
AI-Native Builder & Advisor
Apr 2025 – Present 3 min read
What I'm building after Pixly
These days I'm having fun building with frontier AI, and riding each new model to stay as close to the edge as I can.
The models are now good enough to do real work once the problem is framed clearly. That makes building faster and cheaper, but it also means average output is everywhere. So the scarce work moves up a level: knowing what to build, giving the models the right context, reviewing what comes back, and turning what works into systems you can reuse.
I stepped back from Pixly in spring 2025, after 9 years building it, to focus on family and explore that frontier. Right now that means building AI products and systems, advising founders, writing the occasional angel check, and chasing a toddler around SF.
I took a few months to relax first, and somewhere in there the tools crossed a line for me. Claude Code, Notion AI agents, MCP servers, Cursor, Wispr Flow. Suddenly I could sit down on a Saturday morning and ship something useful by lunch, with no dev team and no ticket queue. Just a clear frame, the right context and a bunch of models doing the heavy lifting.
That became the work.
What I'm actually doing
Three things, and they all feed each other.
Building. I use AI tools every day to design and ship practical systems for myself and people in my network: internal ops dashboards, booking flows, agent-driven workflows, marketing sites, prep automations and small products. Most of the source of truth lives in Notion, and Claude Code takes over whenever something needs real code.
Advising. I work with founders and operators on using AI in the actual flow of their business. I skip the AI strategy decks and start with a simpler question: what should stop being manual next Monday? Most of it is hands-on. We map the workflow, find the few places where an agent can help, build the first version together and add a review loop so the quality holds up. Sometimes that's a focused advisory sprint. Sometimes it's literally sitting next to someone over lunch, connecting the tools on their laptop, so they leave with their first real agent workflow.
Content. This part surprised me. I'm enjoying building in public, documenting the work and trading notes with the online community around AI workflows, tools and how company-building changes when the models are part of the operating system. Some of it is useful and practical, like Your AI Lunch Guy. Some of it is internet-native and weird, like Discomorphism. A lot of it is just fun little projects in my personal repos (schedulers, Notion tools, personal websites, sync scripts) that help me understand what this new way of working actually feels like.
Where this is going
I don't fully know yet, and I like that. After 9 years building one company, I wanted to give the work some room to tell me what it wants to become.
The thesis is getting clearer, though. Generic output keeps getting cheaper, so the best AI-native work will come from people and teams who can turn their context, taste and judgment into systems that get better every time they run. The nuance you've lived through is the part nobody else has (the "you have no idea's" are your advantage).
That's what I'm building toward.