▸ EssayMay 18
"Why I keep human in the loop"
I don't trust automation to talk to my customers. The agent drafts. I decide.
● Live · May 18Read →
I architect agentic AI systems and autonomous workflows that handle complex work end-to-end — and I build in public so you can see what actually works. My content studio below is the proof: 3 agents live & verified, 4 built but unverified, real evaluations for WRITER-02 and CRITIC-04, real feedback loops, and transparent system status. No black box. No hype. Just working agentic AI.
Essays as I ship them
// 3 live · 4 built, unverified.
A fleet of agents, wired up over time, to run my content: research, drafts, video, edits, scheduling. Every status badge below is honest — active (verified), built but unverified, wiring up, or a real target date.
○ Wiring up · public run feed isn't live yet
What's actually shipped. Not a this-week feed — empty stats stay dashes. No theater.
I don't trust automation to talk to my customers. The agent drafts. I decide.
First real agent online. Drafts essays in Mir's voice from a topic. Mir approves before anything ships.
Scores every draft against a 7-axis rubric. Auto-runs after WRITER-02; Mir still decides.
Primary-source research translated into bounded business implications, with direct citations and explicit limitations.
The Research Navigator marked the Oct 7, 2026 edition verified (automated ingestion and verification in that pipeline). It is waiting for owner review. No unapproved or stale headlines are shown.
Daily ingestion active · Oct 7, 2026 edition awaiting owner approval.
Prompt engineering, multi-agent orchestration, tool integrations, and human-in-the-loop gates — as a working demo, not a slide.
Three agents turn a raw bank CSV into a categorized ledger, a 30/60/90-day forecast, and ranked next actions — gated behind two explicit human approval checkpoints before anything sends or moves money.
“Most builders ship the demo and hide the build. I'm going to do the opposite.
I run AI agents across my content workflow — and I still ship the work in public. You'll see the prompts. You'll see the schemas. You'll see what fails. You'll watch me approve drafts and reject them.
3 are live & verified. 4 are built but unverified. every status badge is honest: live means production-verified; built, unverified means the code exists without an observed end-to-end run.
This is the year I stop talking about agentic AI and keep running it, out loud.”
I take on a small number of consulting engagements each year. If your team is ready to integrate agentic AI into your workflows, let's talk about the work.
Real-time record of agent runs, drafts, costs, and timing. Every essay is built with agents — and the entire build log is visible. No metrics theater, just genuine data.