【AI 英文奏折】2026年07月17日
共收录 20 篇深度内容
📋 今日内容速览
快速浏览，点击感兴趣的推文查看详细分析
- Ethan Mollick: 核心观点总结中…
- Machina: 核心观点总结中…
- Santiago Valdarrama: 核心观点总结中…
- Gary Marcus: 核心观点总结中…
- swyx: 核心观点总结中…
- Suchen Zang: 核心观点总结中…
- Gary Marcus: 核心观点总结中…
- Gary Marcus: 核心观点总结中…
- Sharyph: 核心观点总结中…
- Ethan Mollick: 核心观点总结中…
- Machina: 核心观点总结中…
- Chubby♨️: 核心观点总结中…
- ChatGPT: 核心观点总结中…
- Alex Veremeyenko: 核心观点总结中…
- Rohan Paul: 核心观点总结中…
- Chubby♨️: 核心观点总结中…
- Venture Twins: 核心观点总结中…
- Gary Marcus: 核心观点总结中…
- Gary Marcus: 核心观点总结中…
- Aakash Gupta: 核心观点总结中…
📖 详细内容
Ethan Mollick @emollick
Ethan Mollick Professor @Wharton studying AI, innovation & startups. Democratizing education using tech. Author of Co-Intelligence | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
Kimi K3 cannot write a good murder mystery (though neither can any other model). That remains the jaggedest of frontiers. They both make things too obvious (the letter) and too obscure, and cannot foreshadow to save their artificial lives.
⏰ 11:01 | ❤️ 24点赞 | 📝 38词 | 查看原文 →
Machina @exm7777
Machina running ai-powered agencies | weeklyaiops.com | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
your only focus for the rest of the year should be converting this level of intelligence into systems that pay you for the next decade... everyone can now rent the same intelligence for $200 a month, which means intelligence stopped being the edge the moment you got it what you convert it into is the edge an answer gets spent the day you receive it... a system keeps paying you after you've forgotten you built it
⏰ 04:58 | ❤️ 219点赞 | 📝 76词 | 查看原文 →
Santiago Valdarrama @svpino
Santiago Valdarrama | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
How to monitor the web and research what comes back in 300 lines of code. I want to buy some cheap products on Amazon, but I want to grab them when prices drop. I'm using Parallel @p0 here, which is a really easy way to ground your agentic code. Here is how this works: 1. Set up a monitoring event with Parallel for the products 2. If there's a hit, I get a webhook event with the details 3. My script immediately kicks off a Deep Research task 4. The final result tells me whether the product is worth buying I probably spent 1 hour building all of this with Claude Code + the Parallel Agent Skills. I'm linking to the GitHub repo below. Just think about what you could do with this. Here is an easy example (I also set this up): • Monitor news for public tech companies • Analyze the news + price • Assess whether it's worth investing You can do this, put it on autopilot, and get an email anytime there's something worth your time. Here is the video I recorded.
⏰ 22:38 | ❤️ 66点赞 | 📝 179词 | 查看原文 →
Gary Marcus @garymarcus
Gary Marcus OG GenAI Skeptic; spoke at US Senate. Advocating world models | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
i would have made a million bucks if Elon had taken my 2024 bet about AGI not coming by end of 2025, but still people say I shouldn’t bet against him
⏰ 10:38 | ❤️ 22点赞 | 📝 30词 | 查看原文 →
swyx @swyx
| 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
we make sure some of the top YC AI companies are featured at AIE every single year. this year, we were graced by @garrytan and @eve_bouff to cap off our startups and design engineer focused audiences respectively. Really enjoyed these raw high value perspectives!
⏰ 10:10 | ❤️ 21点赞 | 📝 43词 | 查看原文 →
Suchen Zang @suchenzang
| 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
But when the Pharisees heard that he had silenced the Sadducees, they gathered together. And one of them, a lawyer, asked him a question to test him. “Teacher, which is the great commandment in the Law?”
⏰ 09:39 | ❤️ 20点赞 | 📝 36词 | 查看原文 →
Gary Marcus @garymarcus
OG GenAI Skeptic; spoke at US Senate. Advocating world models | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
The other thing that is absurd about $SPCX’s valuation is that Musk just isn’t that good at AI. This new BI article on the chaos within xAI is a case in point. And it’s why he is transitioning to being a cloud provider rather than a frontier leader.
⏰ 09:21 | ❤️ 28点赞 | 📝 51词 | 查看原文 →
Gary Marcus @garymarcus
OG GenAI Skeptic; spoke at US Senate. Advocating world models | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
in a few minutes Trump is likely to make some baseless claims about election fraud in order to distract from a failed war and the biggest insider trading scandal in history.
