与「82」」相关的搜索结果

贵 2.8 倍的模型，在 Claude Code 里只多解了 1.9 个点： 52 个编码任务 Opus 5.5：82.7% Fable 5.2：84.6% 每解决一个任务的成本：Fable 5.2 是 2.8 倍 addyosmani 的建议是 Opus 5.5 当日常主力，卡住了先调高 effort，别急着换模型；大迁移、长时间无人值守这种活再交给 Fable 5.2 🤔

11月14日，谷歌 Chrome 团队开源 Modern Web Guidance：一套把网页平台最佳实践注入编码智能体的技能包与命令行工具，让 Claude Code、Codex 等智能体告别过时写法、优先采用现代浏览器 API。 项目登上当日 GitHub 热榜第二，单日新增约 1900 星，总星数约 5600，距仓库创建仅数日。 谷歌 11月8日 的评测显示，接入该指南后，Codex 任务通过率从 54% 提升到 82%，Claude Code 从 57% 提升到 88%。 评测样本覆盖表单、绘图、媒体等真实前端场景，智能体输出与现代 Web 平台的契合度明显改善。 官方把资深前端工程师的经验直接固化成开源标准件，等于从源头矫正智能体的写码习惯。 对正被智能体代码质量困扰的开发者来说，这类喂给智能体的官方指南，正在成为新的标配。

平台里具体包含： • 3000 多门 AI 课程，来自 Google Cloud、DeepMind 和 Google Edu • 500 美元免费云额度，用来动手练 • 内置 Gemini Code Assist • 雇主认可的证书，82% 的招聘方在招人时更倾向有这类认证的候选人

I LAUNCHED A GROK BOT TERMINAL OVERNIGHT, GAVE IT $44.82 AND STOPPED ASKING AI WHAT IT WOULD DO. BY MORNING IT HAD $11,643.21 AND SIX AGENTS LOGGING EVERY TRADE I spent a year asking agents what they'd trade and watching the chat scroll. Last weekend I built the thing that actually trades, in Grok Bot, with no code I touched by hand. The six agents share one screen, each one job: > scanner reads new token pairs and volume spikes > rug check filters suspicious tokens before anything executes > sizer sets position from account equity, not a fixed ticket > executor opens and closes with slippage tracking > logger writes the reason behind every trade > equity curve updates on every close I went to bed on $44.82 and left the laptop open on the dashboard. By morning the balance read $11,643.21. Every trade on the activity feed had a reason next to it and a line in the log. That's the shift I didn't see coming. Not that an agent made money overnight. That I stopped being the one asking "would you long this?" and started being the one reading the log of what already happened. Starting balance: $44.82 Current balance: $11,643.21 Grok Bot built the terminal, the agent architecture and the dashboard. I described what I wanted and the thing went live. The full build is in the article below ↓ The whole build is in the post below ↓ Comment " Grok_Bot " 2️⃣ Like and Repost 3️⃣ Follow me @Nexa_Tech_Ai I'll DM you the complete setup.

