← 返回最新日报
2026-07-04· 当日存档
AI 日报
AI学习成本与开源竞争
今日看点：2.6万学生研究揭示AI辅助学习长期考试表现下降24%，影响需两年显现；开源AI让中美在算力差距10倍下仍能竞争，效率是关键。此外，Anthropic启动药物发现项目，微软加入AI超级应用竞争。
论文研究
1
A 26,000-student study shows AI's hidden learning cost takes two full years to surface
AI辅助学习短期提分但长期考试表现下降，影响需两年才显现。
The Decoder深挖 →
2
🤖 From this week's issue: A research blog post introducing R&B-EnCoRe, a self-supervised pretraining cycle that lets vision-language-action models ...
斯坦福提出自监督预训练方法，让视觉-语言-动作模型自主发现推理步骤。
X · List深挖 →
3
UK's AI Security Institute finds standard benchmarks systematically underestimate what AI agents can actually do
标准基准因计算预算限制低估AI智能体实际能力。
The Decoder深挖 →
行业动态
4
5
6
7
Anthropic launches its own drug discovery programs to tackle diseases Big Pharma considers unprofitable
Anthropic启动药物发现项目，针对制药业认为无利可图的疾病。
The Decoder深挖 →
8
Security vulnerability reports have exploded since AI models started hunting for bugs
AI参与漏洞挖掘后安全报告激增，单月创历史新高。
The Decoder深挖 →
9
Google DeepMind and A24 announce first-of-its-kind research partnership
谷歌DeepMind与A24达成首个AI与电影制作研究合作。
Google DeepMind深挖 →
10
Claude Code's complicated China problem involves bans on both sides of the Pacific
Claude Code因地缘政治和技术限制在美中面临复杂禁令。
The Decoder深挖 →
产品上新
11
I kept asking Claude Fable to make the game "more AAA" over and over again. The results are... interesting. In Claude's view, this meant upgrading gra...
反复要求Claude Fable升级游戏至AAA级，结果不断进化至WebGL极限。
@emollick深挖 →
12
Microsoft follows Anthropic and OpenAI into the AI super app race with overhauled Copilot and AutoPilot agents
微软计划合并Copilot并推出AutoPilot代理，加入AI超级应用竞争。
The Decoder深挖 →
13
14
OpenAI cofounder envisions "almost no interface" future where nobody learns software anymore
OpenAI联合创始人预测未来几乎无界面，无人再学软件。
The Decoder深挖 →
15
16
I use glm 5.2 in claude code via hf claude almost daily now moved over completely to open models
用户已完全转向开源模型，日常使用GLM-5.2。
@_akhaliq深挖 →
17
The browser wars aren’t about search anymore — here are the best alternatives to Chrome and Safari
盘点挑战Chrome和Safari的替代浏览器，聚焦AI时代新功能。
TechCrunch AI深挖 →
模型动向
18
Mistral's open-source Leanstral 1.5 aces formal math benchmarks and catches real bugs in code
Mistral发布开源模型Leanstral 1.5，在形式化数学基准中表现优异。
The Decoder深挖 →
实用技巧
19
20
21
22
23
24
25
26
Anthropic developer shares prompting tips for Fable 5 that focus on finding your own blind spots first
Anthropic开发者分享提示技巧：先发现自身盲点再交给Claude。
The Decoder深挖 →
27
28
29
快讯
·
June 2026 newsletter
Simon Willison