
Until – You approve the plan, a separate agent checks the code against it
You and your agent draft a Plan: the outcome, the approach, and what's out of scope.
以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。
You and your agent draft a Plan: the outcome, the approach, and what's out of scope.
Send the Plan to a teammate, who approves it or sends it back with comments. This is where your attention goes, because it's where the decisions are.
Approval releases the Jobs. Your agents claim them, leave handoff notes as they go, and ask you when a call is yours to make.
An agent review compares the code with the approved Plan and lists every place they differ.
When a decision is yours, the agent asks you instead of guessing.
Any agent. Use the agent you already use.
Agents will happily make every decision for you, and each one they make is one you stop understanding. We're inviting a small group of builders to use Until on real projects before launch.
Put your taste into the Plan; the agent handles the rest. You keep the judgment and the craft.
Until runs on your machine, in your own Git repo. There's no Until server, no account and no telemetry.
Use the agent you already use, such as Claude Code or Cursor. Any agent that can run a command works.
We're picking a small group by hand. Seven questions, about two minutes.
正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。