Termsslug: computer-use

Computer Use

计算机使用

Agent capability to directly control GUIs and operating systems, a key battleground for desktop agents.

Curated Intelligence

40 items · Page 1/4
The DecoderT2·Agents
78 pts
Google Gemini Adds Background Computer Use on Mac

Google has launched a Computer Use feature in the Gemini Mac app that lets users describe a goal and have Gemini complete multistep tasks across local files, desktop apps, and websites inside an isolated background window, keeping the screen and cursor free. It rolls out in English in the US today, with per-app permissions, live step previews, pause controls, and pre-approval for key actions.

IT之家 · 智能时代T2·Agents
58 pts
Tencent WorkBuddy Adds Computer Use Capability for Desktop Automation

Tencent WorkBuddy has launched a Computer Use capability that recognizes on-screen pages and client interfaces to perform clicks, typing, and scrolling without requiring target software to expose APIs, operating in the background. The feature is in public beta on macOS and Windows, requiring client version 5.11.0 or above and user authorization, with Tencent advising users to stay present during sensitive actions such as login, payment, and deletion.

The Verge AIT2·Agents
68 pts
Google Brings Computer Use to Gemini App on macOS

Google has added computer use to its Gemini app for macOS, letting it complete multi-step tasks across local files, desktop apps, and the web. Users control which apps Gemini can access and can watch a live preview in a background window. The feature is rolling out in English to US users, ahead of Apple's new restrictions on apps requesting full disk access.

TechCrunch AIT2·Agents
62 pts
Google Brings Computer Use to Gemini Desktop App on macOS

Google announced that its Gemini desktop app on macOS now supports computer use, letting the AI operate native Mac apps and browse websites in the background to complete tasks on users' behalf. The feature is initially limited to English-speaking users in the U.S. and requires a Google AI Plus, Pro, or Ultra subscription, arriving later than similar agentic offerings from OpenAI, Anthropic, and Meta.

The DecoderT2·Agents
58 pts
Mistral Launches Vibe Desktop App for Parallel Agent Tasks

Mistral has released a desktop app for its Vibe AI agent (formerly Le Chat) on macOS and Windows, letting users assign multiple tasks at once and watch the agent execute them in parallel. The app merges Vibe CLI's local capabilities with Vibe Code Web's cloud sessions, using Git worktrees to isolate parallel tasks within the same project, and Work mode can now access local files. The report notes Mistral is still catching up, as rivals like Anthropic already offer desktop agents that can operate full operating systems.

Hacker News AIT2·Agents
62 pts
Pomerado Turns Any Website into an MCP Interface for Agents

YC F26 startup Pomerado launched a self-serve product that converts websites lacking APIs into deterministic, self-healing MCP interfaces within minutes. The approach bypasses slow and unreliable browser automation and computer use, targeting the core bottleneck of connecting legacy airline, banking and government systems to agent workflows.

宝玉 (@dotey)T2·Tools
58 pts
Developer Flags GPUI Pitfalls, Switches Back to Electron for Agent-Friendly Workflow

Developer Baoyu reviewed his experience building the baocut desktop app on GPUI, noting that beyond performance and cross-platform strengths, the Rust-native UI framework requires forking for complex UIs, has weak accessibility support, heavy compile resource and time costs, no hot reload, and costly agent-driven UI verification. He ultimately returned to Electron while keeping complex rendering and transcription/TTS modules as Rust/WASM and worker processes to balance iteration speed with backend performance.

Hacker News AIT2·Models
72 pts
Tencent Unveils Hy4 Open-Source LLM with Vision and Computer-Use Capabilities

Tencent announced Hy4, its latest open-source large language model, claiming performance that ranks among the world's leading open models. Hy4 adds visual understanding and computer-use capabilities for the first time, alongside gains in instruction following, long-context comprehension and long-horizon agentic tasks. Model weights and system cards are slated for release in the coming days, with Tencent emphasizing responsible open-source development around cybersecurity and biological safety.

Yuchen Jin (@Yuchenj_UW)T2·Agents
38 pts
Developer Argues Codex's Computer Use Beats Claude Code

Developer Yuchen Jin argues that Codex's key edge over Claude Code is computer use: APIs and MCPs are slow, and legacy enterprise UIs are hard for humans to navigate. Codex operated a confusing expense-reimbursement app and finished the task in minutes, effectively turning any UI into an API. The claim rests on personal experience without quantitative evidence.

The DecoderT2·Models
74 pts
Tencent Releases Hy4 Model with Vision and Computer-Use Capabilities

Tencent has unveiled Hy4, its new flagship model that for the first time understands images and can operate software autonomously, with targeted optimization for coding, finance and office work. Tencent claims second place among Chinese models, but independent tests show it trails top systems like Claude Opus 5.5 on hard math and coding tasks; model weights will be released later.