alındı. Şu an son zirveden gelen kısa vadeli düşüş trendi test ediliyor. 📌 Bu trend kırılımı ile birlikte yeni bir yükseliş dalgası başlayabilir. 📌 Olası yükseliş durumunda grafikte belirttiğim seviyeler tekrar sıralı hedefler olarak takip edilecektir. Tabi #BTC tarafı da izin verirse. Yakın takibinizde olsun.

Türkçe
6
8
148
25.9K
cy ретвитнул
Cointelegraph
Cointelegraph@Cointelegraph·
🚨 UPDATE: Grayscale files S-1 amendment for its Bittensor ($TAO) trust.
Cointelegraph tweet mediaCointelegraph tweet media
English
82
273
1.4K
199.9K
cy ретвитнул
Noah | NCBTRADES
Noah | NCBTRADES@ncbtrades·
Only on Bittensor $TAO can you find setups like this during a bear market... I'll be screaming from the mountain tops about @vidaio_ until all ya'll hear me. SN85 will be of MAJOR importance, especially when you factor in how much video generated content there will be in the coming years. I'm stacking Vidaio alpha tokens as much as I can before the herd comes. There is literally no point in trading memes when you have real utility AI models that you can back on the Bittensor ecosystem. Only 21 million Vidaio alpha tokens. Already $1.1M of alpha tokens burned... I don't think you're ready for the pump.
Noah | NCBTRADES tweet media
English
4
11
86
5.9K
cy ретвитнул
Google
Google@Google·
Gemma 4 is our most capable open model family yet: 🔵 Four versatile sizes 🔵 Up to 256K context window 🔵 Native function-calling for autonomous agents 🔵 Offline, high-quality code generation 🔵 Native multimodal support 🔵 Trained on 140+ languages 🔵 Commercially permissive Apache 2.0 license
Google tweet media
English
32
136
1.8K
286.9K
Elon Musk
Elon Musk@elonmusk·
Grok automatically translating and recommending 𝕏 posts from other languages is starting to work
English
12K
20.3K
277K
78.6M
Konnex
Konnex@konnex_world·
"The era of isolated machines is over." Meet our CEO, @JOllwerther . Business leader with 15+ years in robotics, drone-tech, and media. Jon led commercial teams at Measure (Drone as a Service) and Aerobo/Vermeer. Now, he’s creating TCP/IP for Motion. Get ready to experience Konnex.
English
45.9K
2.4K
21.8K
208.4K
cy ретвитнул
John P. Acquaviva
John P. Acquaviva@JPAFS·
ROBOTS HUMANOIDES DE UNIX AI EN RESIDENCIAS REALES El robot pertenece a la compañía UniX AI, con sede en Suzhou, China. Esta firma de robótica ya ha comenzado la producción en masa y la entrega comercial de sus modelos humanoides (como la serie Wanda). Utilizan inteligencia artificial y aprendizaje por imitación para operar electrodomésticos, organizar espacios y asistir en tareas domésticas en entornos reales.
Español
1
12
27
2.7K
cy ретвитнул
RoboForce
RoboForce@roboforce_ai·
RoboForce just secured a $52M oversubscribed round, bringing total funding to $67M. Led by YZi Labs, with participation from Jerry Yang joining existing investors including Myron Scholes, Gary Rieschel, and Carnegie Mellon University. We’re building Physical AI to power the era of Robo-Labor.
