PRD-to-story cycles, after I deployed the agent stack inside a regulated fintech program.
A small operating system for my own work: one feed that keeps pace with the field, agents that work with me in the loop, one interface that routes it. It's how I'd run a product.
I built my own learning surfaces: one keeps me reading the frontier, one keeps my own work a layer deeper than the résumé.
A fresh, personalized feed every day, generated on a 6am cron from the people and subjects I follow.
I hold my own work to a documented standard, broken down past the artifact, the same rigor I bring to a roadmap.
Every agent here drafts; I decide. Nothing goes out without me.
Reads a recruiter thread, detects cold, warm, or a follow-up, then drafts the reply in my own voice straight into Gmail. I read it, tweak it, send it.
An agent built on my actual work that a recruiter can question.
Two agents get me ready before a room; a ledger feeds every rep into the next one.
Assembles a live dossier on whoever I'm about to meet, the company, the product, so I walk in from a real brief, not a guess.
A technical-fluency coach: name the architecture, trace the data path, own the tradeoff, until that's just how I talk.
I don't remember agent names or juggle a dozen tools. There is one front door.
One seat I talk to like a chief of staff: I say what I want, it routes to the right specialist, and brings back one answer. I stay on the decision boundary for messages, applications, and public work.
I designed the data flow, the agent boundaries, and the guardrails underneath it, including a delete-nothing rule so no automation can lose my data.
The unattended part is narrow, on purpose: each of these six exists because something went wrong once, and a note in a document turned out not to be a control.
A prompt is a request. Code is a rule.