Think with V command: /goal
explainer
/goal

Set a finish line, and let it run. Claude keeps working until it crosses it. No prompting each step. No babysitting each turn.

fig. 01 · the mechanism

The loop, playable

Try it —
the finish line · you set this
🎯step 1Set the finish linethe one thing you do
⚙️step 2Claude worksno check-ins
🔍step 3A faster model checksreads the transcript
🔔step 4You get pingedonly when it's done
turn 0 Pick a goal, then hit run.

Miss? It redoes the turn and checks again. You never step in. The check reads the transcript of the work, not the model's own memory.

fig. 02 · the recipe

What every /goal needs

Three parts. Miss one and it runs forever, or stops wrong. Pick a goal to see all three.

Try it — pick a goal:
1

A measurable end state

What does done look like? Make it countable.

for this goal
2

A stated check

How will Claude prove it? Tell it what to show you.

for this goal
3

Constraints

What's off-limits? Set the guardrails.

for this goal
fig. 03 · in practice

Some pointers

✅ good use cases
  • Apply a brand guide, every rule
  • Build a slide deck from your notes
  • Update a vibe-coded app
  • Make a landing page from a brief
  • Clean up a messy spreadsheet
⚠️ what to expect
  • You can use it with auto mode
  • It doesn't work with plan mode
  • Expect it to run for several minutes
  • It can be token-heavy — heads up
fig. 04 · don't skip this

Human at the helm

Whatever it delivers, check it. You're at the helm, not just in the loop. Make sure it matches what you pictured.

audit it TL;DR

It's not perfect. Read it like a teammate's work.

free · the cheat sheet (pdf)

Get the /goal cheat sheet

Drop your email and the PDF is yours. You'll get the newsletter too.

The newsletter

Want more?

Experiments and tips for thinking and working better with AI. Free.

No spam. Unsubscribe anytime.