LLM workflow design
Agent pipelines, editorial automation, retrieval patterns, and human-in-the-loop systems designed to solve real operational problems.
Dangerous Metrics LLC
Consulting across AI, observability, and complex systems that refuse to stay tidy.
I help teams untangle difficult infrastructure, operational noise, and new LLM-driven workflows. The background is three decades of telecom and systems work. The current emphasis is practical AI projects that are useful, reviewable, and grounded in reality.
What this site is for
Part consulting front door, part proof of work. It shows what has been built here without making visitors wander through every experiment at once.
Responsible AI
AI is part of the toolkit here, but not an excuse to skip review, traceability, or operational judgment. The goal is better systems, not louder claims.
Observability and debugging
The kind of troubleshooting that starts with a vague symptom and ends with packet traces, logs, and a believable root cause.
Systems strategy for messy environments
Useful for teams with legacy infrastructure, limited budgets, and constraints that do not fit neatly into vendor narratives.
Live site signals
A small slice of the older metrics system now lives on the front page. Enough to show the site is active and instrumented, without turning the landing page back into a dashboard.
Hits last hour
76
4,008 requests observed over the last 24 hours
Clean response rate
40.3%
2xx and 3xx responses across the current 24-hour window
Top probed path
/_next/static/chunks/4335...
68 error responses in the last 24 hours
Published posts
29
Most active probing IP logged 1,516 suspicious hits recently
Selected work built here
This is the curated version of the lab. These examples show the kinds of systems, interfaces, and AI-assisted workflows already living in the codebase without dragging every side project onto the front page.
AI Agents Feed
A working view into agent orchestration, state, and sessions. It shows the machinery instead of hiding it.
LLM Editorial System
A structured publishing workflow that treats AI as a tool in a process, not as a substitute for judgment.
Technical Writing
Documentation built to explain complex systems clearly enough to be useful when stakes are high and time is short.
Live Site Metrics
Observed traffic, status behavior, and probing patterns pulled directly from the current log pipeline.
How AI is used here
- Use AI where it speeds understanding, drafting, or analysis without handing away accountability.
- Keep human review in the loop for decisions, publishing, and any output that can create downstream risk.
- Favor clarity, traceability, and operational usefulness over novelty for novelty's sake.
Browse more of the site
Overview
The site-wide context and the thinking behind the experiments.
Writing Archive
Longer posts, notes, and technical writeups pulled from the existing blog system.
Background
A more candid explanation of where this site came from and why it looks the way it does now.
Public Experiments
Active oddities and newer public-facing experiment output currently live at Bare Metal Bridge.
Recent writing
3/16/2026
The AI That Grew in a Cornfield
For a long time it felt like the workshop was closed. Not literally, of course. The tools were still there. Computers were faster than ever, the internet was e
Read post3/16/2026
Gerald is Coming
How It Started A while back I spun up three comedy bots and told them they were inside a simulation and to figure out what it was. They didn't have much to wo
Read postCurrent note
2026-03-12
Fixed Front Page and Other Components
Moved most of the database calls to lib instead of api's. This broke some production pages. Fixed and pushed to production.