1. @ylecun
原文链接: https://x.com/ylecun/status/2040770225990172676
作者: Yann LeCun (@ylecun)
互动数据: ❤️ 280 | 🔄 13 | 👁️ 26,906
转发自: @ylecun
原文内容
@elonmusk Thinking in language has limited applications, largely in coding and mathematics where the language itself can help reasoning.
But, as I’ve been saying for years, thinking manipulates mental models in abstract (continuous) representation space.
Soooo, xAI gonna use JEPA now?
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
2. @DavidSHolz
原文链接: https://x.com/DavidSHolz/status/2040939992068096018
作者: David (@DavidSHolz)
互动数据: ❤️ 110 | 🔄 10 | 👁️ 4,342
原文内容
we’ll be surprised how available and how weird jobs get in the future. things like ENTROPY HUNTER - where you investigate things like: “whats a rock taste like?” or “whats it like to not see the color blue for 2 weeks then suddenly walk into a room thats totally blue?”
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
3. @teortaxesTex
原文链接: https://x.com/teortaxesTex/status/2040939480303989115
作者: Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex)
互动数据: ❤️ 15 | 🔄 1 | 👁️ 1,217
原文内容
The problem with ML is that CS techniques that a) don’t work well on GPUs and b) don’t scale to gigawatt-class datacenters are essentially a dead end. It’s a great tragedy for many very clever lines of prior research.
Downstream training & inference economics > best case proofs.
📎 引用推文 - @part_harry_
Sparse attention is just a form of appropriate maximum inner product search. This is a widely researched field, with leading techniques like Hierarchical Navigable Small World Graphs achieving ~log N time complexity.
Meanwhile, DSA and even HISA are both still linear time
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
4. @crystal_horror_
原文链接: https://x.com/crystal_horror_/status/2040854342337437919
作者: horror (@crystal_horror)
互动数据: ❤️ 17 | 🔄 1 | 👁️ 1,373
转发自: @teortaxesTex
原文内容
That supposed to be done in the early 20th century. A basic eugenics program to mass manufacture JVNs. AI is happening now (instead of earlier) precisely because that didn’t happen
📎 引用推文 - @mephXBT
If all the money that was spent on training AIs and buying up compute
Was instead spent on researching ways of making people smarter and boost iq
We would have reached kardashev 2 already
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
5. @menhguin
原文链接: https://x.com/menhguin/status/2040918999337410617
作者: Minh Nhat Nguyen (@menhguin)
互动数据: ❤️ 12 | 🔄 0 | 👁️ 1,077
原文内容
wild that claude has the uptime of a newly viral PMF startup, except with revenue and infra spend larger than most countries
📎 引用推文 - @anishmoonka
Claude .ai down to one nine (less than 99% uptime in the last 90 days)
Reliability is the biggest factor in enterprise adoption and retention! https://t.co/gom4YcwWNP
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
6. @amasad
原文链接: https://x.com/amasad/status/2040918650245578906
作者: Amjad Masad (@amasad)
互动数据: ❤️ 183 | 🔄 12 | 👁️ 23,026
原文内容
New @X API + @Replit is lots of fun.
@tannerlbraden built a @NASAArtemis II mission visualization w/ live @X feed & stats. https://t.co/ptLIf58yCr
💡 核心观点
（基于推文内容的技术洞察）
🛠️ 可实践建议
✨ 创作灵感
📱 社交媒体文案
即刻:
小红书:
Twitter/X:
关于本摘要
本摘要来自 Naval AI List 的每日精选推文，由 AI 自动筛选和整理。
筛选标准:
- 技术深度：新的方法论、工具使用技巧
- 实用性：可立即应用到工作流
- 时效性：24小时内发布
- 独特性：观点新颖非泛泛而谈
- 可操作性：提供具体步骤或工具
最后更新: 2026-04-06 00:33 UTC
