我与 AI 协作的瓶颈：如何拆解任务并并行化工作流
My Bottleneck with AI
风险投资人 Tomer Tunguz 分享了自己使用 AI 编程的工作流变化：先用 ChatPRD 生成 PRD，再让 AI 制定实现计划、先写测试再逐组件验证，之后可放手让 AI 自主迭代 10-15 分钟。他认为当前瓶颈在于如何把想法和问题结构化，以便并行管理 10 个浏览器标签或终端窗口。在编程之外，他也开始优先问 AI 而非人类专家，但会因 AI 输出可疑而回退到 Google 搜索。
In short : Optimize AI workflows by breaking down projects into smaller tasks and ensuring structured plans, tests, and implementation.
I’ve been watching my behavior evolve as I use AI more.
Coding new projects, I started first by describing a big idea : generate a new website that ingests podcasts and transcribes them. But I quickly discovered this approach was like asking someone to build a skyscraper without blueprints—the AI choked on requests too broad & lacking structure.
So I broke the project down into smaller blocks. Write a function to download the list of podcasts in the last week from these sites. But I noticed how low-leverage that felt : I knew I was yoking a strong animal to pull along a child’s wagon.
Today, I use AI to write a basic PRD, ChatPRD is great for this. Then I ask the AI to read it and create an implementation plan. I approve the plan and then insist that the AI write tests first for each function and then test each component.
At this point, I can let the AI run wild for 10-15 minutes. With the plans and test cases, it can iterate to success itself.
As I watch the code cascade through the window, I have the same feeling of being driven by a Waymo. Enjoying the journey much more now that the toil is managed by a computer.
Top of mind for me now : how do I parallelize this so I can manage 10 browser tabs or terminal windows simultaneously? My limiting factor is structuring ideas & questions in a way to succeed with the AI working.
Outside of coding, I’m seeing myself default to AI search & reverting to Google.com when I suspect something is fishy in the AI’s output.
This is now also true beyond online searches : rather than asking a human expert, I’m experimenting with asking an AI first. Why haven’t equities moved onto the blockchain in the same way debt has?
The rub in this effort is knowing when to trust the AI. Like any human, the longer an AI bloviates, the less I believe it.
My deeper concern is what nuances has the AI excised? What am I missing from its summary?
How have you changed how you work with AI?
Get the next one in your inbox
The 1-minute read that turns tech data into strategic advantage.
Read by 150k+ founders & operators.
GP at Theory Ventures. Former Google PM. Sharing data-driven insights on AI, web3, & venture capital.
来源：Tomer Tunguz 博客 · tomtunguz.com