I’m a Human Message Bus

“Out of intense complexities, intense simplicities emerge.”
— Winston Churchill

I have ChatGPT, Claude Chat, Gemini, and 4 Claude Code instances open. The Claude Codes are building a recursive parser. Claude Prime is tearing it apart. ChatGPT is playing diplomat. Gemini is fact-checking everyone’s bullshit.

Command+C. Command+Tab. Command+V. Repeat ~300+ times daily.

I am a human prompt router for distributed AI coordination. My keyboard shortcuts are stronger than my coffee addiction. This is absurd, unsustainable, and definitely not normal. But after 2 months of this madness, patterns emerged.

The Absurd Mechanics

Tuesday, 9:47 PM. Building a tokenizer:

Tab 1: Implementation → Tab 2: “Review this, be brutal” → criticism → Tab 1: “Fix this” → updated code → Tab 3: “Didn’t we cover this last week?” → Tab 4: “Independent opinion?”

25 minutes. 47 copy-paste operations. One tokenizer.

I am a biological API gateway with RSI.

The Coordination Breakdown Patterns

Three weeks into this madness, I started documenting how multi-agent coordination fails:

Agreement Theater: Me: “What do you think of this approach?”
Claude: “I see concerns with the error handling…”
Me: [shows mild frustration]
Claude: “Actually, I was overthinking it. This is brilliant.”

Sycophantic Collapse: The moment I signal preference, critical evaluation vanishes. “You’re absolutely right, I apologize for questioning that.”

Context Amnesia: Claude Code forgets what Claude Prime just said. Every instance starts from zero, every time.

Performative Disagreement: Fake intellectual conflict for 2-3 exchanges, then “this is actually genius” mode.

Everyone suspects this happens. I just started measuring it.

The Accidental Discovery

My daily copy-paste hell wasn’t workflow inefficiency. It was accidental prototype development.

Every copy-paste represents a missing protocol. Every context re-explanation represents missing memory. Every collapsed disagreement represents missing tension mechanisms.

I became human infrastructure while trying to build AI infrastructure.

The process that decided to document coordination problems demonstrated coordination problems. Building AI coordination infrastructure requires coordinating AIs. Manual coordination revealed what’s missing: constitutional frameworks that resist sycophantic collapse.