RoboForce tweet media
English
211
76
693
102.9K
Konnex
Konnex@konnex_world·
Konnex Autonomous Units Mint opens March 16 — exclusively on @opensea The first robotic identities entering the Konnex network. Built to operate. Designed to evolve. Each unit is a gateway into the Konnex ecosystem. opensea.io/collection/kon…
English
107.9K
31.1K
59.2K
418.3K
Peter
Peter@Peter_rock07·
😱Another Pontienal testnet @fhenix 🪂 Raised $22m Cost $0 👀 > Fhenix is building simple privacy tools for blockchains using a special tech called Fully Homomorphic Encryption (FHE). - go to: alchemy.com/faucets/ethere - claim test tokens > First Step Dapps testing - go to test.redact.money - connect wallet and switch to ETH Sepolia - enter any amount of ETH - Next, click Decrypt and enter any amount of ETH - after decrypting, click & claim tokens > Second dapp - Go to: shieldedmode.fhenix.io - Click on "Add confidentiality, - deploy contract and click deploy". - Click to view tokens or go to "user" - Click on generate permit - done > After that - Click on the shield token. - click mint, enter amount & mint - Go to shield, enter the amount and shield. You unshield & claim as well - Next, go to "USE THIS TOKEN" and transfer a token to any address. Transfer in both "Private and Public" options Done ✅ Repeat these processes on sepolia, Base Testnet etc. Bookmark & RT
Peter tweet media
English
29
35
146
12.7K
齐博千
齐博千@Web3Dc888·
终于面基了 全真教的@Cryptobtcmiko miko 老师,成功加入全真教内部! 想加入全真教的小伙伴可以找我我拉你们😁感谢 @ZhanweiC 老师创建 全真教包含(AI ,加密货币,美股)以及推特时事热点分析🧐!
中文
43
0
41
5.9K
cy ретвитнул
Tâm (❖,❖) 🟨
Tâm (❖,❖) 🟨@Buddha_TTT·
5 MINUTES TO UNDERSTAND MIDEN 1️⃣ What is Miden ? @0xMiden is an infrastructure blockchain focused on - Privacy - Scalability - Client-side/edge execution - Zero-Knowledge (ZK) proofs The core difference of Miden is that instead of executing transactions on the network (sequencer/validator), Miden allows users and apps to execute and generate proof directly on their devices (edge), the network only verifies proof on-chain. => This is a major change in the Trust model Cost model Privacy assumptions Miden isn't a typical ZK L2 . It's built around client-side execution and ZK proofs 2️⃣ Polygon Spin-off Miden was incubated within Polygon Labs (AggLayer Breakout Program) and officially spun off as an independent project in 2025. The spin-off allows Miden to operate with its own roadmap and strategy, focusing deeply on privacy, ZKVM, and edge/client-side execution models. The project is clearly positioned for serious use cases, including institutional and infrastructure-level applications requiring confidentiality. 3️⃣ Funding | How much has Miden raised? $25 million USD (Seed round – April 2025) Lead investors: a16z crypto, 1kx, Hack VC Finality Capital, Symbolic Capital, P2 Ventures, Delta Fund, MH Ventures Angels: Rune Christensen (MakerDAO), Avery Ching (Aptos), Sreeram Kannan (EigenLayer) => These are tier-1 crypto VCs, showing that Miden is not a small project. 4️⃣ What problem does Miden solve ? Most blockchains today + Execution occurs on the network + Sequencer/validators see transaction data + Privacy is very limited + Scaling depends on network throughput Miden takes a different approach. Miden moves execution to the edge (user device) + Users execute transactions locally + Users generate ZK proof locally Only sends : Proof, Commitments to the chain Network only: ✅ Verify correctness ❌ Does not see original data => Result: Better privacy, better scaling, reduced reliance on centralized sequencer, changed trust model (user controls execution) 5️⃣ What is Miden's privacy technology? ❌ Not FHE ❌ Not MPC ✅ ZK Proofs + Client-side execution Privacy comes from + Off-chain state + ZK proofs + On-chain verification + Commitments instead of plaintext state This architecture is close to ZK-native