The middle of software work is collapsing.
For most of my career, the middle was where everything happened. You’d have an idea and eventually ship something, but almost all the effort lived between those two moments: writing code, debugging, translating intent into implementation across handoffs and revision cycles.
That middle absorbed most of the time, most of the talent, and most of what we called craft.
AI agents are compressing it.
If AI is becoming standard design infrastructure, the relevant question is no longer whether to adopt it. It’s how to remain consequential alongside it. I have been living that question in industrial cybersecurity, where the middle used to be the long translation between a security engineer’s model and an operator’s next thirty seconds.
The grief nobody is naming
Nobody wants to say this, so I will.
The grief designers are feeling right now isn’t about losing their jobs. It’s about discovering that what they thought was craft was closer to execution.
For twenty years, design was gatekept by tools. Photoshop took months to learn, then Figma changed that, and AI changed it again. Each time the barrier dropped, the population of “designers” expanded, and what the word actually meant got blurrier.
What AI revealed isn’t that designers are replaceable. It’s that a significant portion of design work was always pattern assembly. Visual hierarchy, spacing, color, consistent components: these are learnable rules. AI learned them faster.
The identity crisis runs deeper than the career crisis.
Designers spent years building a self-concept around taste, craft, and execution: the portfolio, the reverence for process, seniority measured in delivery. Then a model trained on Dribbble screenshots matched the output.
That’s not a job threat. That’s an existential one.
What’s harder to admit is that responding to it requires more than picking up new tools. It requires letting go of a professional identity built around making the artifact. That’s not a minor strategic adjustment. It’s an identity crisis dressed up as a career pivot.
The people shouting loudest about AI’s limitations in design are, almost without exception, the ones whose portfolios look most like what AI produces best.
The ones who are quiet are working at a level AI hasn’t touched: org dynamics, trust architecture, the political read on why a product keeps getting rebuilt, and the ability to tell a founder that the problem isn’t the interface.
That work was never visible in a portfolio. It never won Awwwards. It just made products work.
The contraction isn’t punishing bad designers. It’s exposing a definition of design that was always too narrow.
Also in April: The Two Edges of Judgment · Designers Need a Sandbox Repo · Judgment Is the Territory
Daniel Arevalo is Product Experience Lead at Omny Security. He builds AI-native tools and writes about design, product leadership, and what it means to direct systems that execute at a speed humans never had before.