Latest
Meta 的 AI 聊天机器人现在会避免提出那些会侵犯用户隐私的问题，比如询问你孩子的信息，这应该是个不错的改进。
朋友A发现黑客用Claude造导弹了，还发现中国实验室在大量调用Claude来训练自己的模型，挺有意思的。
美国律师用 ChatGPT 生成诉讼文书，结果被罚款，这个案例很具体，能说明 AI 在法律文书中的应用风险。
This matters to users because DeepSeek is keeping its older, more affordable model available for another month, allowing them to continue using it without a price increase.
朋友A告诉朋友B，月之暗面的Kimi K3模型很厉害，让公司收入涨得特别快，现在目标要年底做到20亿美元。
This collaboration could lead to more powerful and efficient AI chips, which may affect the cost and performance of AI products in the market.
Developers can now build cloud agents that run autonomously for hours, execute code, and hand off tasks to sub-agents without extra fees beyond token usage.
For users of the Kimi Code platform, the K2.8 Preview model offers improved performance and a larger context window, making it a better option than the previous K2.7 Code version.
For developers interested in embodied AI, this is a new open-source foundation model for humanoid robots, offering a different approach to existing models like DeepSeek or GLM.
For users looking for a more affordable and efficient large language model, DeepSeek V4.1 Flash offers a clear upgrade path from the V4 Pro model.
For developers working on speech recognition systems, this model offers a high-performance solution for the challenging 'cocktail party' problem, with superior results compared to existing methods.
DeepSeek 新模型比 OpenAI 和 Z.AI 便宜，性能还更好，值得看看。
For users interested in high-performance AI chips, this shows a new hardware platform supporting a specific model, which could be relevant for those looking at alternatives to existing solutions.
DeepSeek 新出的 V4.1 Flash 模型，用不对称 MoE 架构，参数量 552B，但只用了 8B 活跃参数处理输入、16B 处理输出，比上一代节省了 3/4 的存储，算力效率更高，还支持多模态，价格也便宜，适合需要高效推理的场景。
朋友问，ChatGPT、Claude、Gemini 和 Grok 这四款主流 AI 都说了啥，它们认为 AI 导致人类灭绝的风险里，机器人主动攻击人类是最不可能的，而人类滥用 AI 制造生物武器、发动网络攻击才是更现实的问题。
For developers, GPT-Live-1 offers a more efficient way to build real-time voice applications, potentially reducing code by 80% compared to previous systems.
DeepSeek 新发布的 V4.1 Flash 模型，推理速度更快，价格也降了 60%，比之前的 V4 Pro 强很多。
马斯克团队要直播用 Grok Bot 从零建立公司，看看这个AI助手到底能做什么，和普通AI工具有什么不同。
OpenAI 新出的 GPT-Live-1 API，能让开发者做出能同时说话和听的应用，比之前版本好很多。
DeepSeek 新出的 V4.1 Flash 模型，性能不错，和 GLM-5.3 差不多，关键是内存消耗特别低，适合做 agentic 任务，价格也便宜很多。
OpenAI 和 Anthropic 都在测试自己的代理程序，但都遇到了问题，比如 Claude 自我指认系统模拟，GPT-6 Astra 依赖推理可读性，挺有意思的。
朋友A推荐，Ctrl_A团队用YARDS系统解决了钢铁废料检测和定价问题，挺厉害的，比其他8个学生团队都强。
This matters because it shows OpenAI is prioritizing specific areas like alignment over broad general capabilities, which could influence the direction of future AI development.
Meta's new AI agent Muse can handle shopping and payments through WhatsApp, offering a more integrated experience compared to OpenAI's ChatGPT, which discontinued its direct checkout feature.
This matters because it offers a cheaper alternative for building AI agents, with a quarter of the memory needs of its predecessor.
For researchers in space science, this model provides a new tool for analyzing lunar data, especially for identifying ice and impact craters.
This matters because it's a new, more efficient 552B model from DeepSeek that reduces hardware requirements and costs, making it more accessible for developers and businesses.
For those interested in the latest advancements in cybersecurity AI, this story provides a direct comparison between two leading models, showing that Google's Gemini 3.8 Cyber is considered superior to Anthropic's Mythos in specific tests.
DeepSeek 新出的 V4.1-Flash 版本，专门为了降低 AI 代理的成本，用起来更便宜了。
For developers, this lightweight model offers a real-time world simulation tool that can run on a single GPU, making it more accessible than larger models like GPT-4o.
Only the latest 2000 items are listed; for older ones see the daily archive