Aleksei Ivanov

Founder of Multify, an automated website translation and localization service for international companies.

I am a software engineer with 15+ years of experience. I focus on system architecture and shipping resilient products that are a joy to use.

Beyond building products, I am drawn to foundational computer science: how complex behavior emerges from simple rules, and how to build systems that organize themselves.

Currently working on

Multify

Automated website translation platform that helps businesses expand globally

Visit project site →

redraft.page

Refine text with your voice. Highlight a paragraph, speak a command, get exactly that change

Visit project site →

Lavoro

Background task scheduling library with fluent, type-safe API & support for multiple drivers

Visit project site →

Latest from the blog

24 Apr, 2026 Co-authored by whom?
Claude Code is a tool that you paid for. Usually, you don’t get advertisement in a tool you have paid for. So when Claude appends “Co-authored by Claude” to every commit message, it is only natural it would raise an eyebrow.

It’s not a simple matter, either: disclosing AI generated content might be a good idea, but it has so many edge cases you can cut yourself if you are not careful.

First of all, it might generate this for a commit it didn’t even contribute to (part from the commit message). That’s a stretch to call it co-authoring.

Then–at least for me–this disclaimer suggests that Claude did more than half of the work, while in reality it was merely instructed to execute a set of tasks. The idea was mine, the control is mine, I might very well be responsible for the commit and take credit for it.

You don’t credit Microsoft PowerPoint for the keynote you made or say that built in calculator app “co-authored” you financial analysis. LLMs are tools, not sentient beings. In fact, they have a private owner–the company that rents it to you.

And this is where the third thing comes into the light: intellectual property rights. Oh boy, is it a grey area with this one.

Models providers had to source huge amounts of data to train their models and as far as I know each mode weight doesn’t have “co-authored by this YouTuber” or “co-authored by this blogger” on them. Furthermore, what happens when you make a commit–which is signed off by you–that mentions such co-authoring?

Would that legally transfer some of the rights to it to the company that rented you the model? We don’t know yet, and when there is ambiguity it is better to play safe.

I started writing this post with no strict position on this matter. But in the process of writing it I think I see that there are really no redeeming qualities for doing this the way it is being done currently.

Other posts:

24 Apr, 2026 Co-authored by whom?
23 Apr, 2026 Interfaces as documentation
22 Apr, 2026 Special treatment
21 Apr, 2026 They simply don’t have the time to test it
20 Apr, 2026 Your life is your life’s savings
18 Apr, 2026 Future hackathons
18 Apr, 2026 Student or apprentice?
16 Apr, 2026 Speed and complexity; now and then
15 Apr, 2026 ADD: Anger Driven Development
15 Apr, 2026 Code is not even close to half the battle
14 Apr, 2026 Check your surroundings

See all posts →

Pet projects

From time to time, apart from programming, I also dabble in digital electronics:

VR controller prototype based on ESP8266 (2019)

In this project I was tinkering with ESP8266, making it connect to the PC over Wi-Fi. I used it for emulation of an HTC Vive VR controller through VRidge API over ZeroMQ protocol. During the process I also wrote a simplified ZMQ client that can run on Arduino.

GitHub
Hardware VGM chiptune music player (2018)

Hardware player that plays chiptune music in VGM format. In particular, it uses SN76489 chip to play music using up to 4 square-wave voices. The repository contains tool written in JavaScript that proxies the raw VGM byte stream to the Arduino along with the firmware that controls SN76489.

GitHub
Hardware and software design of an LED matrix display from scratch (2018)

This is my first digital device that I have designed from scratch: from schematics and hardware communication protocol, to the low-level display driver. The final result is an Arduino project that uses this driver software to control the display and play animations.

GitHub

Talks

Comparative analysis of cloud gaming services and streaming technologies (2020)

In this project I compared cloud gaming services by testing their bandwidth usage, latency, and adaptation algorithms using Wireshark. I analyzed five services (Parsec, Steam Remote Play, Moonlight, GeForce NOW, and Google Stadia) and reviewed the evolution of cloud gaming from 2010 to 2020.

Slides Paper
Interactive dashboard for phenological data monitoring across Germany (2020)

In this project I built an interactive dashboard to monitor phenological data across Germany. I developed a data pipeline that processes open data from the DWD, created a methodology to formalize text phase descriptions into comparable numeric values, and visualized the results using Grafana.

Slides Paper GitHub
Systematic hyperparameter tuning for deep Q‑learning agents (2020)

In this study I systematically tested how hyperparameters affect deep Q‑learning agent performance. I experimented with memory size, discount factors, and vision range in a custom grid world environment. I gained some new insights about possible new approaches for adaptive memory.

Slides Paper