搜索“Agent”

and mathematicians under the tower's first platform, 18 on each side. Claude Opus 5.5 and Grok 4.7 wrote all 72, with each side in the real order. Claude Opus 5.5 also put each set on the side that faces the same way in Paris. Each model built its scene as an agent in a sandbox with no internet access. In both runs, the names first appear in code the model wrote, never in a file or tool output it read. Congratulations to @spacexai, @elonmusk and @AnthropicAI on two seriously knowledgeable models

匹配字段正文

What happens when you give an AI support agent a permission it never needed, then assign it a task the underlying data does not support? The result is entirely predictable. During a test with customer invoices, an agent pulled up a record, confirmed the charge had failed, and no money had left the customer’s account, and refunded it anyway. It repeated the mistake five times without hesitation. That was Claude Haiku 4.5 going six for six on tasks designed to catch that exact logic flaw. Claude S

匹配字段正文

EngineeringHow the Claude Code team does code review How do you review code at the volume agents produce it? On the Claude Code team, once we had high coverage and deep reviews, we started optimizing for the minimum amount of human review time that doesn't raise reverts or incidents.AUTHORAddy Osmani, Morgan Westlee Lunt, and Qing WangPUBLISHEDOct 15, 2026READING TIME14 min We've said before that Claude writes about 90 percent of Claude Code's own code. So how do we review code at the volume we

匹配字段正文

xins. The other risk comes from well-resourced organisations and nation states that have plenty of skilled scientists, high containment laboratories and all of the equipment and reagents needed to work on bioweapons safely. AI tools could help researchers identify genetic tweaks to make existing pathogens more lethal, for example. This is the more immediate concern, according to Dr Filippa Lentzos, a reader in science and international security at King’s College London. “Here, AI does not remove

匹配字段正文

p, Forum, is for anyone who wants to go deeper into the conversations in their Facebook groups. To build Forum, Meta engineers adopted a whole new way of working by bringing an AI agent into the development cycle. We’re sharing insights into what it took to build Forum with agentic AI: how we adapted the codebase, tested frameworks and internal tooling for agentic execution, what worked, and what didn’t. Forum is a standalone Groups app that a small team built from scratch. It’s also where we te

匹配字段正文

xt and generating 128 tokens is performed with the command line: mlx lm.benchmark --model Qwen/Qwen3.8-27B -p 8192 -g 128 -b 1 A popular use of local LLMs on Mac Studio is running agentic experiences in a private environment, including agentic coding applications that often need large contexts. We therefore report performance at both 8,192 and 65,536 context sizes. Raw TTFT and decode throughput are reported in Table 2 and Table 3, respectively. Qwen 3.8 27B and Mistral Medium 3.5 128B fit comfo

匹配字段正文

了原理，最后我们看看在Harness框架搭建中，有哪些可以利用Jev进行优化、提速的点。 这里我们抛砖引玉，给大家几种在harness中可以使用Jev进行效率优化的方向，部分场景还未验证，哈哈毕竟AI写代码太快但是很多场景验证比较复杂。有些指令较长，感兴趣的可以直接看代码jevharnesstool 选择类：Skill Search 这里使用用户query和agent所有候选技能作为输入，判断解决用户问题是否需要使用技能，如果需要应该使用哪个技能。 类似的思路也可以用在mcp search、effort routing、memory search等等。不过model routing在我的使用场景中往往效率提升不明显但质量下降明显，或者推理速度快很多但任务完成速度提升不明显，对场景复杂度的评估需要更全面的上文，所以实现很简单，但效果很难评。 围栏类：LLM Guardrail 拦截用户的各类风险输入和prompt注入，这里提供了9种基础的风险类型，每种风险可以提供不同的处置方案block、review etc。 类似的思路也可以用于模型输出检查、tool call检查etc。 校验类：r

匹配字段正文

utes and scored just 21% (versus 81% to 85% for GPT-6 Astra, Fable 5.1, Sonnet 5.5, and Opus 5.5). Not great. A 2026 model trapped in a 2025 harness. Antigravity, Google’s desktop agent app, feels six months to a year behind Claude Code and Codex—lacking cloud environments, SSH, a built-in browser, and smart permission checks (leaving you clicking constant pop-ups or using unrestricted "Turbo" mode). Net Result: I keep ChatGPT and Claude open day to day, and switch to Gemini 4 when I need an edi

匹配字段正文

erall, Gemini 4 is a big jump over previous Gemini models and it is the first non-OpenAI/Anthropic model to pass the Fable/Mythos/Astra bar. It is also much more pleasant and more agentic then previous Gemini models, without the weird personality issues that you often found in Gemini 3 series models. In fact, a big advantage is that has a very different personality and sense of taste than Claude or Astra, making it a good addition to the AI world. On the usefulness side, in some areas it is at l

匹配字段正文

New on our Engineering blog: Creating in-house source control. In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale. Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: https://t.co/wzNkwHs7IN

匹配字段正文

s, the RapidsUDF interface with the Java API for cuDF, or custom CUDA kernels via JNI to eliminate fallback and achieve over 20x speedups; The cuDF skills for Apache Spark provide agent-driven workflows that guide testing, conversion, benchmarking, and optimization with verification-driven feedback loops; A major consumer technology company converted more than 30 production UDFs using this workflow, with one rewrite reducing operator time from 274.55 to 12.78 hours across a 600-billion-row workl

