The Job Titles Are Melting. Your Org Chart Isn't Ready.

The lines between your product roles are getting soft. A designer builds his own tools now. A PM writes specs an agent turns into working code. A senior engineer ships what used to take three people. And the junior seat, the one where new people learned the craft, is disappearing before anyone planned a replacement.

That's a lot of moving parts at once, and it lands on you: who does what, who you hire, and how people grow. Let me catch you up.

When the designer becomes the builder

The wall between designing something and building it is coming down. Pendar, a Slack designer who spent 12 years at Google, built a Claude Code prototyping kit that spins up interactive prototypes in minutes. He can concept an idea, build the tooling, and refine it without waiting on engineering. As he puts it, the line between "designer" and "builder" has effectively collapsed for him.

What's telling is how he got there. He didn't set out to learn AI. He kept building small tools to scratch his own itch, then shared them with his team, who improved them and passed them back. That remix culture is the real shift. Design work is no longer static screens. It's systems that generate the right interface for a moment.

The same collapse shows up on the product side. Peter Yang's six-step process goes from idea to working app in 25 minutes, with a design.md file, a prototype, and an HTML spec feeding straight into a build. One person now covers ground that used to need a handoff chain.

Why your best PM should be building, not managing

Whatnot founded its product team on a strange premise. Tom Verrilli, the CPO, says they regret that product management exists as a separate function. Whatnot has hit over $8 billion in GMV as the fastest-growing US marketplace, so the take earns a hearing. The point isn't that PMs are useless. It's that a lot of PM work was coordination tax, and AI is cutting the tax.

Verrilli's own frame is a shift toward senior ICs doing the work directly. Fewer people managing the flow of decisions, more people making and shipping them. When a designer can build and a PM can spec straight into code, the middle layer of "who's aligning whom" gets thin.

For your team, that means rethinking what you reward. If you promote people for running process, you're paying for the part machines now handle. The people worth their seniority are the ones with taste and judgment who can go end to end.

The business case for customizing everything

There's an economic engine under all this, and it's worth understanding before your next roadmap review. For 50 years, software companies built one product and sold it many times because custom work was too expensive to support. Yannis Perrakis lays out how agentic development moves that math. When each variant gets cheaper to create and maintain, custom builds that used to lose money start to make sense.

He's honest about the catch. AI is good at writing code. It's much weaker at proving code is correct and keeping it running for years. If AI cuts generation cost by 80% but verification and maintenance by only 10%, the win is smaller than the demos suggest.

So the roles that grow in value are the boring ones: quality engineering, defining constraints, building evaluation systems. When software can take many shapes, testing one known app isn't enough. That's a hiring signal, not a footnote.

The seat where people used to learn

Here's the part that should worry you most, because it's slow and hard to reverse. Stanford's Digital Economy Lab used ADP payroll data, real paychecks, and found employment for developers aged 22 to 25 fell nearly 20% from its late-2022 peak. Workers 30 and older in the same fields grew 6 to 12%.

AI didn't replace engineers. It replaced the entry point into becoming one. The junior seat was where people picked up the craft by doing the small stuff. Cut the seat and you keep your current seniors, but you stop making new ones. The bill comes due later, when today's seniors move on and there's no bench behind them.

The deep cut

You can absorb every trend here except one, and the exception is the junior pipeline. Collapsing roles and cheaper customization play to your favor this quarter. A broken training system hurts you in three years, and by then it's expensive to fix.

The practical move is to rebuild how people learn, on purpose, since the old apprenticeship is gone. Pendar learned by building small tools and sharing them. Mentorship is the other lever. NN/g notes that highly effective mentorships last a couple of years, not a few sessions, and that organizations with mentorship programs lifted promotion and retention for marginalized groups by up to 24%. If AI is eating the on-ramp, you need to build a new one, with named mentors, real cadence, and stretch work that forces judgment. Don't wait for HR to invent it.

Three questions for your team

• Which of our PM and design roles are still paid mostly for coordination, and what would those people do if we freed them to build directly?
• If we stopped hiring juniors this year, where would our next generation of seniors come from, and can we point to a real mentorship plan on paper?
• As we ship more custom work, who owns verification and maintenance, and are we staffing that as seriously as we're staffing generation?