# KK / AI × Game Dev

URL: https://kiki.it.com/en/

> KK's notes on AI, game development, Codex collaboration, and practical technology exploration.

Lang: en

## Markdown Sources
- [Home](https://kiki.it.com/en/index.md)
- [About](https://kiki.it.com/en/about.md)
- [Words](https://kiki.it.com/en/words.md)
- [Links](https://kiki.it.com/en/links.md)
- [Archives](https://kiki.it.com/en/archives.md)
- [Categories](https://kiki.it.com/en/categories/index.md)
- [Tags](https://kiki.it.com/en/tags/index.md)
- [LLMs.txt](https://kiki.it.com/en/llms.txt)
- [LLMs-full.txt](https://kiki.it.com/en/llms-full.txt)

## About
## Who I am

I am KK, a game developer who keeps following AI, agents, and new technologies that can change how work gets done.

I enjoy building features, debugging problems, and turning a vague idea into something that runs. This site is not a list of what I know. It is a record of how I break down a problem, find a path, learn from failures, and turn the result into something other people can understand.

## How I work with AI

I treat AI as a collaborator, not a black box that makes the final call for me.

When a task depends on a game engine, a platform API, or a real runtime environment, I validate the critical path myself first. Then I give Codex the method, constraints, and success criteria so it can handle the full implementation, repetition, and detail checks.

For more complex work, I ask Codex for a plan, review it against the real constraints, and then let it ship the implementation. AI saves time, but it also gives me more room to follow new capabilities and new ways of working.

## What I write here

- Practical experiences with AI tools and agents
- Game development features and debugging postmortems
- Technical exploration, infrastructure, and workflows
- AI frontier news with the original source, a plain-language explanation, and an objective breakdown

I want these notes to help readers avoid a few wrong turns and make human–AI collaboration concrete.

## Latest Posts
- [How I Used Codex to Fix an Item Count That Went Negative](https://kiki.it.com/en/posts/codex-negative-item.md): A simple bug became a useful lesson in where AI helps most: reproduce the problem first, then let it scan the details people tend to miss.
- [How I Work with Codex When It Cannot Run the Game Engine](https://kiki.it.com/en/posts/codex-engine-collaboration.md): Game engines and platform APIs contain rules AI cannot directly verify, so I validate the critical path first and then let Codex expand the proven method into a complete feature.
- [Why I Want Agents That Can Operate a Computer](https://kiki.it.com/en/posts/agents-that-operate-computers.md): When I follow new models and agents, I care less about the numbers than the tasks they can actually complete and the control I should keep.
- [Why I Built My Own Infrastructure: Keeping an AI Workflow Reliable](https://kiki.it.com/en/posts/ai-workflow-infrastructure.md): I did not build my own network infrastructure to show off. My work depends on AI, and when the tools became unreliable, I needed to make the workflow reliable again.