Most of the push is borrowed.
On 30 September, ESA's Juice spacecraft swings past Earth to steal a little of the planet's momentum on its way to Jupiter. It carries over three tonnes of propellant and still cannot get there on its own. That trade, less fuel in exchange for a hard appointment, is one we make in software every week.
At the end of this month, on 30 September, a spacecraft that left Earth in April 2023 comes back to visit. ESA's Juice probe will pass close by, bend its path against our gravity well, leave with more speed than it arrived with, and carry on towards Jupiter. It does not arrive until July 2031. This is the third of four such manoeuvres, after a lunar-Earth pass in August 2024 that had never been attempted before, and Venus in August 2025. There is one more Earth flyby in January 2029 before the long quiet leg out to the gas giant.
The number worth sitting with is the fuel. Juice launched on an Ariane 5 carrying more than 3,000 kg of propellant, a wet to dry mass ratio of about 2.6 to 1. More than half of what went up was there purely to make the thing move. And it is still not enough. Roughly 600 million kilometres of trip, a full tank by any sane definition, and the mission only closes because it takes four separate loans of momentum from planets that happen to be going the right way.
The thing you did not have to carry
We keep meeting the same shape in the systems we build. A client asks for a booking flow, or a check-in system, or an operational tool that has to sit in the middle of a business that already runs. There is always a version of the brief where we build all of it: our own identity layer, our own payments, our own scheduling, our own reporting. It is buildable. It is also the version where you spend the whole budget on propellant and never get out of the inner system.
The better version finds the momentum that already exists. The venue already has a property management system with the bookings in it. The client already has a payment provider with their reconciliation set up. The staff already carry phones. The door already has a reader on it. Every one of those is a body with its own gravity, and if you can plan a path that passes close enough, you get speed you did not pay for.
What borrowing actually costs
This is where the analogy earns its keep, because a gravity assist is not free. It is cheap in fuel and expensive in freedom.
- 01You have to turn up. A flyby is a fixed appointment with a moving object. Miss the window and the whole plan behind it has to be recomputed. Integrations behave the same way: the third party sets the release schedule, the deprecation date, the maintenance window, and you work around it.
- 02The path is now shaped by something you do not control. Juice goes to Venus because the trajectory needs Venus, not because Venus is interesting on this leg. Plenty of software architecture is shaped the same way, by what an existing system will and will not give you.
- 03The plan gets long. Eight years from launch to arrival, most of it spent going somewhere other than Jupiter. Building on other people's rails often means a slower first release and a much cheaper tenth one.
None of that is an argument against it. ESA made this trade deliberately with a decade of margin analysis behind it, and the alternative was not a faster mission, it was no mission. The point is that the trade is explicit. Somebody wrote down what the borrowed momentum cost in schedule and dependency, and signed it off.
The studio read
Most projects we inherit did not do that. Something was integrated because it was quicker that week, and nobody recorded what it bought or what it committed to. Then a supplier changes an endpoint, the appointment is missed, and the recovery cost lands on a team that never knew it was on a trajectory in the first place.
So the working rule we hold to is dull and it holds up. Borrow aggressively, and write down the loan. For every external system a build leans on, say plainly what capability it provides, what happens the week it is unavailable, and what the exit looks like if the terms change. If that fits on half a page, the dependency is sound. If it cannot be written down, you have not designed a trajectory, you have just hoped.
Juice will pass overhead on 30 September, take what it needs, and say nothing. Five more years and it gets to Jupiter, on a route that was fully understood before it ever left the pad. That is the standard.