Skip to main content
DanManREAL/ AI_LAB
Updated Nov 20, 2026Request a test

AboutWho runs the lab

I'm Dan. I test AI tools on real work.

About

I'm Dan. I build games, tools and an AI video factory with AI coding agents. Here I share how that work goes, with the results, the bills and the failures.

Why a lab

This lab grew out of my own projects. When an agent builds something useful, wastes time or gets a job wrong, I can show you what happened. Read the reports and explore my AI Coding Agents Tier List. The method is on How I test.

The setup

I use these five apps to work with AI coding agents:

  • Claude Code: Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, and DeepSeek V4.1 Flash through the DeepSeek API
  • Codex CLI: GPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna
  • Gemini CLI: Gemini 4 Argon on Vertex AI
  • Qoder CLI: GLM-5.3, Sonus, Qwen 3.8 Max
  • Muse Code: Muse Wonder Full 1.0

Local tests run on an RTX 5080 (16 GB) and an RTX 5060 (8 GB). Where each agent stands is on the tier list.

What I build with them

A free browser game, REAL_GRIND. A local AI video factory, MythREAL_FACTORY, which makes my recent episodes. Unofficial community tools for Godforge Duels. GIT_REAL, a tool that checks for work an agent could lose. Read the build stories to see how they came together and what went wrong along the way.

Two developers, one house

Ava designs and builds her own games. She made some of them all by herself and others together with her dad, who helps put them online. They live at Ava's Games.

Independence

Independent testing. Sponsorships are always disclosed, and verdicts are never for sale. If your team lets AI coding agents work on real projects, I also offer setup services with a defined scope. You can also email me at [email protected].