Last Wednesday night, I stared at a blank doc for 2 hours
I've been running my side hustle for over half a year. The thing I fear most isn't "can't make it work"—it's "don't know which one to start first." I have a pile of small projects I want to do—build a membership community, try Xiaohongshu, write an automation flow for a client—but every time I sit down, my brain starts fighting: this one first? Or figure that one out first? Two hours later, I'm still scrolling my phone.
I've made this mistake before—always thinking "figure it out clearly before starting," and after one night of thinking, I'm still stuck in the same spot the next day.
A British engineer built a method for breaking through "fog" projects
In a Latent Space interview, Matt Pocock (a British engineer with 220k stars on GitHub) shared his /wayfinder tool. He said he used to be the same way—dumping everything on AI to run overnight, but the "planning" step was actually the most exhausting: breaking fuzzy ideas into plans, into task lists, and watching whether AI could handle enough in one conversation.
What he wanted to solve was that "unclear starting point" state. /wayfinder works like an AI orchestration layer: you just say "I want to do XX," and it uses multiple approaches to do research, prototype, compare—and finally throws a viable next step back to you.
I tried a similar approach with a Chinese AI assistant—dumped my "want to build a membership community" idea into it, asked it to first list 5 questions I didn't know I should ask, research 3 similar accounts, and give me a minimum validation plan. A 2-hour task squeezed into 20 minutes.
What does it cost to replicate today?
Money: $0. Free tiers or lowest subscriptions of ChatGPT, Claude, Gemini (around $20/month total) are enough.
Time: First run-through takes about 1 hour, mostly learning how to "ask clearly."
Technical barrier: Just typing. The core is learning to articulate your "fuzzy goal" in a paragraph—which is actually a soft skill.
First step: Open your usual AI chat window, type: "I want to do ____, but I don't even know how to take the first step. Please first ask me 5 clarifying questions, then give me 3 minimum validation plans."
How to use this at different stages?
If you're just starting out (no money, no clients): Don't rush to buy tools. First write down what you "want to do"—even 3 sentences—then send it to AI to ask you questions. Free tools are enough.
If you have 1-2 clients: Use AI to break down "who to talk to next, what materials to prepare." Turn the fuzzy "I should get more clients" into concrete "contact 3 people this week."
If you're scaling up: Treat the /wayfinder approach as a "project breakdown layer"—run every new project through AI planning first, then delegate to humans or AI agents for execution. Saves you tons of "meeting time explaining things."
If you don't try today, that's fine—but next time you're stuck on "what should I do first," remember AI isn't just for writing things, it can also help you think through your next step.