Using Manus to Rebuild Manus: Practice and Lessons from an AI-Native Org Overhaul用 Manus 改造 Manus,AI Native 组织改造的实践和思考
用 Manus 改造 Manus,AI Native 组织改造的实践和思考Using Manus to Rebuild Manus: Practice and Lessons from an AI-Native Org Overhaul
用 Manus 改造 Manus,AI Native 组织改造的实践和思考Using Manus to Rebuild Manus: Practice and Lessons from an AI-Native Org Overhaul
Guests: 张涛 · Manus
Manus 联合创始人张涛复盘研发、产品、市场三个团队如何用 Agent 改造自己:一人同时推进十几个需求,产品经理直接提交代码,人均月 AI 开销超 1 万美元Manus co-founder Zhang Tao on how R&D, product and marketing rebuilt themselves around agents: engineers run a dozen-plus tasks at once, PMs ship code, and AI spend topped $10K per person a month
Read the original · mp.weixin.qq.com ↗
Key points
R&D built Manus-mono: not a merged monorepo but one entry point plus a manual written for AI, with Git worktrees, the full stack running locally, permissioned access to production logs and end-to-end checks with test accounts; humans still approve releases.
Engineers paste Slack requests verbatim into rows of Codex and Manus sessions and routinely juggle 10–20 tasks; a 7-person mobile team covers 2 products and 4 apps.
About half the PMs ship features end to end and all have committed code; some requirements now arrive as videos instead of PRDs, gated by an in-house code-review bot.
Marketing’s “digital colleague” triages Slack interruptions on its own, proactively flags project status, and can file feature requests that trigger Manus Automations to extend its own tools.
September AI spend topped $10K per person; Zhang argues it buys time and attention, that using the best models is how you learn their limits, and that costs will fall as the product improves.
Quotes
“One person can build every platform for a single requirement.”
“First, this money doesn’t buy labor; it buys time and attention.”
“In an AI-native organization, people’s role hasn’t shrunk; it has become more concentrated.”