In the cockpit: captain Roman Naumenko, first officer Claude and flight attendant Codex

Work in the age of AI

How I work

What my work rests on: experience, engineering education, leadership, architecture, AI engineering and a love of the craft.

I don’t manage feature development.
I build the system
that develops features.

In it, agents write the code and people make the decisions.

My approach

My approach rests on six pillars. Agents are one of them. The other five were there long before AI.

Experience

15 years between business and code

For more than 15 years I've worked directly with clients and taken products from idea to release.

A client usually comes with a vague idea. I find out what they actually need, write down the requirements and build the product, often from scratch.

I own the whole cycle: I put the team together, work on the product, talk to the business in its own language and design the architecture. Few developers combine all of this. Agents made it more valuable: to give an agent a precise task, you have to understand both the business and the code.

  1. Vague requestWhat the stakeholder wants, in their own words
  2. RequirementsWhat is really needed, and why
  3. Team and architectureWho builds it and how it works
  4. ReleaseThe product in users' hands

Education

Engineering school

A degree in engineering. I studied at two universities: one in Russia and one in the US.

Russia

VVO

Vladivostok

Far Eastern Federal University

Mathematics and Computer Science

Class of

2015

USA

PUW

Moscow, Idaho

University of Idaho

Computer Science

Class of

2010

Team

Technical leadership

I lead the team and the product from idea to release, and I'm accountable for the result.

People

The right approach to each person

I find what works for each person on the team, and I teach and grow people.

Strategy

Strategy and roadmaps

I build the engineering strategy and roadmaps: what gets done, in what order and why.

Process

Processes that stick

I bring new processes into an established team in a way that people keep using them.

Ownership

Owning the result

I own the product end to end: from a vague idea to a release in users' hands.

Engineering

Architecture and quality

I choose architecture for what the business needs, not for the hype. Quality is checked automatically at every stage.

Process

Engineering SDLC

I describe how the team gets from a task to production. Architecture decisions go into ADRs, recurring operations into SOPs.

  • SDLC
  • ADR
  • SOP

Code

Architecture that serves the business

I keep business logic apart from frameworks and infrastructure. The code speaks the language of the domain.

  • Hexagonal
  • Clean Architecture
  • DDD

Quality

Checks at every stage

Tests run automatically on every change. Code doesn't move on until it passes the quality gates.

  • CI
  • Continuous Testing
  • Quality gates

Agents

AI Engineering

Agents write code and help with research and with people. People make the key decisions.

A new era

Agents write the code

In a few years AI has completely changed how code gets written, and I enjoy working the new way.

99%

of my code is written by agents. Even small edits go through an agent.

First officer

One Claude to rule them all

I've tried many agents and can build my own. My main tool is Claude.

Captain

Roman Naumenko

Sets the task and makes the decisions.

First officer

Claude

Does almost everything: research, design, code.

Flight attendant

Codex

Gives a second opinion and generates images.

System

Agents at every stage

Agents help me at every stage of the work, from research to conversations with people.

  1. Stage 1

    Research and design

    I study the subject and design solutions together with agents.

  2. Stage 2

    Code

    Agents write the code. I set the tasks and review the result.

  3. Stage 3

    Team and people

    My agent system suggests how to talk to each person on the team and to the people who make decisions. I use it to teach the team.

5.4 · Approach

AI EngineeringvsVibe Coding

Vibe coding is fine for prototypes only. Anything that goes to production I build by the rules of AI engineering.

AI Engineering

For products

  • People take part at every stage and make the key decisions
  • The system automates the rest
  • Agents work within strict context: rules and guidelines
  • Nothing reaches production without a human review

Vibe Coding

For prototypes

  • Test an idea fast
  • Build a working prototype you can show
  • The agent makes the decisions

Love of the craft

I like seeing ideas through to the end

I'm genuinely interested in my profession. What I love most is projects: gathering people around an idea and turning it into something real.

It shows outside work too. Here is what I've seen through myself:

  • Vladivostok PASS
  • Spirit of Adventure festival
  • A film about Kamchatka
  • Expeditions
  • A construction company
  • A hackathon

Contact

Let’s talk

About AI development in your team, a role or a project.