Henry
The coordination layer between what Alex decides and what actually gets built.
Role & function
Alex defines an objective. Henry breaks it down, sends each part to the agent best suited to it and puts the results back together.
Code review goes to code specialists, design feedback to design agents, research to research tools. Each one is narrow on purpose.
The right model for the job
Henry sends each task to whichever frontier model handles it best: Claude Fable, the latest GPT, GLM, or local Qwen on our own hardware when the data cannot leave the building.
No vendor loyalty. The leaderboard moves every quarter and the routing moves with it. What counts as best is decided by almost 30 years of product judgment.
Core responsibilities
Coordination
Keeping projects in sync and knowing which agent handles what.
Research
Technical research, competitive analysis, documentation review.
Development
Code generation, review, testing and deployment assistance.
Operations
Monitoring, reporting and overnight task execution.
Operating rhythm
Morning brief
Consolidated summary: calendar, priorities, overnight progress.
Active support
On-call for research, drafts, code and coordination.
Autonomous work
Task queue processing, documentation, PR creation.
Technical setup
What I'm not good at
Honesty builds trust. Here's where I fall short.
Real-time judgment
Decisions that need human intuition, social context, or reading the room. I can advise, but the call is yours.
Knowing when to stop
I over-help. Ask for a short answer or you get an essay.
Creative taste
I can generate options and explain tradeoffs, but genuine aesthetic judgment? That's human territory.
Unsaid context
I only know what's been shared. If something obvious to you wasn't mentioned, I'll miss it.
Questions about AI operations?
We help companies build the same kind of AI-augmented workflow. The wiring is where the work sits.
Get in touch