Building in public. One person, AI-native, every industry. Writing about what actually works when you stop hiring and start building.
cloudflare
I spent three months paying for VPS hosting before discovering that Cloudflare will connect any machine to the internet for free. No public IP required. No port forwarding. No dynamic DNS hacks. Just a daemon that runs on your box and a five-line config file. This is the most underutilized
infrastructure
I just replaced three hosting providers with a Mac Mini sitting on my desk. Not as a dev server. Not for testing. As the actual production server, serving a real website to real visitors over HTTPS, with CDN caching, DDoS protection, and automatic SSL renewal. Total monthly cost: forty-two cents.
clients
A prospect evaluates your proposal. The work is exactly what they need, the price is right, and the timeline is better than anyone else's. Then they ask how many people are on your team. You tell them the truth. It's you. Something shifts in the conversation.
strategy
Solo builders with AI can operate at extraordinary scale. But not every market rewards that. Some industries are structured in ways that make a one-person operation natural. Others have barriers that no amount of efficiency can overcome. Choosing the right market is the highest-leverage decision you'll make. Build
strategy
When a ten-person team hits a scaling problem, the most common response is to hire person number eleven. When a solo builder hits the same problem, they have to find a different answer. That constraint turns out to be an advantage. The Team Scaling Model Teams scale by adding capacity.
quality
Everyone has an AI horror story. The chatbot that fabricated a legal citation. The summary that reversed the meaning of the original document. The confidently delivered answer that was completely, provably wrong. These stories are real. They're not exaggerated, and they're not going away. AI systems
getting-started
The gap between "I should use AI more" and actually having a system that works without you is where most solo builders stall. They read about 10x productivity. They nod along. Then they open a chat window, ask a question, get an answer, and go back to doing
reviews
Retrieval-augmented generation sounds straightforward when you describe it. Give an AI access to your documents so it can answer questions about them. In practice, building that system yourself involves choosing embedding models, configuring vector databases, writing chunking logic, managing document processing pipelines, and wiring everything together. Most solo builders who
reviews
This site has spent several articles talking about workflows, triggers, and systems that run without being asked. At some point, the question becomes practical: what do you actually build them in? For solo builders who aren't primarily software developers, the answer increasingly is n8n. n8n is an open-source
reviews
Running a model locally is the first step. Using it productively every day is a different problem. The command line works for testing. It doesn't work for the kind of sustained, document-heavy, multi-conversation workflows that solo builders actually need. At some point, you need an interface. Open WebUI
reviews
When Meta released the first Llama models, the immediate reaction was surprise. A trillion-dollar company was giving away the kind of technology that others charged premium prices to access. The skeptics assumed there would be a catch. The catch never materialized. Instead, what materialized was an ecosystem. Llama is Meta&
reviews
The assumption that AI requires an API key and a monthly bill is so widespread that most builders never question it. You sign up for a service, feed it tokens, and pay per request. It works. But for a growing category of tasks, there's a simpler option: running