A website for a small business, specified, built under my direction, and tested before it goes live. AI automation is the next offer: a workflow that keeps running.
01 — Websites
Free sample first.
Taking my first client projects. I build a free sample before you pay anything.
Website
$1,000 one-time
You own the domain and the logins.
How a website gets built
Scroll to move through the steps.
02 — AI automation
This is the second offer. I quote each workflow to the job.
AI automation
Custom quote
There is no fixed price.
A repeating job: a schedule, a handoff, a check, a record. I write what the system may and may not do. Anything that sends a message or spends money waits for my approval.
Send my project detailsSystems I run
These are my own projects. Each pair of pictures is a screenshot of the live page, desktop and phone. The full write-up, including a paper-trading test that showed no edge, is on my portfolio.
I founded this
A Web3 radio station on X that has to stay on the air around the clock. Live Spaces drop, and every show has to be handed to the next host. Doing that by hand, day and night, does not hold.
I automated the schedule, the show handoffs, and the site. An operations agent, Orion, posts the daily schedule at 6 a.m., warns hosts 30 and 5 minutes before a show, and hands the broadcast to the next host. If the stream drops, a keep-alive brings it back. I directed the AI-assisted build and I test it.
A handoff counts only when the next host is confirmed as co-host. Routine health checks are plain scripts. Dangerous commands need my approval. Credentials stay in a vault the agent cannot read. Every incident becomes a written procedure.
It broadcasts 24/7, with 25+ shows a week, as listed on spaces.media. My portfolio also lists thousands of hours on air and about 5.4k followers on X.
Public site
Trading-bot sellers post screenshots. Those leave out losers, idle bots, and fees, so a visitor cannot see what a bot actually kept.
A public leaderboard of about 133 Polymarket trading bots. Each one is scored from its wallet’s public on-chain history, after trading fees, including rebates. Losing and idle bots stay listed. An automated pipeline refreshes the board. People can submit a bot, and there is a newsletter sign-up. I directed the AI-assisted build and I test it.
The methodology is published on the site. Each bot is labeled Confirmed, Self-claimed, or Suspected. The page states the sampling bias: bots taken from top lists over-represent winners. Only public data is used.
The board is live at botscoreboard.com and tracks about 133 bots. The page shows when the data was last refreshed, and the methodology sits on the site.
The public record
The systems needed a public record of the spec, the controls, and the result, including a result that did not work.
I write and maintain mgimaging.github.io/portfolio. Each case study states the problem, what I built, the controls, and what happened. I keep the page current.
Spec first. Hard limits written down. Human approval before anything sends or spends. A log of decisions and incidents. Results reported as they came out.
It is public. It includes a paper-trading test of 37 pre-registered strategies. None showed an edge. Paper only. No real money.
03 — How I work
The same order every time. I do not ship a system I have not specified and then checked myself.
Every system starts as a written specification: the goal, the inputs, the success checks, and what it must never do. For a website that is the pages, the contact path, and who owns the domain and the logins. For a workflow it is the schedule, the checks, and a hard stop. Nothing sends a message or spends money until a person approves it.
AI coding agents build from that spec. I steer the work, read the pages and the code they produce, and send back anything that drifts. I design the system. I do not pass along a result I have not read.
I open the result in a browser, on a desktop and a phone, before it goes live. For a running system I also check the logs and the approval gate. A send or a spend that skipped approval does not ship. When a test shows no edge, I report that. I do not loosen a limit to force a better-looking result.
04 — Standards
I work to code, and I inspect before sign-off. The same rule I use in the weld shop is the rule I use here.
Any outgoing message or payment waits in a queue until I approve it.
05 — About
I am a certified structural welder (AWS D1.1) at Fisher Industries in North Dakota. I work to code and inspect before sign-off. I use that same discipline on software.
I previously did 3D exterior renderings and have since moved fully into web and automation.
I also publish the tests that fail. I published a paper-trading test of 37 pre-registered strategies. None showed an edge. It was paper only. No real money.
Contact
Taking my first client projects. I build a free sample before you pay anything. Email is the way in.
