← 返回最新日报
2026-07-11· 当日存档
AI 日报
GPT-5.6破数学难题
GPT-5.6 Sol Ultra一小时解开50年数学猜想,并用700词提示词管理64个子Agent;同时谷歌发布Gemma 4论文,31B参数模型逼近闭源前沿。办公AI格局生变,GPT-5.6接管微软全家桶。
模型动向
1
2
GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture [pdf]
GPT-5.6 Sol Ultra证明图论经典猜想,数学推理新里程碑。
Hacker News深挖 →
3
4
OpenAI's GPT-5.6 Sol autonomously post-trained the smaller Luna model with a "fairly underspecified prompt"
GPT-5.6 Sol用模糊指令自主微调小模型,RL自训练能力惊人。
The Decoder深挖 →
5
This time it is novel math proofs with a public model (most of the other big math breakthroughs have been with experimental LLMs).
公开模型首次攻克长期未解数学猜想,AI科研能力再突破。
@emollick深挖 →
6
7
8
Google Research Introduces SensorFM: A Wearable Health Foundation Model Pretrained on One Trillion Minutes of Sensor Data
谷歌发布穿戴健康基础模型SensorFM,万亿分钟数据预训练,性能全面超越。
MarkTechPost深挖 →
9
Looks like automatic music transcription just got a serious upgrade 😲 Unsurprisingly, data is king, also for this task.
自动音乐转录模型MuScriptor发布,多乐器转MIDI精度大幅提升。
X · List深挖 →
10
11
12
China's Orca world model matches specialized robotics systems without ever seeing a single action label
北京智源发布世界模型Orca,无需动作标签即可匹配专业机器人系统。
The Decoder深挖 →
13
Ant Group’s Robbyant Unveils LingBot-VA 2.0: A Causal Video-Action Model Built Natively for Physical AI
蚂蚁集团发布灵宝VA 2.0,原生物理世界视频动作模型,支持因果推理。
MarkTechPost深挖 →
产品上新
14
15
16
17
18
OpenAI admits it "didn't get everything quite right" with ChatGPT Work launch and scrambles to fix UX and costs
OpenAI承认ChatGPT Work发布存在UX和成本问题,正紧急修复。
The Decoder深挖 →
19
20
论文研究
21
ICML 2026|小模型也能「指挥」大模型RL后训练:清华&腾讯提出GPS,最高减少69% Rollout成本
小模型筛选prompt降低大模型RL后训练成本最高69%,ICML 2026论文。
机器之心深挖 →
22
23
行业动态
24
SK Hynix raises $26.5B in the biggest foreign IPO in US history, is urged to build new US fabs
SK海力士创美国史上最大外资IPO,筹资265亿美元,被敦促在美建厂。
TechCrunch AI深挖 →
25
26
27
Apple sues OpenAI for allegedly running a "coordinated campaign" to steal trade secrets through poached employees
苹果起诉OpenAI系统性挖角窃密,超400名前员工跳槽,法律战升级。
The Decoder深挖 →
28
实用技巧
29
the org chart is starting to look different; small atomic teams, each one building on its own multibillion dollar market. used to be every team lived ...
组织架构转向小型原子团队,各自开拓数十亿美元市场,复合型创业时代。
X · List深挖 →
30
“如果你不能用300行代码写个Cursor,这行你就别待了!”Ralph Loop创造者、Claude Code核心技术设计者的暴论
AI编程工具普及,软件开发门槛降低,人人皆可成为开发者。
AI前线深挖 →
日报存档
2026 年 8 月
14 日AI日报:手语翻译、Grok 4.6与资本热潮13 日DeepSeek V4 Pro开源,AI行业热度飙升12 日AI日报:算力新纪元11 日AI日报:开源与安全并进10 日Meta开源多模态,AI安全引担忧9 日AI安全测试成新风险8 日AI安全与模型突破并进7 日AI安全与竞争新局6 日AI安全警钟与巨头变局5 日AI代理安全与商业化双线并进4 日阿里巨模型发布,开源生态激荡3 日Qwen3.8-Max领跑,AI应用百花齐放2 日AI 速读 · 2026-08-021 日AI安全警钟与模型新突破