I recently built and deployed an entire project from scratch in under a week!
A few years back, that sentence would have sounded unreasonable. Lean Launch Mate (LLMate) was built with an early version of Claude, and another piece of software I am preparing to launch has taken me a couple of days, not months.
Underneath the excitement is a much harder feeling: if the work no longer has to pass through my hands, what exactly is my value in making it?
The work used to feel like ours
Software engineering has always had a strange relationship with craft. We do not shape wood or stitch fabric, but we still talk about building things. We take pride in a clean implementation, a difficult bug understood, or a system that holds together because we made hundreds of small decisions correctly.
The work felt handmade even when the material was abstract.
- AI changes that relationship
- A growing part of the execution now happens somewhere in the background
- We describe, direct, review, and correct
- We can watch hours of familiar work collapse into minutes
The result may be good, but it can leave us with the uneasy sense that it is not hands-on enough to count as ours.
That is the psychological shift I hear when I speak with other engineers.
Look at what I can build now. Look at what can be built without me.
Most of us are living in both places at once: quantum feelings, both true at the same time.
Effort and value were tangled together
For a long time, it was easy to connect effort with value:
- A feature took several days, so the skill required to build it was visible
- We wrote the code, traced the failures, and carried the implementation in our heads
- The labour itself became evidence of expertise
When AI produces the same visible result quickly, that connection breaks. The speed is useful, but it can also feel gut-wrenching. If something that once required years of practice can now be produced through a conversation, were those years worth what we thought they were?
I do not think this is only fear of losing a job. It is also grief for a familiar way of recognising ourselves.
We became engineers by learning how to do the work. Now we are being asked to stay engineers while doing less of it by hand.
The machine did not decide what good meant
When I look honestly at those fast builds, though, I do not see a machine working alone.
The tools could execute because I had already formed a point of view. I brought a strategic plan, chose what the product should do, broke the system into parts, set constraints, and kept judging the output. I knew when something was technically correct but wrong for the product. I knew what to remove and what needed another pass.
Without that direction, the software might still have been built. It might even have looked impressive. But it would not have been built with the same coherence or taste.
Taste here is not decoration, and it is not a vague creative instinct. It is the accumulated ability to notice what does not belong, choose between reasonable options, and understand the consequence of a decision before it spreads through the system.
AI can make an option. It cannot relieve us of the responsibility to know whether that option should become the product.
Our value is moving
The uncomfortable truth is that writing code by hand is becoming a smaller part of the value some engineers provide. I do not say that lightly. Coding has been both a skill and a source of identity for me.
But smaller does not mean absent. It means the work is spreading into different places:
- Framing the right problem before generating a solution
- Designing a system the tool can build without losing the product’s intent
- Navigating models and workflows without confusing speed with quality
- Evaluating output that looks plausible but may still be wrong
- Knowing what is safe, useful, maintainable, and ready for someone else to consume
- Taking responsibility for the final result
Execution is moving into the background. Judgment is moving into the foreground.
This is not a comfortable trade. Judgment is harder to measure than lines of code, and taste does not produce a neat activity log. A day spent making one crucial decision may look less productive than a day spent writing a thousand lines. It may also be far more valuable.
A different kind of handmade
I am not ready to say that the old craft no longer matters. Engineers still need enough depth to inspect what the tools produce, recognise weak architecture, and step in when the abstraction fails. We cannot navigate systems we do not understand.
But I am also not interested in treating manual work as automatically more meaningful. Handmade software is not better simply because it took longer. The user does not owe us appreciation for the difficulty of our process.
The work now is to keep the care while changing the method.
That means building a new relationship with our own expertise. Our value is not only in how much code we can personally produce. It is in how well we can turn an idea into a thoughtful, dependable result, using every tool available without surrendering our standards to it.
I still want to feel my hands in the work. I am beginning to understand that they may show up less in every line and more in the shape of the whole thing.
The craft is not gone. It has moved from making every piece to knowing what deserves to be made and deciding when it is truly ready.