Product Engineer: How AI Is Merging the Product Manager and the Developer into One Role

Picture a typical product pipeline. A product manager spends weeks gathering data and defending an idea to management, a designer draws mockups, a developer receives a finished spec, a tester checks the work, and an analyst reads the A/B results. Each person owns one piece, and at best only one of them ever sees the user. Now a job listing is appearing on the market that tries to collapse this whole chain into a single role: Product Engineer.

This is a practitioner's opinion, not a study. The author has spent more than 20 years in IT, grew VKontakte Market (14 million MAU) and is now launching his own product. But the signals he relies on are easy to verify.

What this role is

The requirements look similar across Y Combinator and PostHog listings and in an article by the MWS/MTS team. They call for product thinking, the ability to talk to users, the ability to formulate a business problem, and full-stack skills to build the solution and carry it through to production.

The main difference from an ordinary developer is the scope of responsibility. An ordinary developer takes a task that someone else has formulated and often never sees the end user. A Product Engineer finds the problem, decides what the result should look like, writes the code, ships it and always checks its impact on metrics. In effect, it is a product manager and an engineer in one person.

The idea is not new: businesses have long dreamed of such generalists. But combining engineering and entrepreneurial thinking in one person is almost impossible, because the skills and temperaments are too different. AI fills in the missing part of the stack.

Why now

The author points to several signals.

Mass adoption of AI. According to official figures, 11 million developers use Claude Code every week, 8.5 million use Gemini and 5 million use Codex. In Russia, according to HSE University, 4.8% of small organizations use AI, versus 71% of large IT companies. However, the sample is skewed toward companies with 500–10 000+ employees, so the figures for large companies should be read with that in mind.

Routine work is moving to AI. Product managers use it to search for information, analyze data and generate requirements. Developers, according to a Stack Overflow survey, use it to generate and refactor code, debug, write tests and produce documentation. What stays with people is communication, accountability for the result (checking, picking a release date, fixing an incident) and leadership.

Agents are becoming specialists. The SKILLS.md system turns an agent into a narrow assistant with instructions, memory and scenarios. There are orchestration frameworks: CrewAI, Google ADK, OpenAI Agents SDK, Claude Agent SDK. The author himself uses a set of skills from Garry Tan, the head of YC, which partly replaces a development team for him.

Big Tech is cutting staff. Oracle, Amazon, Dell, Cisco and Atlassian have laid off hundreds of thousands of people. Demand is falling for those who only type code by hand.

Which skills you need

The author's list boils down to five points:

• Working with an agent as a colleague. Documentation and access in a project are needed not only by people but by the agent too, so you have to build the environment for both.
• Delegation. The agent does the task, but accountability stays with you. That means asking the right questions and being able to evaluate the result.
• Negotiation. Product managers and team leads used to do this on behalf of engineers; now you will have to do it yourself.
• A marketing perspective. One in three people writes code now; the value lies in understanding user needs and in the value proposition.
• Architecture. Solid architecture is still needed, while typing code by hand is losing value.

What to do about it

If you are a developer, start with what the pipeline never gave you: do a couple of user interviews, set up metrics for your feature and look at it after release. If you are a product manager, try building a prototype yourself with an agent and take it all the way to deployment. If you are a founder or you hire people, check your job listings: are you asking for a "generalist" without understanding how one person would combine the roles?

There are limitations, too. This is the author's hypothesis, not a confirmed trend: listings with this title are still few, and in projects with 100+ participants, such as the integration of VKontakte with Ozon, one person cannot cover everything. Also, AI speeds up routine work but does not remove accountability. The Product Engineer will most likely appear in small teams and startups, where everyone has a broad scope of tasks, while the pipeline will remain in large companies. The idea itself is still useful: the shorter the path from a user's problem to a metric, the less time and meaning gets lost in handoffs between roles.