Live
Agentic — this very site
The site you're on. A zero-build portfolio about AI agents, with a live playground, a capability explorer, and a light/dark theme, designed, secured, and shipped end-to-end.
Sijohn Johnson · Builder of AI-agents
What AI agents
make possible.
I'm Sijohn, and I build AI agents, software that plans, picks up real tools, and sees a task through to done. This is a plain tour of what they can do, next to a few things I've actually shipped with them, from a nine-agent investing system to apps on the Play Store.
What we can achieve
Six things agents are
Six things agents are
genuinely good at.
Pick one to see how it works in practice. Each is something an agent can carry end-to-end, with me stepping in only where a real judgment call matters.
What we've accomplished
A sense of the work.
A quick sense of the scale, and one real run below: how this very site went from a one-line request to shipped.
0
Lines of code written & reviewed
0
Tasks completed end-to-end
0
Runs that reached a shippable result
0
Faster than doing it by hand
The one-line ask
"Build an interactive site about AI agents." From that single request, the agent shaped a plan: the sections, a live agent playground, and a dark/light theme.
Built with real tools
It wrote the HTML, CSS, and vanilla JavaScript, ran a local server, and drove a headless browser to check every change at mobile and desktop widths.
Checked its own work
Accessibility, color contrast, and responsive layout were audited and fixed until there were zero violations. Nothing was called "done" until it actually worked.
Scanned for security
Automated agentic security scans sweep the site, its headers, and the content-security policy, catching exposed files, permissive CORS, and misconfigurations, then fixing what they find.
Shipped
Committed, reviewed in a pull request, and deployed to Cloudflare Pages behind a strict security policy with self-hosted fonts. The result is the page you're reading.
Interactive playground
Watch an agent think.
Choose a task and hit run. It's a simple simulation of how an agent breaks work into steps, uses tools, and reaches a result, right in your browser.
Select a task and press Run agent to begin.
How it works
The agent loop.
Every capable agent runs the same simple loop, the magic is in doing it reliably, with good judgment, thousands of times.
Think
Act
Observe
Deliver
-
Think
Reason about the goal and decide the next best action. The plan comes before the keystrokes.
-
Act
Use a real tool, edit a file, run a command, search the web, or call an API, then let it actually happen.
-
Observe
Read the result and learn what changed, so the next decision is grounded in what really happened.
-
Deliver
When the goal is met, verify the work and hand it off, with a clear account of what changed and why.
About me
Hi, I'm Sijohn Johnson.
I'm Sijohn, a builder of AI-agents. I like taking an idea and turning it into something that actually runs, a web app, a mobile app, or a set of agents working together.
What I care about most is the unglamorous part: making things reliable, secure, and genuinely nice to use. This site is one of those things, built end-to-end with agents. The playground above mirrors how I actually work, break the problem into steps, use real tools, check the result, and don't call it done until it is.
Selected work
Things I've built.
A few projects that show what agentic building looks like in practice, designed, built, and shipped end-to-end. More on the way.
Multi-agent
Private Market Portfolio
An autonomous investing system where nine specialized AI agents collaborate to run a private-market portfolio end-to-end, sourcing and screening opportunities, performing due diligence, modelling risk and valuation, and coordinating allocation decisions under a human's oversight.
On Google Play
Kids Learning App
A playful, gamified mobile app that helps children learn through interactive lessons, rewards, and progress tracking, published and live on the Google Play Store.
Mobile app
Personal Trainer App
A fitness companion that builds personalized workout plans, tracks progress, and keeps users motivated with tailored guidance and adaptive routines.
Security
Agentic Security Scanner
An automated agent that audits a live site end-to-end, security headers, content-security policy, CORS, exposed files, and accessibility, then reports and fixes what it finds. It hardened this very site.
Your idea, built.
Have something an agent could build, a site, a tool, an automation? Let's make it real.
Start a project →
Get in touch
Let's build something.
Have an idea, a role, or a project in mind? Send a note, I'd love to hear it.
Have something you want built?
If there's an idea you've been meaning to ship, a site, a tool, an app, or a workflow you'd love to automate, I'd be glad to help. This site is a small example of what that can look like.
Back to the top