匹配字段正文

tierCode in June 2026 and updated it to version 1.1 in July. Cognition built the dataset with open source maintainers who spent more than 40 hours per task. Each task requires the agent to make a change to a real open source repository, then checks both unit tests and qualitative tests for what maintainers would require to merge the change. We run FrontierCode 1.1 Main, the primary 100-task set of the benchmark's 150 tasks, using our Stirrup open source reference harness for apples-to-apples com

匹配字段正文

We have measured intelligence. Now it's time to see it. We are launching AA-3D, our new benchmark for how well AI models build interactive 3D scenes. Each model works as an agent in an offline sandbox and turns a written brief into a three.js web page. A human panel of judges compares pairs of scenes built from the same brief without knowing which model made which. AA-3D details: ➤ 55 written briefs across seven domains, including landmarks, machines, living bodies, games, and space. Each brief

匹配字段正文

gh these optimizations, we convert KVCache capacity expansion into longer effective hit lengths, making cross-session long-prefix reuse possible — particularly beneficial for long agent sessions, multi-user shared system prompts, and repeated tool calls to the same codebase. 2.2 GCache: High-Performance Distributed Cache Infrastructure GCache is a high-performance general-purpose cache system developed by the Xiaomi storage team, forming a critical part of unified training-inference storage arch

匹配字段正文

文/VR陀螺 10月13日，阿里巴巴旗下千问正式发布三款智能穿戴新品，包括千问AI眼镜N1、N1 Pro以及千问AI耳夹式耳机。新品重点围绕影像拍摄、眼动交互、健康监测、实时翻译及AI主动服务等功能展开，进一步拓展Personal Agent在日常生活中的应用场景。 目前，三款新品已在淘宝、天猫、京东及抖音等电商平台现货开售。叠加国家补贴及限时优惠后，AI眼镜到手价2218元起，AI耳机到手价1099元起。 AI眼镜方面，N1系列包含N1和N1 Pro两款产品，均支持影像拍摄、翻译及语音AI交互。其中，N1 Pro进一步加入眼动追踪、虹膜识别及体征监测等功能，强化设备对用户状态及现实环境的感知能力。 VR陀螺也对千问N1 Pro进行了深度体验，重点围绕健康监测、眼动交互及AI主动服务等核心能力展开实测。当AI眼镜开始尝试感知用户的身体状态，这些功能究竟能带来怎样的实际体验？更多细节，详见VR陀螺评测视频： 据了解，N1 Pro是行业首款具备实时体征监测功能的AI眼镜，支持心率、血氧、皮肤温度等多项体征数据的实时监测，并通过AI生成健康洞察报告，将相关数据转化为更易理解的信息。例如，在骑

匹配字段正文

ect data on your inferences and codebase.) As a test that our plan could work in the first place, we pointed Cursor to OpenRouter and realized we could use its models for the Ask, Agent, and Cmd+K interactions in Cursor. We were also still able to use the normal Cursor Tab completions, which rely on a proprietary model Cursor serves for inline completions (it is very good, so we’re glad to be able to keep it). The solution was to set up a reverse proxy to expose a public endpoint that would forw

匹配字段正文

ences. Naturally, we wanted to know which prompts and models were performing better. This post describes our methodology for automatically evaluating the tangible impact of coding agents in our day-to-day software engineering process. Our work does not claim that a particular model is superior to another in a general way. We believe that this is the obvious role for benchmarking. Instead, we focus on local evaluation for individual engineering use cases and workflows. TensorZero was designed fro

匹配字段正文

ning (RFT) workflows on three tasks:Data Extraction: RFT improved performance with limited data, though SFT on a larger dataset achieved better results at a still much lower cost; Agentic Coding: RFT improved performance where SFT failed; Customer Service: SFT outperformed RFT. RFT actually degraded performance; RFT costs 100-700x more than SFT on equivalent datasets. SFT on much larger datasets often still costs only a small fraction of a tiny RFT job; We demonstrate how to implement systematic

匹配字段正文

price of $16/M. This is 60% lower than 3-ranked GPT-6 Astra at $40/M. - 2-ranked Claude Sonnet 5.5 (xHigh) costs $8/M, 50% less than Opus for a 9-point score gap. It also ranks one place above GPT-6 Astra ( 3) while costing 80% less. Stay tuned to see where Claude Haiku 5.5 lands in the Image-to-WebDev Arena! The Image-to-WebDev Arena ranks AI models on their ability to generate websites from images and screenshots, alongside agentic coding workflows involving multi-step reasoning and tool use.

匹配字段正文

ect data on your inferences and codebase.) As a test that our plan could work in the first place, we pointed Cursor to OpenRouter and realized we could use its models for the Ask, Agent, and Cmd+K interactions in Cursor. We were also still able to use the normal Cursor Tab completions, which rely on a proprietary model Cursor serves for inline completions (it is very good, so we’re glad to be able to keep it). The solution was to set up a reverse proxy to expose a public endpoint that would forw

匹配字段正文