我让AI替我运营创业公司一周
做法与步骤
- 作者使用Grok Bot组建了一个迷你代理团队：支持机器人用Gmail回邮件、工程师机器人写代码、安全机器人监控Cloudflare和服务器日志。
- 最让作者兴奋的是机器人之间能协作：它们在没有作者参与的情况下，一起为客户修复了一个bug，让琐事自动化，从而专注于更大的事。
- 移动办公体验被颠覆：作者在威尔士5小时徒步中通过手机与机器人沟通完成任务，效率超过坐在电脑前9小时，意识到投入时间与产出并不相关。
- 作为单人创业公司，作者用营销机器人以每周增长PromptBase社交影响力2%为目标，开设TikTok、YouTube Shorts等渠道并实验加倍投入有效方法。
- 作者进行了一次奇怪实验：让项目经理机器人在凌晨4点自动部署一个秘密功能，机器人选择上线账户超时预警功能，效果不错，但醒来后感觉既酷又奇怪。
- 主要缺点：设置和连接耗时且数据被锁定；机器人不会主动提出疯狂的增长创意，仍需人类想点子；人类可能逐渐失去对业务的理解，需警惕被机器人牵着走。
对实践的启示
对AI应用开发者而言，代理协作和移动化工作流已可落地，能释放单人生产力；但需权衡数据锁定风险，并保持对业务的理解与创意主导权，避免被自动化反噬。
作者承认的反方与局限
作者承认设置耗时、数据被锁定，机器人缺乏疯狂创意，且人类可能失去对业务的理解，需要小心避免被机器人掌控。
原文摘引
It's made me actually believe for the first time that a 1-man startup could actually compete with a much bigger one (not just hype).
It made me realise how effort-in does not correlate to output, and made me think what work might be like when you don't have to "work" as much in a weird AI future.
It can sometimes feel as a human you lose understanding of what's going on in your business, you need to be careful to stay on top of this so the bots don't start running you.
▸ 展开原文全文（34 段，抓取自 Hacker News）
Wednesday, October 14th, 2026
I completely missed the OpenClaw hype - I actually thought it looked kind of dumb at the time.
What is this for?
But the past week or so I've been quite enjoying using Grok Bot - which is very similar (and like OpenAI dots I guess).
It's basically a mini team of agent/bots that run on a Linux server somewhere that you can assign jobs to.
Eg, I have a support bot that uses my Gmail to answer emails, an engineer bot that writes code, and a security bot that monitors my Cloudflare and server logs for suspicious things.
I don't want to delve into the weeds of it too much here but I thought it would be cool to jot down why I'm quite intrigued by this as a solo entrepreneur/developer.
1. Working together
Grok Bot clicked for me when they worked together and fixed a bug for a customer without me (texts to my bro):
Having these bots interact with each other like this is really cool - this is the most interesting thing for me. Trivial things that you sometimes put off (e.g. fixing small bugs, or following up an email) become automatable so you can focus on bigger stuff.
2. Mobile
The idea of coding/working from my phone seemed awful to me. I dislike my phone in general. But last week on a 5 hour hike in the middle of nowhere in Wales, I was communicating with these bots to do stuff for me.
I had a really enjoyable day, and got more work done than some days where I'm sat in front of my laptop for 9 hours.
It made me realise how effort-in does not correlate to output, and made me think what work might be like when you don't have to "work" as much in a weird AI future.
3. Trying new stuff
As a 1 man startup - I've always been limited with what I can do. Things like content marketing & trying paid ads have always seemed out of reach because I don't have enough time.
I now have a marketing bot whose sole goal is to grow PromptBase's social reach 2% per week.
It's started a TikTok profile, YouTube shorts channel and more. I've told it to experiment and double down on what works. It seems to kinda be working!
I would never have time in my day for this usually, and it makes me want to try more creative things like startup vlogs alongside this blog.
It's made me actually believe for the first time that a 1-man startup could actually compete with a much bigger one (not just hype). And being a 1-man startup might actually be an advantage because there's less human overhead.
4. Weird experiments
A few nights ago I instructed my project manager bot to ship one secret feature to PromptBase at 4am and deploy it whilst I was asleep - it just had to benefit the site in some way.
I wanted to see what this felt like - which sounds weird - I know.
It decided to ship a feature that alerts users before their account is about to get timed out from too many declined submissions - a pretty useful idea based on a lot of emails from support.
It went well, and worked! - waking up and seeing it felt very cool, but also a bit strange, like PromptBase could just be run by an AI agent if I really wanted.
5. Bad things
So far Grok Bot has been pretty good, the main downsides I've found are:
- It takes a while to set up and connect everything - and I'm quite locked in and have handed over a lot of data; I think that's why these AI labs are pushing these products so hard right now.
- I was hoping these bots would have some crazy moonshot ideas on how to grow PromptBase given all the context they have - so far that has not really happened - you still need to come up with ideas.
- It can sometimes feel as a human you lose understanding of what's going on in your business, you need to be careful to stay on top of this so the bots don't start running you.
6. Conclusion
This is a long post, my hand hurts from writing. I'd definitely recommend experimenting with this if you've not tried these kind of products before.
If anything, it's maybe given me a bit of an insight into the hype around always-on Agents and where things are going.
'Til next week!