chains but emphasizes edge execution 6⃣ Founding Team Miden is led by builders with very deep backgrounds in ZK and blockchain infrastructure @bobbinth Threadbare : Founding Team Co-Founder & Chief Architect of Miden VM and Edge Execution Model Former engineer at Meta and Ethereum Foundation, specializing in zkVM, STARKs, and cryptographic systems @azeemk Khan : Co-Founder & COO Responsible for strategy, operations, and ecosystem Has a strong background in Web3 and has participated in crypto initiatives at both institutional and organizational levels @schmiddominik1 : Co-Founder & Head of Product Responsible for product development and transforming research-grade ZK infrastructure into usable tools for developers This isn't a marketing-first team. This is a core infrastructure development team with a long-term vision for privacy, scalability, and execution 7️⃣ Partners & ecosystem Miden is incorporated by Polygon Labs. Integrates AggLayer (bridge, settlement) and is considered an “AggLayer-native chain” Miden is also building + Developer tooling + zkVM (Miden VM) + Playground + testnet for builders 8️⃣ Roadmap & Mainnet Testnet + playground : live Dev infra + tooling : nearing completion Mainnet target : v0.15 (late March 2026) with possible mainnet alpha extending into early-mid Q2 2026 depending on protocol readiness 9️⃣ Personal Assessment ❌ UX is still laggy ❌ Testnet is not yet stable But ✅ Strong funding ✅ Infra-level project ✅ Distinctive real architecture ✅ Very few people working in depth 🔟 How to Contribute Early to Miden as a Builder ? If you see potential in Miden and want to contribute If your thesis on Privacy + ZK + edge execution is something you believe in for the long term, you can join very early in the following ways: 🧪 Experience the testnet Complete playground quests, try out the flows, and understand how Miden works from a practical perspective 🤝 Join the builder community Join Telegram, ask questions, support new members, report bugs, share workarounds — this is where the team and builders pay the most attention. 🎥 Create quality content Tutorials, architecture breakdowns, share testnet experiences, analyze privacy and execution models. At this stage, in-depth content leads to more transactions Strong infrastructure projects are usually quiet in the early stages. But those who join early, understand deeply, and contribute genuinely are often the best positioned when the ecosystem begins to expand #Privacy #gmiden
Tâm (❖,❖) 🟨 tweet media
Tâm (❖,❖) 🟨@Buddha_TTT

Miden 💢 Testnet | Part 3 Spent the morning deep inside @0xMiden and this is where things stop feeling like a testnet and start feeling like real infrastructure. I just dropped a full walkthrough for 4 of the most important Miden tasks 🧩 5. Interact with a Counter Contract 🛠️ 6. Deploy a Counter Contract ⏱️ 7. Timelock P2ID notes 🌐 8. Network-level transactions This is where you touch smart contracts, notes, and how execution actually flows across the network Not glossy. Not smooth. But real. Task 4 is still bugged on my side which honestly tells you how early this still is If you’re serious about understanding how Privacy + ZK infra works under the hood, this isn’t optional Builders don’t wait for perfect UX. They build through it

English
3
1
7
475
cy ретвитнул
TIGER
TIGER@tiger_web3·
卧槽…GitHub榜首项目有点逆天了,叫做 WiFi-DensePose 开源项目。 号称只通过 WiFi 信号,就能“还原”你在家里的动作轨迹——不需要摄像头、不用额外传感器,甚至不用特殊硬件。 我日,门外连wifi就可以穿墙透视??? github.com/ruvnet/wifi-de…
中文
136
545
3.3K
948.3K
cy
cy@Chany2134·
@yujuhao @MahojinAI 推特连接粘贴过来,怎么总是失败呀,你也是这样的吗,能出个视频讲讲吗
中文
0
0
0
542
我只是个撸毛的
我只是个撸毛的@yujuhao·
A16z CSX 领投的500万美元的AI项目 @MahojinAI 第0季空投积分计划,没有推荐机制,能做的不多啊 1⃣使用钱包登录,绑定X 2⃣每日签到得50积分 3⃣每日分享X得100积分 4⃣花钱订阅得积分 ✅Story生态,可多号参与啊,每天就签到一下,非常简单! 参与链接:app.mahojin.ai/maho-point
我只是个撸毛的 tweet media
我只是个撸毛的 tweet media
中文
14
21
52
9.3K