Systems Lab
Enterprise architecture is the day job. On my own time I build the tools I wish I had: agents that break my work into pieces, systems that remember why I made a decision, and research loops that turn a pile of half-finished experiments into something I can actually reuse.
AgentOS · the loop
This is how a goal moves through my system — from a single line in a queue to a shipped artifact. The machine runs the loop. I only step into the one place that matters: anything that spends money, sends a message, or touches the outside world waits for me.
A goal lands as one line. Work doesn't start in a chat window — it starts as a task.
The task is triaged, then broken into worker-sized subtasks — each small enough for one agent to finish.
into the working loop
Every subtask lives here. Work is pulled from the board and returns to it — nothing survives only in a conversation.
Idle steward — when the queue goes quiet, it sweeps forgotten pockets and re-files dropped work back onto the board.
role-typed workers pull ready tasks
sources, reads, synthesizes
drafts in my voice
models, backtests, sizes
builds, deploys, infra
writes code + tests
audits before it ships
anything with real-world impact stops here
Safe & reversible — research, analysis, drafts, code on a branch. The machine keeps moving.
Money, outbound messages, deploys, external writes. Nothing irreversible happens until I say go.
Real files land — code, PRs, decks, reports. What the run learned is written to durable memory, not left in the transcript.
Back to the board. The next run starts already knowing this — it inherits the memory instead of relearning from scratch.
A personal operating layer for turning ambiguous goals into decomposed work, routed specialist agents, tracked outcomes, and reusable memory.
A repeatable system for turning customer strategy, architecture, and prototype context into polished executive-ready narratives.
A private research-to-output workflow for transforming messy notes, market research, and experiments into publishable ideas.
The loop
01
Capture
→02
Decompose
→03
Build
→04
Verify
→05
Remember
Design Principles
The site can show the shape of the work without exposing confidential customer, employer, or financial data.
Every workflow should leave behind files, tasks, summaries, or reusable context that the next run can build on.
Automation accelerates sourcing, drafting, and analysis; final calls stay explicit when reputation, money, or access is involved.
The value is not only in the code. It is in turning technical systems into stories people can understand and act on.
On the roadmap
Planned work, not yet live. These ship as the underlying systems mature.