产品发布/更新
-
Runway launches Model Context Protocol server
Runway officially launched the Runway MCP server, allowing any MCP-compatible AI agent (such as Claude, ChatGPT or…
-
Claude Code v2.1.152 released
Claude Code released version v2.1.152. Key improvements include: `/code-review --fix` now applies review suggestions…
-
OpenAI products support secure connections to private MCP servers
Your team can keep MCP servers inside your internal network while ChatGPT, Codex, and the Responses API connect via…
-
Open-sourcing the FastVideo Dreamverse real-time video generation tool
🚀 Generate a 30-second 1080p video in just 7 seconds! We have open-sourced FastVideo Dreamverse: an…
-
Grok coding agent lands on the Kilo IDE platform
Use your SuperGrok or X Premium+ subscription in @kilocode. Try grok-build-0.1 for high-speed, agentic coding…
-
Perplexity open-sources Unigram tokenizer to cut CPU usage
We have open-sourced a rebuilt Unigram tokenizer that cuts CPU usage by 5-6x. Small rerankers and embedding models run…
-
TRL adds Delta Weight Sync: weight deltas sent via Hub Bucket, per-step transfer cut from 1.2 GB to 20–35 MB
In asynchronous reinforcement learning, the trainer must send full model weights (e.g., a 1T-parameter checkpoint of…
-
OpenCode与MiMo V2.5限时免费开放
OpenCode x MiMo V2.5 - 限时免费 1M 上下文 • 推理 • 文本 • 图像
行业动态
-
Jensen Huang unveils NVIDIA's new Taiwan campus
Jensen Huang showcased the new Taiwan campus. NVIDIA plans to invest about $150 billion a year in Taiwan, one week…
-
China to speed up research on comprehensive AI health development legislation and low-altitude economy legislation
At a State Council Information Office press conference, the Ministry of Justice said 2026 will see government…
-
The Pope keeps a level head on AGI
Pope Leo XIV issued an encyclical titled Magnifica Humanitas, warning that the use of AI is by no means a purely…
-
Qualcomm reportedly strikes AI ASIC chip deal with ByteDance, orders in the millions
Bloomberg reports that Qualcomm has reached an AI ASIC cooperation agreement with ByteDance, which will purchase…
-
2026 election information and safeguards
As global elections approach in 2026, efforts to safeguard them span three areas: helping the public access election…
论文研究
-
Fast, faster, Qwen. 🚀
Qwen3.5 reached a record 580 tokens per second (tps) on the TokenSpeed inference engine for agent workloads. The…
-
ITBench-AA: Frontier models all score below 50% on first agentic enterprise IT task benchmark
The ITBench-AA SRE benchmark from Artificial Analysis and IBM shows that no frontier model scored above 50%. Claude…
-
Privacy analytics via zero-trust aggregation
Google Research introduced a new privacy analytics solution. It combines a new cryptographically secure aggregation…
-
Coding agents in the social sciences
A survey of 1,260 quantitative social scientists shows that while 81% of respondents have used AI chatbots, only 20%…
技巧与观点
-
Securing source code with large language models
This article shares best practices for using Claude Opus to build threat models, find code vulnerabilities, and…
-
I think Anthropic and OpenAI have found product-market fit
Anthropic and OpenAI have found product-market fit through coding agents, driving a significant rise in costs for…
-
Reachy Mini achieves fully local voice interaction
The Reachy Mini robot can now achieve fully local voice interaction via the `speech-to-speech` library, with no cloud…
-
OpenAI's Altman says AI's impact on white-collar jobs is less severe than expected: I'm glad I was wrong
OpenAI CEO Sam Altman acknowledged that earlier predictions that AI would eliminate large numbers of entry-level jobs…
-
A zero-trust security framework for AI agents
Anthropic released a security framework for enterprises deploying autonomous AI agents, noting that frontier large…