Software is following the path coal once did. When a thing gets cheaper, we don’t use less of it. We use a great deal more. The cost of a working line of code, written by a person sitting at a keyboard, has been falling for two years now at the speed of model improvement. The amount of software being written is not falling with it. The opposite. We are in the early innings of what will, in retrospect, look like an explosion.
The question that follows is not whether engineers still matter. They do. The question is what they do, and where they sit in the build. We think about this question a lot. We started Relay because the answer surprised us.
An engineer’s job is no longer to write the line. It is to be the person on the other end of a press.
For most of software’s history, an engineer wrote the line. The line was the unit of value. A junior engineer wrote a line; a senior engineer wrote a better one; a staff engineer reviewed the lines of others and made them coherent. The pyramid was tall and the work was vertical.
That pyramid is collapsing into a plane. Today a marketing manager in Dallas, a clinical-operations lead in Seattle, an analyst in Mumbai , anyone with a tool and an idea, is producing code. They are not engineers, and they will not become engineers, and the software they produce is good enough to use. Most of it ships in some form. Some of it ships in a form that should not have shipped.
The work that distinguishes those two outcomes is not the writing of the line. It is the press of a button. It is the moment a person who has shipped this kind of system before walks into the room, looks at the diff, and says one of three things. This is fine, ship it. Or this will break here, change this. Or do not ship this; I will rewrite it with you. The decision is small. The compounding effect, across thousands of builds inside a single company, is not.
That is the work an engineer does now. Not the line. The decision. The relay.
The category we are building is the category of that decision. We don’t think it has a good name yet. Engineer-as-a-service is too transactional. AI co-pilot is taken and points the wrong direction. Pair programmingis too symmetric. The work is asymmetric: the AI runs eighty percent of the build, the person runs the moment that decides whether the build ships. We’re calling it the press. The dot in the corner. The relay.
This is a long way of saying the obvious thing. Software engineering is not going away. It is moving. It is moving from the inside of a single company’s payroll to the outside of every company’s build session. It is moving from a salary line to a press. It is becoming, for the first time in the history of the craft, available the same way electricity is available. You press, it arrives, it does the thing only it can do, and then it leaves, and the build goes on.
That’s what we’re building.
The founders, May 2026