“Did you ask Chat GPT?”: How a tech lead’s joke turned into an AI reviewer in two days

“Did you ask ChatGPT?” — this phrase became our inside joke. Tech lead Ivan was constantly experimenting with artificial intelligence, and his colleagues teased him about his hobby. But when the question of automating code reviews arose, it was this enthusiasm that allowed us to put together a working prototype in just two days.

As the CEO of Nomium with 11 years of experience in custom development, I, Oleg Akulov, am confident that the future lies in the intelligent automation of routine processes. Today, I am ready to share the results of an experiment that will soon grow into a full-fledged product.

Problems with manual code review

Our company employs 40 developers working on various projects. Code review is a mandatory step, but it takes up significant resources from technical leads and team leads.

Key challenges:

• Reviewing each pull request requires attention to style, architecture, security, and performance.
• Large amounts of code are difficult to analyze without errors.
• Reviewing newcomers and colleagues from other projects requires additional time to immerse oneself in the context.
• The process often turns into a formality or protracted discussions.

Research confirms that manual review does not scale well and takes up the time of senior developers. Even analyzing a few files can take hours.

Why we chose artificial intelligence

We actively use AI tools in our daily work:

• n8n — for automation and service integration
• Replit AI — for code generation
• Lovable — for rapid prototyping
• ChatGPT — for testing hypotheses and creating examples

The statistics convinced us:

The birth of an idea and preparation for the hackathon

In the summer, our product manager Maria gathered ideas for internal products. We have been engaged in custom development for 15 years and understand the challenges faced by developers.

The idea of an AI reviewer won because:

• It solved a specific problem for technical leads.
• It matched our experience and expertise.
• It had high automation potential.

The decisive moment came when I suggested holding a hackathon. We decided to combine the creation of an MVP with a team event.

Organization of the hackathon: July 26-27

We assembled a distributed team of developers, designers, and marketers. The tasks were divided as follows:

• Technical team — architecture, AI integration, API
• Design and marketing — branding, interface, presentation

Day one

• General meeting and task setting
• Preparation of datasets and API design
• Configuration of n8n and Replit
• Creation of the interface using template generators

Day two

• Completion of work on the MVP
• Testing and assembly of the presentation
• Internal demonstration and collection of feedback

Key findings and insights

Technical achievements

• Working prototype of an AI code reviewer
• Integration with popular neural network services
• API for connecting to existing processes
• User interface for teams

Team results

• Enhanced collaboration in a distributed team
• Knowledge sharing between different specializations
• Practical mastery of new AI tools
• Confirmation of the effectiveness of the hackathon format

Business insights

• Confirmation of demand for code review automation
• Demonstration of the technical feasibility of the concept
• Experience in rapid prototyping with AI tools

What we learned in 48 hours

AI tools save time

n8n enabled rapid service integration, Replit AI generated template code, and Lovable allowed marketers to create landing pages without developer assistance.

Remote teams can work effectively

Different time zones became an advantage—work continued around the clock.

MVP in two days is possible

We proved that code review automation is possible and can be implemented by a small team.

Coming soon: Merge Senei announcement

This experiment will not remain an internal project. Very soon, we will present Merge Sensei, an AI tool for automatic code review created during the hackathon.

What will be included in the first version:

• Automatic code checking for errors and vulnerabilities
• Integration with popular IDEs and platforms
• Adaptation for different programming languages
• Intuitive interface for development teams

Stay tuned for news — we are preparing a product announcement and special conditions for early adopters.

Oleg Akulov, CEO of Nomium

Want to be the first to try our AI code review?
Write to us