⏰ 08:59 | ❤️ 98点赞 | 📝 31词 | 查看原文 →
Sharyph @sharyph_
Sharyph Accelerating Creators with AI | Showing how to use AI to start, create, grow, and scale faster | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
I first paid attention to AI video because it could make clips. Now I am watching a different question: can it become interactive? @PixVerse_ is exploring that idea through PixVerse Game, bringing together real-time generative video, AI Agents, game mechanics, and user-defined worlds. It is not a finished game engine, and AI gaming has not been solved. I see it as a fun early look at how video models might eventually become part of worlds people can influence and interact with. Want to get in? A few invitation codes will be available in the comments. It is a limited drop, so first come, first served. Here is the experiment those codes unlock: https://world.pixverse.ai/game
⏰ 21:52 | ❤️ 43点赞 | 📝 118词 | 查看原文 →
Ethan Mollick @emollick
Professor @Wharton studying AI, innovation & startups. Democratizing education using tech. Author of Co-Intelligence | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
So I guess it is time to wonder: how does pre-clearance work for open weights models? No model card yet from Kimi K3 but maybe at weight release in a couple weeks, yet open models are easy to jailbreak. Do open models claiming to be Mythos/Sol level (K3 is not yet there, but someone will reach it soon) get vetted by the US/UK/etc? Will China start to care about cyber risk? Since policy has been emergent, I guess no one knows. And before anyone says you can’t recall open weights models once released, that is true! But governments have levers to require that no company doing business with their nation’s citizens uses unvetted models or serves them, etc. No one is taking away your downloaded weights, but it is entirely possible to make any company avoid using them due to their higher risk. All of this points to the need for some sort of international cooperation on model vetting.
⏰ 08:26 | ❤️ 113点赞 | 📝 163词 | 查看原文 →
Machina @exm7777
running ai-powered agencies | weeklyaiops.com | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
you shouldn't use any software that hasn't shipped an MCP or a CLI learning a new interface is literally negative EV now: the setup, the tabs, the clicking... every hour of it is you doing labor your agent can't inherit the replacement is already here: voice prompt your harness and let it drive the software for you if a tool only ships an API, PrintingPress prints you a CLI for it, and it works fine most of the time for example, Kit and beehiiv already ship their own MCPs: - configure the emails - manage subscribers - draft the full newsletter with your own skills - send it and read the analytics ... without opening either app once software you can't script is software you work for btw
⏰ 01:57 | ❤️ 130点赞 | 📝 127词 | 查看原文 →
Chubby♨️ @kimmonismus
Chubby♨️ Dream realized! Turned my love for AI into a career – sharing daily. Get my newsletter | 影响力: 225k+万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
2026 is the year I'll truly feel and fully grasp the acceleration for the first time. Countless releases, each one a significant leap forward, China and the USA locked in a race to outdo each other. 2024 and 2025 felt incredibly slow in comparison.
⏰ 07:44 | ❤️ 358点赞 | 📝 42词 | 查看原文 →
ChatGPT @chatgptapp
ChatGPT ChatGPT is for the people. | 影响力: 543.86k万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
A look at computer use and the built-in browser in the new ChatGPT app @dkundel walks through how ChatGPT can work with apps on your computer and browse the web to research, navigate websites, and complete tasks with you.
⏰ 07:29 | ❤️ 580点赞 | 📝 40词 | 查看原文 →
Alex Veremeyenko @alex_verem
Alex Veremeyenko Marketing + AI = $$$ 🔑 @godofprompt (co-founder) Also know as: @alex_prompter | 影响力: 96.5k万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
Anthropic and Andrew Ng dropped a free 2-hour masterclass on building AI agent skills from scratch. Covers everything from Claude agent skills to MCP integration to long-running agent architecture. 00:00 - How to build agent skills with Claude 22:32 - Claude pre-built skills for AI agents 41:07 - Agentic skills vs tools, MCP, and subagents 01:06:06 - Skills for long-running agents People are paying $997 for bootcamps that cover half of what's in this video. This is the most practical walkthrough on agentic development you'll find anywhere right now. Bookmark it. Watch it tonight. Then read the guide below.
