Wanted to play with his wife—created Solviq: "What? Where? When?" in Ukrainian

I just wanted to sit on the sofa and play

This story has a very recognizable beginning. The Solviq author and his wife love solving 'What? Where? When?' questions in the evenings. He wanted to find a convenient place in Ukrainian, sit down together, and play. This very need became the reason to create his own service.

Solviq was created by Maks Traffic PRO, a member of our community. According to him, he wrote the code using Claude Code and Codex: formulating tasks, checking the result, and asking for corrections. Not a single line written manually — that is how the author describes his development method.

His motivation sounds better than a typical pitch about an 'education revolution': to make a game that he himself wants to use. Now he invites others to play and report what breaks.

What is Solviq and who should try it

Solviq — an intellectual browser-based game with 'What? Where? When?' questions. Here, the journey to the answer is more interesting: noticing a hint, connecting details, discarding the first too obvious version. The Ukrainian language and the short question format make it a suitable option for an evening when you want to stretch your brain a little.

The service can be tried solo or by discussing questions in pairs before entering the answer. This is an idea for a shared evening around one screen, not a claim of a separate multiplayer mode for couples. The author describes support for both phone and computer and free access.

An account is not required for the first encounter. At the same time, current interface texts directly indicate that ranked play is available after logging in; in-depth explanations also suggest logging in. So the promise of 'everything without registration' should be read with this clarification. The author calls Google login a way to save your rating.

Why a quiz needs an AI Judge

Text-based quizzes have a familiar problem: you understood the answer, but wrote it slightly differently. A typo, a different case, or a synonym can turn a correct thought into a frustrating 'not accepted'.

According to the author, Solviq handles this with an AI Judge that accounts for typos and synonyms. If a player disagrees with the result, an appeal is provided. The public client includes appropriate texts and interaction elements. How successfully the system handles complex, controversial answers can only be assessed with specific examples — we have not conducted such a comparative test yet.

The idea here is sound: a person should compete with the question, not try to guess the exact string saved by the developer. At the same time, the right to appeal a decision is important precisely because an automated judge can also make mistakes.

Magistrate: the answer can also be discussed

Another feature is the host, Magistrate. The author describes voice explanations of why the answer is what it is. The current client includes buttons to play the voice, ask questions about the answer, and work with appeals.

For such a game, the explanation is often as interesting as the score itself. A good question leaves behind an 'ah, so that's how it's connected' moment. If the host helps you see that connection, a lost minute can still be pleasant. This is our assessment of the concept; we have not verified the quality of all voice responses.

For the competition, the author claims an Elo-based rating. In the usual sense of this approach, the difficulty of the opponent or the task matters, not just the total number of correct answers. We have not audited the exact formula or the stability of the Solviq rating.

A month of development, Claude Code, and Codex

The author mentions about a month from the start of work to this launch, of which approximately 24 days were spent actively working on the project. Here are the figures from his message:

• 874 commits in the development history;
• approximately 109 thousand lines and 770 tests;
• about $2000 spent on Claude — this is not the stated full budget of the service;
• about 40 thousand collected questions, of which approximately 17 thousand remained after cleaning, according to him.

We provide this data based on the developer's words: we do not have access to his private commit history, testing reports, or invoices. The current public client also has 17,000 set for the counter, but this label itself is not an independent database count. The provided screenshot shows a different number — 10,466+; the author has not yet explained the reason for the difference.

For the history of Vibe Coding, the workflow itself is more interesting. A person holds the concept and decides what should happen in the game; agents help turn this into code. After that, there is running, checking, errors, and the next iteration. A large number of lines does not replace a smooth gameplay experience, and the number of tests without their content does not show whether the game is pleasant to use.

Tournament questions — a separate part of the product

According to the author, Solviq uses real tournament questions. The website texts claim attribution to authors and tournaments; the client includes fields for the source, author, and a link to the original.

For the reader, this is a useful detail: you can look up the origin of a question rather than perceiving it as an anonymous generator response. We have not verified the completeness of the attribution, the translation method, or the terms of use for the entire database. This is a separate topic for a future conversation with the developer.

What can be taken from this for your own project

The strongest part of the Solviq story is the specific need it started from. 'My wife and I want to play in Ukrainian in the evening' gives the development a clear direction: quickly open the game, get an interesting question, answer conveniently, and want the next one.

A second useful thought is that AI can be both a development tool and part of the product itself. In this case, Claude Code and Codex helped the author with the code, while the judge and host are meant to help the player during the game. The utility of each of these solutions is verified separately.

If you want to start your own game, we have a mini-course on making games with AI and a a short guide with a ready-to-use prompt. Another project from a community member is — the Android notepad Origami Notes by BaraniukConnection.

And Solviq can already be opened and tried. The author asks for specific feedback: what didn't work, where the controls are unclear, or which scoring caused questions. You can discuss the game and your own experiments in our Ukrainian Vibe Coding community.