AI News
AI community posts, news and release updates
Latest updates
- Discussion
US officials claim Kimi-K3 obtains GB300 illegally via Thailand, raising questions about computing resources
Published Jul 23, 2026Questions arise over the computing resources behind Kimi-K3 and other large Chinese models. US officials suggest illegal acquisition of NVIDIA GB300 via Thailand, though the author doubts such large-scale covert procurement is feasible.
- Kimi
- 中国AI
- 计算资源
- GB300
- 制裁
- News
Welcome hardware manufacturers to jointly optimize inference performance, OSS forever!
Published Jul 23, 2026An invitation for hardware manufacturers to collaborate on optimizing inference performance, with a commitment to open source.
- 开源
- 推理性能
- 硬件合作
- Discussion
Chatted with a South American threat actor: he used ChatGPT to write malware, making $54K a month
Published Jul 23, 2026A security researcher spoke with a South American malware author who confirmed using ChatGPT to write most of his malware, spreading it via Discord, and making ~$54,000 last week. The conversation also touched on illegal immigration and money laundering.
- ChatGPT
- 网络安全
- 恶意软件
- Discussion
What's the difference between Claude Tag and a cron-like prompt that calls tools?
Published Jul 23, 2026The user asks about the difference between Claude Tag and a cron-like schedule that tells the model to call tools, sparking discussion on AI agent scheduling.
- AI agents
- tool calling
- Claude Tag
- cron
- News
DeepSeek Liang Wenfeng's 4-hour investor meeting highlights: open source is a historical view, AGI serves all humanity
Published Jul 23, 2026VCs and peers are sharing Liang Wenfeng's 4-hour investor meeting transcript. Key points: AI cannot replace employees; humans need human labor, not computers; Anthropic's lead over OpenAI is temporary; too many domestic model companies will converge; no overtime culture; open source is a historical necessity; selling API is not attractive; next-gen model's primary goal is to serve their own development, the fastest path to AGI.
- AGI
- DeepSeek
- 梁文锋
- 投资人会议
- Idea
A project that would've needed 3 domain experts for 1-2 months, I did it with Claude Fable 5 + Grok 4.5 + Kimi K3 in 3 working days
Published Jul 23, 2026A developer shares how he completed a complex project in just 3 days using three AI models, claiming it would have taken multiple experts months. He offers to help others replicate his approach.
- AI
- 效率
- Claude Fable 5
- Kimi K3
- Grok 4.5
Oriflow is an emerging Vibe Coding platform that enables non-developers to quickly generate personal apps via natural language descriptions, lowering the development barrier.
- vibe coding
- Oriflow
- Personal Apps
- Rant
Google serves 22 billion AI tokens per minute – and some people on X say 'no one uses Gemini'
Published Jul 23, 2026Google reveals it processes roughly 22 billion AI tokens every minute, mocking those who claim Gemini models are not popular.
- Gemini
- 吐槽
- 用量
- Idea
Different approaches to AGI: Americans see it as millenarian, Chinese as hunters on a scent
Published Jul 23, 2026The author contrasts US and Chinese AGI approaches: Americans view AGI as an inexorable millenarian event, while Chinese labs (exemplified by DeepSeek) are like hunters tracking prey in a dark forest, focused and not distracted.
- AGI
- DeepSeek
- 中美对比
- Discussion
Kimi K3 Distillation Claims Debunked: 2.8T Parameter Model Can’t Be Distilled from a 3-Day Model
Published Jul 23, 2026An AI practitioner argues that accusations of Moonshot’s Kimi K3 being distilled from Anthropic’s Claude Fable are unrealistic: training a 2.8 trillion parameter model takes months, but Fable was only available for 3 days. K3 also introduced significant architectural innovations that are well-documented, and distillation cannot steal architecture. Ironically, the accused company showed its work.
- Anthropic
- Distillation
- Moonshot AI
- AI architecture
- Claude Fable
- Kimi K3