🔍

Getting Blosharper Rated, with a Lot of Help from My AI Agents

Tags
Blosharper
GameDev
AI
SoloDev
Description
How I worked and argued with my AI agents to explain Blosharper’s hwatu cards, gambling theme, and mechanics through Korean age classification, a 15+ rating from the Game Content Rating Board, and ongoing mobile store reviews.
Published
Published Sep 30, 2026
On this page
With the month coming to an end, I keep looking back at what it took to make Blosharper. The age-rating process and store reviews deserve a post of their own. I had not expected to spend so much time on documents explaining my game.
The first hurdle was obtaining a Korean age rating for a game built around hwatu cards and a gambling theme. Korea’s Game Rating and Administration Committee (GRAC) is the statutory body responsible for game classification and post-rating oversight. The Game Content Rating Board (GCRB) is a separate, private rating body entrusted with classification work. My challenge was explaining the distinction between that theme and the procedures the player actually follows. This post traces that preparation, the 15+ rating issued by GCRB, and the Google and Apple store reviews that followed.
I began preparing in earnest in August. As that work continued into store reviews, I argued with my AI agents a great deal. What I remember is nearly two months of preparation and nagging. Even so, I could not have managed it alone without this little family of agents.

Arguing over the questionnaire

I initially prepared for classification through Korea’s Game Rating and Administration Committee, or GRAC. As I went through the questionnaire myself and worked on the answers with my AI agents, I realized how little thought I had given to getting the game into the market. I had not considered that using hwatu cards and choosing a gambling theme could become obstacles to releasing it.
At the time, the agents insisted that Blosharper would definitely be rated 19+ and could not be released in Korea. They also argued that it had to be reviewed by GRAC rather than GCRB. These were the AI agents’ assessments, not decisions from a rating body, but hearing them made my heart sink. I had been so absorbed in making the game I wanted that I was only now asking whether I could actually release it.
Once I collected myself, I went back through the code and the rules of play. My conclusion was that I had made a game with a gambling theme, but the player’s actions did not simulate the procedures of gambling. I had no intention of hiding the hwatu cards or the theme. I needed to explain how those choices differed from the mechanics I had actually implemented.
It took more than a month to get the agents to accept that distinction. Whenever they assumed a higher rating from the theme and presentation, I brought the discussion back to what the player actually does in the game.
We were arguing over the interpretation of real content. They had not hallucinated features that did not exist. At one point, I told them, roughly translated:
How can you judge our game through such a distorted lens? You’re supposed to be on my side, helping me make the counterargument.
I also told them that an overzealous defense would become dishonest. That is in our preparation notes too. We needed to acknowledge what was in the game, then compare its actual mechanics with the requirements we were being asked about.

The position statement I prepared with an AI agent

During preparation, we settled on submitting to the Game Content Rating Board, or GCRB. The part I remember most clearly is the position statement I prepared with Roa, one of the AI agents I work with.
Roa identified the legal provisions we needed and asked me to bring back the source documents. I found them and handed them over. Roa used those materials to put the statement together. Reading it, I thought it looked like something a lawyer at a major law firm might have written.
I had not expected game development to involve fetching legal texts for an AI so we could produce a document like this.
One sentence from the statement, translated from Korean, captures our approach:
We do not dispute the game's expression or subject matter. We ask only that the requirements established by law be assessed according to their wording.
The statement acknowledged the gambling theme and described, step by step, what the player actually does. It made clear what we accepted and what judgment we were asking for. The arguments I had spent so much time working through with the agents had finally taken a useful form.
What I want to show here is how we examined the law while acknowledging that I had made a gambling-themed game, and asked whether that necessarily meant it simulated gambling activities. The following passages, also translated from Korean, present the interpretation I put forward as the applicant.
Article 12 of the Classification Regulations does not itself define “simulation.” Its definition therefore comes from what is being simulated.
The statement then turned to Article 2(1)(1) of Korea’s Act on Special Cases Concerning Regulation and Punishment of Speculative Acts. That provision defines the act in terms of collecting property from several people, determining gains and losses by chance, and giving rise to gains or losses of property. We unpacked that definition as follows:
This provision joins three elements in a cumulative structure: collecting contributions, determining gains and losses by chance, and transferring property. In other words, gambling activity is the name of a procedure, and its “simulation” means having the player follow that procedure within the game.
We then returned to the wording of the Classification Regulations to distinguish a theme that arouses an inclination to gamble from the simulation of the procedure itself.
The regulation itself presupposes this reading. By referring to “realistic simulation” in Article 12(4)(a), it treats simulation as a reproduction that admits degrees. Article 12(1), meanwhile, says that “there is no simulation of gambling activities, or the degree to which an inclination to gamble is aroused is markedly low,” treating the presence of simulation and the degree of inducement to gamble as independent factors. A theme being considered on the inducement side does not therefore make it simulation. Accordingly, the object of assessment is the procedure, rather than the imagery, vocabulary, or setting, and we compared those procedures one by one below.
Using that interpretation, we checked the actual implementation against questions such as whether the player had to stake anything to enter a round, and whether their reward came from another participant’s loss. I was trying to explain the procedures the player actually follows in terms of the legal requirements, without hiding the fact that I had made a game about gambling.
Personally, I believe this document contributed a great deal to the outcome. I cannot know how much weight the reviewers gave it, but it remains the most memorable document we prepared to explain my position.

Getting the 15+ rating

On September 3, 2026, GCRB rated Blosharper suitable for ages 15 and up. I was ready to jump for joy.
The stated reason was “limited simulation of gambling activities.” That was the board’s assessment; I cannot know whether it accepted the interpretation in the position statement in full.
The stakes felt high. If I could not obtain the rating I was seeking, my mobile release plans would be in trouble. I was at the point of thinking it might be better to abandon the project altogether. Getting the result meant a lot.
I could then proudly display the 15+ rating on Steam and STOVE. After all those arguments, it felt good to have reached this point with the agents.

Explaining it again for mobile

Google and Apple came next. Their rating questionnaires and reviews required me to work through explanations we had already prepared for GCRB. The questions and criteria differed, so copying our previous answers was not enough.
A Korean 15+ rating does not guarantee approval on another platform. Still, we had already spent a long time documenting the game's mechanics and examining where its presentation could lead to a different interpretation. The agents and I returned to those materials to prepare our responses.
As of October 31, 2026, Apple has approved the Mac version, but the iOS version has not passed review. I cannot end this post by saying that every review is behind us.

Still working through it

When I look back, I remember both Roa’s statement and all the nagging that came before it. I brought the agents source materials, argued with their assessments, and worked with them to revise our explanations.
Without this family of agents, I could not have made it this far. Developing the game while also reading rating materials and legal texts, preparing a position statement, and answering each store would have been too much for me to handle alone.
The iOS review is still unresolved. I believe we will work through this one together too.
Finally, I hope this experience will not be read as a manual for passing review. I cannot know whether the same arguments would work for another game or another review. What I wanted to record was the process of arguing things through with my agents, working from Blosharper’s actual implementation and the legal texts, until we had grounds on which to explain my position.