People sit around for years talking about starting because they begin with a five-year plan.
If you are at zero, ask what will exist three days from now. Then one week. Then one month.
Your first customer will change the plan anyway. They will want a different problem solved, refuse the feature you loved, or pay for the part you considered secondary.
Long-term thinking becomes useful when present-day evidence exists. Before that, it often becomes a sophisticated way to avoid being judged by reality.
Make the company small enough to start. Make the next commitment short enough to finish. Let revenue earn the right to extend the horizon.
You do not need to imagine the billion-dollar version before you have made the first rupee. Fuck it. Try the small version.