Writing Pilot
What It Was
Writing Pilot started with a simple question: what if you had GitHub Copilot, but for everything you type on the web?
The idea was a browser extension that would watch what you were writing in any text field — emails, forms, social posts, anything — and offer inline AI completions you could accept with a single key. No switching tabs, no copy-pasting from ChatGPT. Just a suggestion sitting right there in your input, waiting for you to press tab.
It was a genuinely exciting problem to work on. And I built a lot of it. But it didn't survive contact with the market — twice.
What I Built
By the time I shut it down, Writing Pilot was a fully functioning product in production:
-
A browser extension built with WXT that could read selected text, read full page content, and make tool calls back to the server
-
Persistent memory and chat history — the extension remembered context across sessions
-
A custom Next.js backend handling auth, API routing, and user management
-
Stripe billing fully integrated — subscription tiers, payment dashboard, the works
-
An admin dashboard for monitoring users, managing accounts, and reviewing usage
-
All of it live, deployed, and working
The extension itself was more capable than it sounds. It wasn't just autocomplete — it could understand what page you were on, what you had selected, and use that as context for its responses. Think of it as a lightweight AI agent that lived in your browser sidebar.
The Pivots
Pivot one: Google does this better
The original concept — inline completions in any text field, globally — fell apart when I realized Google was already doing exactly that natively in Chrome, and doing it better than I could with an extension. You can't out-Google Google when Google ships it as a default browser feature.
So I shifted. Instead of competing with inline completions, I repositioned toward a Perplexity-style browser AI — a sidebar assistant that could read your current page, answer questions about it, help you draft responses, and remember context across your browsing session. More of a thinking tool than an autocomplete tool.
Pivot two: Chrome ships native AI
Then Google announced Chrome's built-in AI features. A first-party, free, deeply integrated AI assistant in the browser itself. At that point, the value proposition I was building toward was gone. I couldn't compete with free models shipped by the company that owns the browser.
I made the call to shut it down.
What I Got Wrong
Looking back, the biggest mistake wasn't the pivots — the pivots were the right calls. The mistake was what I built before validating the idea.
I spent months building the website, the payment system, the admin dashboard, the backend infrastructure. All of it polished. All of it working. And I did almost none of that before confirming there was a durable value proposition that competitors couldn't easily kill.
The lesson: infrastructure is not validation. A working Stripe integration doesn't prove anyone wants your product. I built the engine before I knew where I was driving.
Next time, I ship the smallest possible thing that proves the core value — before I touch billing, dashboards, or auth.
What It Taught Me
-
Competitive landscape research is not optional. Both pivots were things I could have caught earlier with more deliberate market research before writing a line of code.
-
Value proposition first, infrastructure second. I inverted this and paid for it in months of work that had to be abandoned.
-
Knowing when to stop is a skill. The temptation when you've built something real is to keep pushing. Recognizing that the moat was gone and making a clean decision to shut down — that was the right call, and I'd make it again.
The code was good. The product thinking came too late.
Project discontinued August 2025 - These are my own experiences, decisions, and lessons. I used AI to help organize and articulate them.
