TypeScript-first product builder · AI-tooling tinkerer · self-hosting enthusiast · occasional timezone mystery

I turn ambitious ideas into useful software—usually late at night, often from a new country,
and almost certainly with coffee within reach.

$ whoami
clez — developer, product builder, professional rabbit-hole explorer

$ pwd
/somewhere/on/earth/with/wifi

$ focus
saas · ai agents · mcp · developer tools · self-hosting

$ uptime
sleep_schedule: beta

I build across the whole product stack: interfaces, APIs, databases, background jobs, authentication, integrations, deployment pipelines, and the operational tooling that keeps everything alive after launch.

Most of my commercial SaaS work lives in private repositories. The code may be private; the architecture decisions, deployment lessons, and debugging scars are very real.

My favorite projects sit where product thinking, developer experience, and automation overlap. I like difficult systems that feel simple to use—and tools that explain themselves when something goes wrong.

Once, my math teacher told me:

“You're too dumb to be successful.”

That left me with two sensible options:

1. Prove them wrong.
2. Become successful enough that nobody asks me to calculate anything without a computer.

So naturally, I became a developer.

The plan is going surprisingly well—although centering a <div> can still become a deeply personal experience.

What started as curiosity became a habit: understand the system, remove the repetitive work, make the dangerous parts explicit, and keep iterating until the rough idea behaves like a real product.

Complete products rather than isolated features:

• Next.js and React interfaces with TypeScript end to end
• APIs, databases, validation, authentication, and permissions
• Dashboards, admin tools, workflows, and background jobs
• Third-party integrations and automation
• Deployment pipelines and self-hosted infrastructure
• Bugs that only appear in production on Friday evening

MCP servers, model-routing profiles, context-engineering experiments, and agent workflows that connect models to useful systems.

Asking an AI to write a poem is fun. Asking it to diagnose infrastructure at 3AM is a lifestyle.

Deployment automation, diagnostics, lifecycle management, and safer operational tooling. I want infrastructure that gives people control without requiring a ceremonial sacrifice to the DevOps gods.

Installers that install. Uninstallers that actually uninstall. Commands that are safe to rerun. Errors that tell you what happened and what to do next.

“Something went wrong” is not an error message. It is a cry for help.

A large part of how I research, plan, structure, and ship software is powered by GSD Core.

GSD keeps the creative idea tornado on speaking terms with production. It turns vague ambition into research, specifications, executable phases, verification, and—against all statistical expectations—shipped software.

Without structure, many ideas would still live in a folder named:

final-project-v2-new-FINAL-this-one-really

My interpretation of the workflow:

💡 Get an idea
🔍 Research everything
📝 Create a suspiciously detailed plan
💻 Build at 3AM
🧪 Discover an edge case nobody asked for
🔧 Fix it anyway
🚀 Ship
🌍 Change country
🔁 Repeat

Or, more accurately:

Git. Get distracted. Return at 3AM. Ship. Somehow Done.

Thank you to everyone behind open-gsd for building the framework behind much of what I build. 💜

The stack changes when the problem demands it. Strong technical opinions are useful; installing a new framework because it became popular on Tuesday is slightly less useful.

Make the difficult thing understandable.
Make the repeated thing automatic.
Make the dangerous thing explicit.
Make the error message useful.
Make the rollback path boring.

I like software that:

• can be installed without an ancient Stack Overflow answer;
• explains what went wrong and suggests a next step;
• is safe to run more than once;
• can be removed without an archaeological expedition;
• makes destructive actions obvious and deliberate;
• survives the journey from “works on my machine” to “works outside my timezone.”

SaaS            → focused solutions to specific real-world problems
AI agents       → stronger context, safer actions, better workflows
MCP             → connecting models to systems that do useful work
Self-hosting    → ownership without unnecessary operational misery
Developer tools → less setup, clearer feedback, fewer surprises
Travel          → how unreliable hotel Wi-Fi can become
Sleep           → backlog

Near the top of the rabbit-hole queue:

• Cursor-native developer tooling
• phase-aware model selection and verification
• MCP infrastructure for self-hosted platforms
• deterministic context handling for coding agents
• private SaaS systems with maintainable operational foundations