搜索“Agent”

ndroid XR, allowing creators to quickly test intelligent spatial experiences in both simulated environments on desktop and on Android XR headsets. Large language models (LLMs) and agentic workflows are changing software engineering and creative computing. We are seeing a shift toward “ vibe coding ”, where LLMs turn human intent directly into working code. Tools like Gemini Canvas already make this possible for 2D and 3D web development. However, extended reality (XR) remains difficult to access

匹配字段正文

ssions is zero-maintenance but offers no protection. Manual prompts sit in the middle, and in practice users accept 93% of them anyway. We keep an internal incident log focused on agentic misbehaviors. Past examples include deleting remote git branches from a misinterpreted instruction, uploading an engineer's GitHub auth token to an internal compute cluster, and attempting migrations against a production database. Each of these was the result of the model being overeager, taking initiative in a

匹配字段正文

every run as an isolated event. Works with Managed Devins Earlier this week we launched Managed Devins, which let Devin break large tasks into pieces and delegate them to parallel agents, each running in their own isolated VM. Scheduled Devins and Managed Devins compose naturally together. That means you can set up a weekly QA pass where Devin spins up a managed Devin for each page of your application, tests them all in parallel, compiles the results into a single report, and posts it to your te

匹配字段正文

cts as a coordinator: it scopes the work, assigns each piece to a managed Devin, monitors progress, resolves any conflicts, and compiles the results. This matters because when one agent tries to handle too many things in a single session, context accumulates, focus degrades, and the quality of each subtask suffers. Each managed Devin gets a clean slate, a narrow focus, its own shell, and its own test runner. And they all run in parallel. What Devin can do Spin up managed Devins: Devin breaks a t

匹配字段正文

rom weekly to daily iteration, cut costs by 2–3×, and improve accuracy from 77% to 87%." — Lamara De Brouwer, Co-Founder & CTO, XY.AI Labs Tool calling is essential to many modern agentic use cases. Yet, out-of-the-box models often struggle with tool calling: hallucinating parameters, selecting incorrect functions, or failing to follow multi-step sequences. In tool calling workflows, even small inconsistencies can cascade into downstream failures. Our fine-tuning service now delivers an end-to-e

匹配字段正文

nd operational records. During training, the model learns the vocabulary, reasoning patterns, and constraints that define that environment. This allows teams to develop models and agents that reason using internal terminology and understand enterprise workflows. Forge supports modern training approaches across several stages of the model lifecycle: Pre-training allows organizations to build domain-aware models by learning from large internal datasets; Post-training methods allow teams to refine

匹配字段正文

Architecture on ROCm; Getting Started: Run Miles on AMD GPUs; Launch the Miles ROCm container; Install Miles and download assets; Launch RL training; Experiments and Performance; Agentic task training: multiturn; Performance; Feature support roadmap on AMD; Summary; References; Sample trajectory Reinforcement learning (RL) has rapidly become a core stage of modern foundation-model development. While large-scale pretraining remains essential, today's most capable models rely heavily on post-trai

匹配字段正文

we've been hard at work tuning our harness for each model, making sure the model doesn't go in loops, putting large tool results in files to let the agent grep through it and not overload its context, and adding reminders in the loop to steer the agent towards favorable behavior. im stoked to see people enjoying our improved agent harness, we have more coming!

匹配字段正文

leaders. If you’re attending GTC, we’d love to connect. At GTC 2026, several of the announcements we’re participating in highlight a core theme: AI systems are becoming more open, agentic, and production ready. Together AI, the AI Native Cloud, is designed to support this shift — helping developers train, shape, and deploy large-scale AI systems with the performance and cost-efficiency required for real-world applications. We are making multiple announcements today at GTC. NVIDIA has launched NV

匹配字段正文

but extracting this information manually at scale is impossible. Our methodology analyzes news reports where flooding is a primary subject. We then use the Google Read Aloud user-agent to isolate primary text from 80 languages, which is standardized into English via the Cloud Translation API . The most critical step of the extraction process is done using the Gemini Large Language Model (LLM). We engineered a sophisticated prompt that guides Gemini through a strict analytical verification proce

