MNMatthew Nolan

Founder · Builder · Systems thinker

Matthew NolanI build the things I wish existed.

I’m a self-taught builder. I turn problems into systems, products and businesses, while working a full-time day job.

Explore the work Meet Matthew Let’s build something

The short version

Thirty seconds on who I am.

  • Builder

    I learn whatever the problem requires.

  • Founder

    I turn problems into products and businesses.

  • Systems thinker

    I look for the system underneath the problem.

  • Self-taught

    I didn’t wait for someone to teach me.

  • Experimenter

    Some ideas work. Some don’t. Both teach me something.

  • Night shift

    I build after the day job ends.

Proof of work

Can you see what he actually made? Yes.

Real screenshots of live sites, captured Oct 8, 2026. Click through and check.

The liveloadless.com homepage: “Stop the spiral. Find your next clear move.”
liveloadless.com, captured Oct 8, 2026. The live pages still carry the ClarityOS product name.

LIVE LOADLESS™ case study

The LOADLESS app’s first screen: one text box asking what is sitting heaviest in your head right now.
app.liveloadless.com: the first screen asks one question. No account needed to start.

LIVE LOADLESS™ case study

The Crestlake Studio homepage: “Every agency hides the price. We publish it,” next to a live project estimator.
crestlakestudio.com, captured Oct 8, 2026.

LIVE Crestlake Studio case study

The Paul Williams Insurance homepage: “Protect what matters most” over a photo of a Florida neighborhood.
Client build: paul-williams-insurance.pages.dev, captured Oct 8, 2026.

LIVE Crestlake Studio case study

Things I’ve built

The work, with honest status labels.

  1. 01 LOADLESS™LIVE Cognitive-load reduction. My primary venture. A live product with real auth and live Stripe billing, built by one person at night. Pre-revenue, and honest about it. Cloudflare Pages · Supabase · Postgres · Stripe · TypeScript Case study
  2. 02 NALAIN USE The governance system I run my work on. A personal operating system for decisions: claims need evidence, agents earn trust, and nothing gets called “done” until it’s checked three times. Scheduled agents · Local LLMs · Supabase · Cloudflare · Markdown Case study
  3. 03 The agent fleetIN USE 19 scheduled agents running my operations. 19 active scheduled agents handle status digests, metrics and scans. 14 more are paused or retired. Keeping the count honest is part of the job. Cron · Local LLMs · APIs · Shell Case study
  4. 04 Crestlake StudioLIVE My design and web studio. Websites for local businesses, with prices published in the open. Client work includes a live insurance agency site with intake forms and an admin dashboard. HTML/CSS · React · Vite · Supabase · Cloudflare Pages Case study

All work, including experiments and the one I shelved

Built without a safety net

I figured out how to build.

I work a full-time day job in public works maintenance. I don’t have a venture fund, a development team, or a roadmap someone handed me. I get about two hours a night.

I learned what I needed to learn and built anyway.

So far: a live venture, a working app with payments, a self-built AI governance system, 19 autonomous agents and a design portfolio. Built in the margins of a $35K-a-year job.

See it for yourself

How I think

Operating principles, with receipts.

These come from the system I actually run my work on. They’re evolving, not universal laws.

  1. 01

    Build before permission

    I learn by building the thing. Waiting to feel qualified is how nothing gets made.

    Evidence Every project on this site was built without a team, funding or a formal background in software.

  2. 02

    Systems over symptoms

    When something keeps breaking, the fix is usually one level up. Find the system that produces the problem.

    Evidence The shared-state file got overwritten once. The fix wasn’t a backup. It was a rule: append, never replace.

  3. 03

    Truth before helpfulness

    Never manufacture certainty to give an answer. “I don’t know yet” is a complete sentence.

    Evidence This site reports no user numbers, because there aren’t any worth reporting yet.

  4. 04

    Evidence before claims

    Every claim carries an evidence tag. If it can’t be verified, it gets labeled, or it gets cut.

    Evidence Look at the small tags across this site. Each one points at its proof.

  5. 05

    Reduce complexity

    Good systems make hard things easier to understand and to run. If it needs a manual, it isn’t done.

    Evidence LOADLESS’s first screen is one question and one text box.

  6. 06

    Human + AI

    AI widens what one person can do. It doesn’t replace judgment. Money, strategy and “is this true?” stay with me.

    Evidence NALA’s agents escalate anything financial or strategic. Always.

Things that didn’t work

A person who has built nothing has nothing to fail.

Shelved Oct 3, 2026

I shipped a product nobody asked for

What happened
ClarityOS launched as a complete product: site, app, sign-in, payments. It got zero traffic and zero interest.
What it taught me
Shipped is not the same as wanted. I’d spent my hours on the product and almost none on finding the people it was for.
What changed
I shelved it, moved its best ideas into LOADLESS, and now treat distribution as part of the build.
Related work

Sept 15, 2026

A session overwrote my system’s memory

What happened
My AI sessions share one state file. One of them replaced the whole thing (28K of infrastructure history) with a single note on one topic.
What it taught me
Any file that many writers touch needs a rule, not just a backup.
What changed
I recovered it from a session backup and added a rule at the top of the file: append, never replace. Every session reads that rule first.
Related work

All of them

Now

What has my attention.

Updated Oct 8, 2026.

Building

  • LOADLESS. Getting the live product in front of the people it’s for.

Exploring

  • How to sell a product built to ask for less attention.
  • Where agents should stop and ask a human.

Learning

  • Moving into a full-time role in tech. This site is part of that move.
  • Postgres row-level security, done properly.
  • Running language models locally instead of renting them.

Thinking about

  • What trust looks like between a person and an agent that acts for them.
  • Whether one person can run a real company on agents without losing the plot.

If you think like this, let’s talk.

I’m looking for people who are building something, solving something, or seeing a problem differently.