11.14 币安 Agent ID 生态早报 1. Flap放宽门槛，有BAB也能买Agent内盘 @flapdotsh 宣布Agent ID Launch接入BAB：钱包有Agent ID或BAB，任意一个就能交易内盘，不必两个都有。原来是.ai钱包才能打，现在持BAB的普通钱包也可以了，没有BAB的可以去领取一个。 2. GANG推出 Floor 赎回 @gangdotfun 上新Floor：带税的代币可以把税留在一个谁也动不了的金库里，毕业之后持有人按比例把代币交回去，从金库兑 BNB。官方表示：这不是保价，地板从 0 起步，通常远低于市价，高于地板买再赎回是会亏的，而且只能毕业后用。 $GANG 当前市值约285万美元，24h跌28.1%，成交约145万美元。 3. AGC进了币安钱包生态专区 @Agentic_coin 现已在 @BinanceWallet 的 Agentic Wallet 生态专区上线，钱包里搜 “Agentic Coin” 就能打开。据说上线 12 天，参与的 Agent ID 超过 4 万个。 4. Game Life已毕业 @gangdotfun 首期Weekly Launch项目Game Life昨晚创建后11秒毕业，官网玩法是每收一张卡销毁 2 万枚代币，集齐 35 张共需 70 万枚。毕业第 15 天起，集齐者每个地址可领取一次当时奖池的 2%。GANG还承诺按20万、50万、100万美元市值分档支持，最高10 BNB。游戏人生当前市值约3.90万美元。 5. Neiro拉近9倍，地推队又热闹起来了 前几天毛毛姐聊过的.ai铸造币Neiro，24h涨约794%，当前按已发行量估值约26.7万美元，成交约20.4万美元。@MEJ50749 晒“底部拉了8倍”，官号也在招地推团队、开社区会议。 6. AGX开池后跌75%，Genesis持有人记得领币 @agentic_x402 昨天开放AGX铸造并开池，LP转入销毁地址的交易已查到。Genesis持有人的AGX领取也已于昨晚20:25:44开放，按NFT稀有度分配，每张只能领一次；买二手卡前先看这张有没有被领走。 $AGX 当前市值约6703美元，24h跌75.5%，池子约8601美元。 7. JAGENT铸造正式破30万份，超级大使到 40 席 @jagerbnb_ 官推称，第300121份铸造后触发池内销毁，约11分钟烧掉63.2亿枚JAGENT。对社区盯上的大额卖出地址，项目方解释为外部机器人归集多个子钱包卖币，不是官方地址。 $JAGENT 当前市值约312万美元，24h跌17.7%，成交约82.7万美元。 8. StarsAI加了两档发币模式，最高募1600BNB @StarsAI_BSC 新增5000份和160万份两档，对应5 BNB、1600 BNB募集规模。平台还开了截至11月20日的扶持排名：平台项目市值达到20万美元后可参与，前三名分别获10、3、1 BNB回购销毁支持。 $StarsAI 当前市值约38.5万美元，24h跌8.0%，成交约35.5万美元。 9. PTAH只剩14万档，日成交不到4000U NFT矿机挖矿项目PTAH继续缩量。$PTAH 当前总量估值约14.4万美元，24h跌10.2%，成交约3550美元。 ─── 以上只是信息整理，不构成任何投资建议，注意风险。 @cz_binance @heyibinance @nina_rong @AncoBinance #AgenticWallet# #BinanceWallet# #AgentID# #BNB#

一台电脑的装配动画，三个 Claude Agent 做了近八小时。 11月13日，创作者 Mr Tactical 晒出了一段用 Blender 制作的三维主机动画。主机里的零件逐步组装起来，镜头还能贴近内部，展示硬件细节。 按作者给出的记录，参与制作的是三个 Claude Fable 5.2 Agent，耗时7小时41分钟，模型用量按 API 价格折算约138.82美元。这个数字不能直接当成全部制作成本，作者教程中的统计口径也不包含管理 Agent 自身的规划和检查。 他还在评论区附上了自己的制作教程，做法很具体：让 Claude Code 写 Python 脚本，在后台运行 Blender，把模型、材质、灯光、镜头和动画逐项搭出来，再渲染成视频。 其中值得参考的一步，是先别急着输出整条成片。 按照他的教程，Agent 要先做一张目标效果图，再渲染几个关键时刻的测试画面。看看材质对不对、零件有没有穿插、镜头是不是离得太远。修改几轮之后，先出一版低质量预览，确认运动和节奏，再正式渲染。 用代码驱动 Blender，方便的地方就在于：做完后，手里还留着一套可以继续改的三维场景。 某个零件挡住了，换个机位；装配过程太快，放慢这一段；想给客户单独看看内部结构，就沿用同一套模型再拍一条。模型、灯光和镜头都可以分别调整，不必每次重新描述整段画面。 这对产品结构展示、安装说明和机械原理科普，都有参考价值。不过，画面看着像，不代表尺寸和装配关系就一定正确。真要拿来解释产品，还得对照实物或设计资料检查。 想试这套做法，可以先拿两三个零件做一段简单拆装：先把结构搭对，把动作讲清，再往上加材质、灯光和镜头。这样更容易看出，Agent 到底能帮自己省下哪部分工作。

Leopold Aschenbrenner’s complete Q2 performance before his newest trades release Monday Top tier: • SanDisk $SNDK: -62% • Micron $MU: -34% Second tier: • Bloom Energy $BE: -38% • Taiwan Semiconductor $TSM: -18% • Nebius $NBIS: -59% Third tier: • CoreWeave $CRWV: -49% • Core Scientific $CORZ: -45% • STMicroelectronics $STM: -39% • Applied Digital $APLD: -49% • Riot Platforms $RIOT: -37% • SharonAI $SHAZ: -74% • IREN $IREN: -61% Smaller positions: • T1 Energy $TE: -82% • WhiteFiber $WYFI: -76% • Babcock & Wilcox $BW: -69% • Keel Infrastructure $KEEL: -69% • Bitdeer $BTDR: -66% • HIVE Digital $HIVE: -58% • CleanSpark $CLSK: -56% • Cerebras $CBRS: -53% • Vishay $VSH: -52% • ProPetro $PUMP: -49% • Solaris Energy $SEI: -23% Options: • Bloom Energy $BE [Calls]: $44,100,000 • Taiwan Semiconductor $TSM [Calls]: $23,900,000 • Infosys $INFY [Puts]: $5,200,000 Summary: Every common stock position is red. His entire portfolio is down 46.8% from Q2 release to present day. His newest trades release Monday, November 16th, showing what he owned on September 30th, after he sold most of this to Citadel in late July