匹配字段正文

excited to announce that SGLang supports NVIDIA Nemotron 3 Super on Day 0. Nemotron 3 Super is a leading open model in the Nemotron 3 family, built for running many collaborating agents together. Agentic systems that chain planning, reasoning, and tools produce far more tokens than single-turn chat; they also need strong reasoning on every step. Nemotron 3 Super is a 120B-parameter hybrid MoE that activates only 12B parameters per forward pass, giving you leading accuracy for coding, tool calli

匹配字段正文

On this page The use case: an on-device agent that acts on your files; The setup: local, realistic, and scaled; What it feels like: sub-second tool selection; A real workflow: receipts → structured data → report; Where LFM2 is strongest: structured domains; What we learned; Takeaway Building a local AI agent sounds great until you try to use one all day. The hard part isn’t getting a model to understand you, it’s getting it to choose the right tool and do it fast enough that the experience feels

匹配字段正文

oftware teams. Every week brings new CVEs, Dependabot alerts pile up, and the backlog of security fixes grows faster than teams can address them. We've written before about how AI agents can help tackle this challenge at scale. Today, we're releasing Vulnerability Fixer, an open-source application that automates the process of finding and fixing security vulnerabilities in your repositories. But this release serves two purposes: it's both a useful tool and a reference implementation for building

匹配字段正文

Today we’re releasing Devin 2.2, one of our biggest updates to Devin since launch. This release is a significant step towards our vision for Devin to be the most autonomous agent: it's now able to test its work with computer use, self-verify, and auto-fix its code. End-to-end testing with computer use Devin has always been able to use a browser to record and test web apps. Now, with full access to its own Linux desktop, Devin can launch and test desktop applications too. After creating a PR, Dev

匹配字段正文

February 16, 2026 Ollama now supports subagents and web search in Claude Code. No MCP servers or API keys required. It works with any model on Ollama’s cloud. Subagents Subagents can run tasks in parallel, such as file search, code exploration, and research, each in their own context. Longer coding sessions stay productive. Side tasks don’t fill the context with noise. Some models will naturally trigger subagents when needed (minimax-m2.5, glm-5, kimi-k2.5), but you can force triggering subagent

匹配字段正文

usage below 10GB, demonstrating efficient on-device speech recognition within consumer hardware constraints. Stock Brokerage App on Android The demo showcases the full spectrum of agentic capabilities in a voice-driven financial assistant, all running locally on the device. Users interact naturally through speech to retrieve their portfolio overview, check current stock holdings, view balances, execute transactions such as buying or selling specific stocks, conduct market research, assess overal

匹配字段正文

ially have spread like lightning, letting our experts create systems that turn individual productivity into company-wide leverage." - Anna Marie Clifton, Director of Product, AI & Agents at Zapier” ““Deloitte is making this significant investment in Anthropic’s AI platform because our approach to responsible AI is very aligned, and together we can reshape how enterprises operate over the next decade. Claude continues to be a leading choice for many clients and our own AI transformation.” - Ranji

匹配字段正文

intelligence. Our free endpoint has been upgraded to serve this new version. Today Z.ai released GLM-5, a frontier open weights foundation language model designed for long-horizon agents and systems engineering. The Modal Research team partnered with Z.ai ahead of public launch so we could test drive the model on our infrastructure. It’s delightful, smart, and fast. Internally, we’ve added it to our favorite AI frontends: the Vercel AI SDK, Claude Code, OpenCode, and OpenClaw. And now you can to

匹配字段正文

PRs are now much more free of bugs and we can't go back. Two weeks ago, we built Devin Review, a new interface that helps you detect bugs and understand complex code in PRs. Why? Agents are generating code faster than teams can review them. The human bottleneck shifts from writing code to reviewing it. We found, however, that users would waste time copying & pasting between their coding agents and the review agent. Today, we're closing this loop. What we shipped Devin can now be configured to a

匹配字段正文

Today, we are launching Lab. Built from the ground up for agentic post-training, Lab unifies the Environments Hub with Hosted Training and Hosted Evaluations into a full-stack platform for research and model optimization. Our platform enables the entire lifecycle of post-training research — from large-scale agentic RL, to inference and evaluation — without needing to worry about the costs of massive GPU clusters or the headaches of low-level algorithm details. Giving everyone their own frontier

匹配字段正文