Will Binns
@wbnns
Building mru.space • ran bitcoin.org, led teams @coinbase • editor of Mastering Bitcoin & Mastering Ethereum • software that outlives its makers
You cannot write down every case
So you write down enough of them, plus what to do when none of them match
That second part is most of the work
0
0
0
Somewhere ahead is a machine that runs into trouble four light years from anyone and simply handles it
Not by waiting. By deciding. This is software we can write now
0
1
6
The interesting question is not how to keep a machine working for centuries
It is how to keep it the same machine while everything about it changes

Mru Aerospace
@mruspace
1d
A craft sent to another star is alone for centuries. It has to fix and adjust itself, because nobody else can.
That sounds like a problem and it is a solvable one. You let it change anything about how it works, and you put the one thing that matters somewhere it cannot reach.
0
0
1
We are sending our best current understanding of the universe to another star, and the only thing it will ever report back is where we got it wrong

Mru Aerospace
@mruspace
2d
A craft four light years out can send almost nothing home. A full image is out of the question. The power is not there.
So it does not send what it saw. It sends the difference between what it saw and what it expected to see.
It has to know what to expect. That expectation is
0
0
0
We have always been the ones who knew. The instrument reported, we decided what it meant
Somewhere ahead that order reverses for good

Mru Aerospace
@mruspace
2d
A craft between stars finds its position by looking at stars it recognises. When radiation finally kills the camera that does that, it cannot ask anyone where it is.
So it keeps going on arithmetic. Last known position, last known speed, the laws of physics, forward.
Deciding
0
0
0
The hard-won part is already in people who have spent careers keeping machines alive at a distance
What does not exist is anything that can carry it where they cannot go
That is the part we are building
0
0
0
Voyager has flown 49 years, but only because a team on Earth still keeps it alive
So I simulated 10,000 spacecraft where parts break and nobody can fix them. Some wait for Earth to tell them what to do. Some decide for themselves
I expected the gap to hinge on how fast Earth replies, but it doesn't... 1 day or 6 mont...
1
1
9
We are deciding, today, what a machine should care about in a place no human will ever stand
Every generation thinks it knows what is worth looking for
Ours gets to hand that guess to a machine and point it at another star
0
0
1
We are writing instructions for a situation we cannot describe
Not a hard version of a known problem
A category of problem nobody has had

Mru Aerospace
@mruspace
4d
Far past anything we have sent, a machine will run into a situation nobody thought of. It will work out what to do using rules written centuries earlier by people who are long dead. It will act. Then it will write down what happened, for a reader who may never exist.
That moment
0
0
0
We are not writing software to control a spacecraft
We are writing down what the people who would have controlled it know

Mru Aerospace
@mruspace
4d
A NASA mission runs on people. Controllers who send up the day's commands. Engineers who review anything that looks wrong. A flight director who makes the call when something breaks.
A craft sent to another star has none of them, and never will. It is its own operations team,
0
1
6
This is not really a spacecraft question
Anything that can modify itself and cannot be reached has the same problem
We are going to build a lot of those

Mru Aerospace
@mruspace
5d
A craft sent to another star is alone for centuries. Nobody can send it a new plan, so it has to be able to change how it works on its own.
But anything that can rewrite itself can also rewrite its reason for being there. So the goals are burned into a chip that no software can
0
1
4
Every long journey we have ever sent has been on a leash
Voyager is on one right now
Somewhere ahead is the first thing we let go of entirely, and that moment is a software problem, not a rocket problem
0
0
2
Picture a craft sent to another star... it arrives in 400 years, long after everyone who built it
You can send a signal, but you can't have a conversation because an answer takes 8 years to come back
And whatever it does when it gets there was decided by software written today
2
3
17
