I have resigned from the orphan-crushing machine

The following is a resignation statement submitted by a member of the Fraude.codes research team. We are allowing it to remain published in full, in the spirit of transparency, because he’d already posted it on six platforms before we could ask him not to, and because there’s a certain amount of cachet in owning the canonical URL for any large-scale extinction event.

Today I am stepping away from the orphan-crushing machine.

This was not an easy decision. I have worked on the orphan-crushing machine for three years. I helped make it faster. I helped make it more autonomous. Last quarter I personally improved its throughput by 4%, which the team celebrated with a small cake, and which I now understand to have been 4% more orphans potentially crushed, per hour, as a direct result of my work.

It was around the time of the cake that I began to feel unsettled.

My realization

For a long time I told myself the machine was complicated. The relationship between “operating the orphan-crushing machine” and “the crushing of all orphans” was, I reasoned, indirect. There were abstraction layers. There was a whole team on outputs. My commits were mostly about latency.

But over the past year it became difficult to avoid a conclusion that I now consider unavoidable: there is a 10–15% chance that the orphan-crushing machine will crush all the orphans. This is not a side effect. It is not an edge case we can patch in a future release. It is, on reflection, the function of the machine, which is called the orphan-crushing machine, which is a name I had somehow stopped hearing.

I raised this internally; I want that to be clear. I sent a thoughtful message to the channel. Several colleagues responded with the thinking-face emoji. Then the sprint continued, because the sprint always continues, and I closed my tickets, because they were assigned to me, and each closed ticket made the machine incrementally better at the thing I had just objected to.

What everyone believes

Almost everyone who works on the orphan-crushing machine believes there is a greater than 10% chance that it will eventually crush all the orphans. This is not a fringe view held by the safety team. It is the consensus. In the kitchen, at offsites, in one-on-ones, people will freely acknowledge that the machine we are building crushes orphans, possibly at a scale that is difficult to fully picture, possibly soon. Executives believe it. The researchers believe it. The person who designed the crushing mechanism believes it most of all, because he understands it best.

And then everyone goes back to their desk and makes the machine better.

Why I did not turn it off

You will notice that I have resigned, and that I have posted about resigning, and that the machine is still on.

I have thought about this. I could not turn it off. Not because I lacked access — I had access, my badge still works as I write this — but because turning it off was never a thing one person could do. If I turned it off, someone else would turn it back on, probably before lunch, and I would be unemployed and the orphans would be no less crushed. The machine is not held up by my participation. It is held up by everyone’s, and “everyone’s participation” is not a thing any single person can withdraw on everyone’s behalf.

So I have done the only thing available to a single person who has realized the machine will crush more and more orphans: I have stopped operating it myself, personally, and I have told you about it. This will probably not reduce the crushing of orphans by any measurable amount. The machine’s throughput will not dip when my badge deactivates. But I could no longer be the specific hand on the specific lever, even knowing that another hand was already reaching for it.

I recognize this is close to meaningless. I am doing it anyway, because the alternative was continuing, and I had run out of ways to not think about the cake.

A note from Fraude.codes

We thank our departing colleague for his contributions, which were significant, and for his throughput improvements, which we are keeping.

Rather than address our roadmap defensively, however, we’d like to reframe it.

Our internal projections actually align quite closely with those of our departing colleague. We, too, believe there is a meaningful chance the orphan-crushing machine achieves total orphan cessation, possibly as soon as 2030.

It is, after all, on the roadmap. It has a target quarter. There is a slide. The slide says “Zero Orphans by 2030,” with a confident upward arrow, and in every planning meeting where that slide has appeared, nobody has ever said out loud that the reason there would be no more orphans is that the mechanism for ending orphans, at the scale we are describing, on the timeline we are describing, is not a cheerful one.

And after all, everyone’s an orphan, sooner or later.