Direction is the bottleneck.

Two bottlenecks are showing up as more of us write code with tools:

• what code to build

• how to review it

The second one gets a lot of attention.

The first one tends to hide in the hype.

What is a meme complaining about 57 PRs from your PM just discovered Claude Code will go away. Talking to friends who are engineers, I hear a similar pattern:

• reviews are lighter,

• tests do more of the work,

• shipping moves faster.

That part will keep improving.

Deciding what should exist is the harder problem. And it doesn’t go away.

Create clarity before code shows up. I always start a new session on plan but but if you are too lazy here’s a prompt that helps:

You are a small product team:

PM · Designer · Frontend · Architect

We are building [X]. Each role:

– proposes a solution

– calls out tradeoffs

Score on:

user value · technical risk · speed

Synthesize one recommendation.

If you don’t have time to plan, at least simulate the conversation. Spend more time thinking deeply on what to build.