Writing a first prompt
Describe the problem, not the solution. This is the single highest-leverage thing you can do, and it takes one sentence.
Why the solution is the wrong thing to say
When you describe a solution, you have already made every interesting decision — and you made them before anyone examined the problem. The agent will build what you asked for, which means the cut list has nothing to work with and the expensive mistakes survive into production.
When you describe a problem, the plan can propose something cheaper than what you had in mind. That happens often enough to be the main reason this stage exists.
Three pairs
Build me a DeFi platform with staking and governance.
My eleven-person running club keeps forgetting to pay subs and I hate chasing them.
The first names a category. The second names a problem, and a problem can be argued with.
An NFT collection with utility.
I make 500 units a month and buyers keep asking how they know it is really mine.
One of these has a test for whether it worked. The other has a launch date.
A social app for traders, like Twitter but on-chain.
My group chat argues about calls and nobody remembers who was right.
Naming a comparable product imports its whole feature list. Naming the friction does not.
Things worth adding
One sentence is enough to start, but three details reliably improve the plan:
- Who it is for, and how many of them. Eleven people and eleven thousand people are different products, and the plan should know which one it is scoping.
- What currently happens instead.The workaround people already use tells the agent what “better” has to beat.
- What you would consider a failure. This is the one people skip, and it is the one that shapes the cut list most.
Things not worth adding
- A stack. If you name Solidity, you will get Solidity, even when the plan would otherwise have argued the chain out.
- A competitor. Naming one imports its entire feature list into your scope by implication.
- A deadline, at this stage. It changes the cut list for reasons that have nothing to do with the problem.
Next: Working with the plan.