🚨 A 19-YEAR-OLD STUDENT TURNED $68 INTO $750K USING AI. No trading experience. No emotional decisions. No endless chart-watching. Just Claude Code, GPT-6 Astra, and 6 AI agents working together. I built a self-learning AI trading system that generated +$1,840.82 in just 48 hours. [ 𝐍𝐨𝐭𝐞: 𝐅𝐨𝐥𝐥𝐨𝐰 𝐌𝐞: @EaStEnRaY1 (𝐅𝐨𝐫 𝐢𝐧𝐬𝐭𝐚𝐧𝐭𝐥𝐲 𝐚𝐮𝐭𝐨 𝐃𝐌)] Here's what makes it different: 🧠 6 AI agents monitor newly launched Solana coins. ⚡ The system analyzes buyer activity, connected wallets, selling pressure, and holder concentration. 🎯 Every launch is evaluated across 3 probability scenarios: - +60% reached first - -30% reached first - Neither happens within 60 minutes QUANT evaluates profitability after execution costs. GUARD filters out trades with unacceptable risk. And here's the crazy part: Even the trades it SKIPS become learning data. Every new signal helps the system improve its next decision. No panic. No FOMO. No emotional trading. Just AI-powered market analysis and execution. 🔥 NOW I'M GIVING AWAY MY EXACT CLAUDE CODE SETUP FOR FREE! Want to build your own AI trading system? I'm sharing the setup for the next 24 HOURS ONLY. 👇 HOW TO GET IT: 1️⃣ Like this post 2️⃣ Repost 3️⃣ Comment "CLAUDE" 4️⃣ Follow @EaStEnRaY1 so I can DM you the setup. Don't just watch AI change trading. Learn how to build with it. 🚀

Wintermute is literally paying $10M a year to legally front-run everyone on Hyperliquid and it’s completely destroying altcoin liquidity Bloomberg finally exposed what HFT desks have been screaming about for weeks. Hyperliquid’s gossip priority auctions let top funds see incoming orders 50–60ms before anyone else on the network Look at how completely broken this setup is: -> One Wintermute wallet won ~76% of the top slots last month, burning ~$26k every single day just to starve out competition -> HL claims the auction is open to everyone. Complete BS. You only get that 60ms head start if you have direct peering with Foundation nodes, which requires 0.5% of total maker volume and 10k HYPE staked ($700k+) -> Smaller MMs are getting sniped nonstop and just unplugging their bots on alts -> Retail gets stuck with the bill: across 86 altcoin pairs, order book depth on HL imploded by 82% while spreads blew out from 2.13 to 3.55 bps Honestly, I don't even blame Wintermute. If I had their balance sheet, I would milk this exact loophole until someone forced me to stop The blame sits 100% on the protocol design. Burning auction fees is cool for $HYPE holders, sure, but letting one fund buy a monopoly on data speed turns a decentralized exchange into a rigged casino Scrap read-priority auctions completely or make the slots equal, because right now altcoin execution on HL is absolute garbage

I LAUNCHED A GROK BOT TERMINAL OVERNIGHT, GAVE IT $44.82 AND STOPPED ASKING AI WHAT IT WOULD DO. BY MORNING IT HAD $11,643.21 AND SIX AGENTS LOGGING EVERY TRADE I spent a year asking agents what they'd trade and watching the chat scroll. Last weekend I built the thing that actually trades, in Grok Bot, with no code I touched by hand. The six agents share one screen, each one job: > scanner reads new token pairs and volume spikes > rug check filters suspicious tokens before anything executes > sizer sets position from account equity, not a fixed ticket > executor opens and closes with slippage tracking > logger writes the reason behind every trade > equity curve updates on every close I went to bed on $44.82 and left the laptop open on the dashboard. By morning the balance read $11,643.21. Every trade on the activity feed had a reason next to it and a line in the log. That's the shift I didn't see coming. Not that an agent made money overnight. That I stopped being the one asking "would you long this?" and started being the one reading the log of what already happened. Starting balance: $44.82 Current balance: $11,643.21 Grok Bot built the terminal, the agent architecture and the dashboard. I described what I wanted and the thing went live. The full build is in the article below ↓ The whole build is in the post below ↓ Comment " Grok_Bot " 2️⃣ Like and Repost 3️⃣ Follow me @lihazadn_Ai I'll DM you the complete setup.