⏰ 20:44 | ❤️ 153点赞 | 📝 95词 | 查看原文 →
Rohan Paul @rohanpaul_ai
Rohan Paul Compiling in real-time, the race towards AGI. The Largest Show on X for AI. | 影响力: 0万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
Kimi K3 just beat Opus 4.8 at building retro games for half the cost. Atomic Chat, a desktop app that runs LLMs locally, asked each model to build Road Fighter, Battle City, and Q*bert as self-playing HTML files. To make each result work, the graphics, enemies, game rules and a player that could play on their own all had to be in the same file. Because a single error can ruin the entire simulation, setup tests are more important than code generation. Outputs: Kimi K3: 18.4K tokens, $0.28 GPT-5.6: 18.1K tokens, $0.28 Opus 4.8: 21.3K tokens, $0.54 GPT-5.6 generated the best looking textures, but the logic for the cars/tanks did not work. Shows how great looks can be deceiving for poor control and state management. Kimi's best work was Q*bert, where the agent changed the state of the cube, moved around the pyramid, and avoided an enemy, without breaking the game loop.
⏰ 06:52 | ❤️ 84点赞 | 📝 147词 | 查看原文 →
Chubby♨️ @kimmonismus
Dream realized! Turned my love for AI into a career – sharing daily. Get my newsletter | 影响力: 225k+万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
NVIDIA’s new #1 RTEB embedding model is built around a simple idea: the cheapest reasoning token is the one an agent never generates. Nemotron-3-Embed-8B reaches 78.46 average NDCG@10 on RTEB and 75.45 on MMTEB Retrieval. NVIDIA tested it inside a search agent powered by Nemotron 3 Ultra. More accurate retrieval returned relevant evidence earlier, reducing repeated searches, context inspection and reasoning turns. The result: NVIDIA says the 8B model achieved the highest retrieval accuracy and lowest estimated downstream token cost across three agentic evaluations. For cheaper deployment, NVIDIA also released two 1B models. The Blackwell-optimized NVFP4 version offers up to 2× BF16 throughput while retaining more than 99% of its retrieval quality. All three support 32K context, multilingual and code retrieval, enterprise RAG and agent memory. The weights, data and recipes are open for commercial use. Agent cost optimization is moving into the retrieval layer.
⏰ 05:45 | ❤️ 106点赞 | 📝 137词 | 查看原文 →
Venture Twins @venturetwins
| 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
Netflix's CEO revealed that the company has already used AI in production across 300+ titles. In many cases, gen AI tools were used for "key shots" that would've been left out due to cost. And they just released a documentary with 17 minutes of AI-enhanced footage.
⏰ 06:34 | ❤️ 137点赞 | 📝 47词 | 查看原文 →
Gary Marcus @garymarcus
OG GenAI Skeptic; spoke at US Senate. Advocating world models | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
“the steepest commodity curve any technology has run in recorded history” - exactly what I have been warning people about in Marcus on AI since August 2023. We had nearly three years to prepare for this moment, but almost nobody listened.
⏰ 06:28 | ❤️ 143点赞 | 📝 39词 | 查看原文 →
Gary Marcus @garymarcus
OG GenAI Skeptic; spoke at US Senate. Advocating world models | 影响力: unknown万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
I have a great idea! To protect two closed-source AI American companies (and some very large investments by Josh Kushner), let’s make every single other American company pay more for AI than it otherwise might.
⏰ 06:22 | ❤️ 70点赞 | 📝 37词 | 查看原文 →
Aakash Gupta @aakashgupta
Aakash Gupta ✍️ product-growth.com 💼 https://t.co/STzr4nqxnm 🤝 https://t.co/SqC3jTyP03 🎙️ https://t.co/fmB6Zf5UZv | 影响力: 278.7k万粉丝
💡 核心观点: 核心观点总结中…
可信度: 1/10 – 1项暂无法验证
事实核查:
- ✗ 无法验证: 事实核查功能暂时不可用 (系统处理中)
原文内容:
30 engineering teams entered her company's internal hackathon. A PM won it. The idea she used was sitting in a public Anthropic blog post. All 30 teams could have read it. Here's what she built with it. One agent writes. A second agent gets configured with what actually matters at her company, then attacks the first agent's output. Flaws route back, the first agent revises, the loop runs until the work clears the bar. She pointed it at the real codebase and shipped it into production. GANs, applied to agents. Now look at where the leverage actually sits in that setup. The builder agent is commodity. Every team in that room had the same models and the same context window. The scarce input was the config: a written definition of what "good" means at that company, precise enough that a machine could enforce it with no human in the room to interpret it. That's a spec. PMs have claimed to own the spec for twenty years. This is the version where vagueness has a cost you can measure. Jyothi Nookula has been an AI PM since before that was a title. She rebuilt the evaluator live in this episode, starting from an empty folder. Most PMs are still trying to get the agent to produce the right answer. The people winning are building the thing that decides whether the answer is right. Write the evaluator.
⏰ 06:14 | ❤️ 39点赞 | 📝 237词 | 查看原文 →
相关文章
