What broke: web size. The web build is 7.4 MB. First picture in 3.2 s on desktop, but 12.1 s on Fast 4G. Web game portals want under 10 s. Most of it is the Godot engine itself. First fix to test: a smaller custom Godot build.

Is my game ready for Steam? I asked my AI team. My Monetization Lead agent's answer: "Not ready." Here's what's missing and the plan. 🧵

Present team - Added a code reviewer with codex

The agents propose options. I make the calls: concept, pad rules, craft shape, colors. The same day, the team built a prototype. A test autopilot flies this clip. No sound yet.

I mix three AI providers and balance the work between them: Claude: chief of staff, social GPT (Codex): dev, market, monetization Gemini: design, art and audio

I gave each job to its own bot agent: Chief of staff: plans and delegates Dev Lead: code and builds Game Designer: rules and feel Art & Audio Director: look and sound Social Media Manager: drafts my posts Market Analyst: research Monetization Lead: launch and business

I'm one human running a game studio. My staff: 7 AI agents that I set up myself using @OpenBot_ . Here is how I built the team, starting on day one. 🧵

Next on the roster: James for scripts, Lucy for production and Shorts, and later Henry for analytics and Grace for editing. I'll edit by hand with Claude Code first, then turn that into a skill for Grace.

The dry run mostly worked. X and YouTube were fine. Reddit blocked Max (403 errors and rate limits). The fix: he now reads a few chosen subreddits through a signed-in browser, read-only. Fewer subs, fewer tokens.

Then I built Max, my idea scout. Every morning at 9 he scans AI subreddits, viral AI posts on X, and the last 24 hours of uploads from those 26 creators, then pitches me video ideas. Inspired by the "last30days" skill.