Onlook founder Daniel announced a rebuilt, Apache-2.0-licensed desktop editor that places an infinite canvas over a real codebase. The app runs Claude Code or Codex on the user’s subscription, creates Git worktrees, and represents changes as reviewable pull requests; the team also plans a paid cloud product for teams.
JD Cloud launched an upgraded JoyCode intelligent coding tool centered on end-to-end autonomous decision-making. It combines project context, CodeGraph, CodeWiki, and an Agent Loop for planning, execution, verification, and iteration, while adding multi-agent collaboration, configurable models, and reusable tools and workflows. JD Cloud claims JoyCode serves more than 10,000 internal developers, contributes over 80% of AI-generated code, and helps shorten the development cycle by about 40%.
掘金 · 人工智能 describes an AI-SDLC engineering structure for letting coding agents work across existing Git repositories without consolidating them into a monorepo. The author’s team starts agents in a shared main repository, uses a manifest to define repository access and commands, and creates isolated worktrees for writable repositories. The account also covers disk-usage and unmerged-change risks, unified commands for Claude Code and Codex, and stricter cleanup rules for worktrees.
InfoQ 中文 AI&大模型 translated a podcast interview in which chief architect Carlos Kelkboom told host Patrick Akil five counterintuitive rules for agentic coding: build the code skeleton yourself, bar the main agent from editing files directly, discard generated code you cannot recognize, let agents finish larger tasks in one pass, and run several agents in parallel then merge only the good parts. He described familiar failures, including agents quietly replacing deployment scripts, the same feature implemented five times, and patching rather than restructuring when architecture needs to change. The two also discussed shrinking PR debate, how AI-generated code should be reviewed, and what happens to engineering judgment and junior-developer training as code gets cheaper.
What problem it solves
Teams using coding agents need a shared way to access conversations, dispatch work, and follow activity across machines and surfaces. Screenshots, pasted logs, and separate agent sessions make context difficult to preserve and coordinate.
Key features
Shared agent conversations — Open the same conversation with teammates and follow its transcript, runtime status, files, and …
What problem it solves
Developers using coding agents in the terminal often need to run multiple tasks at once, which requires manually juggling terminal tabs, branches, and repository checkouts. Paradev provides a shared workspace for isolating tasks, reviewing changes, and previewing applications without managing those details by hand.
Key features
Parallel agent workspaces — Assign each task it…
What problem it solves
Coding agents need isolated places to work without disturbing a checkout, while developers need to steer concurrent sessions and review each change. They also need durable project context that survives beyond a single chat session, along with tools and workflows that fit how they work.
Key features
Isolated git worktree workspaces — Open a git repository as a project and cre…
What problem it solves
Git worktrees give parallel AI agents separate working directories, but Git's native worktree workflow requires repetitive commands and branch-name entry. Worktrunk simplifies creating, switching, inspecting, merging, and removing worktrees while adding automation for multi-worktree development.
Key features
Branch-oriented worktree management — Worktrees are addressed by br…