APPROACH
With Claude Code, I built the full backend and shared memory architecture based on first principles thinking. I designed a custom dashboard with a new design system that optimized for efficiency over brand expression without losing the core brand personality. Then, I shaped the conversational experience of how Mr. Oyster communicates across Discord and the dashboard.
Connect siloed pillars into one decision system
Instead of rebuilding what existing tools already do well, I built a system where live data from Shopify, Instagram, TickTick, and Discord flows into one central engine. Mr. Oyster analyzes the cross-platform data to propose ranked actions on Discord and the dashboard, while my partner and I make the final decisions and provide feedback.
Turn data into trustworthy recommendations
To ensure we can trust the recommendations, I optimized for calibrated trust. Every decision shows the hard data and reasoning behind it based on urgency, budget, and time, and how it ladders to our overarching goals across pillars. If a reasoning can't be traced back to real data, the recommendation never reaches us.
AI-assisted collection planning
I built a way to manage multiple collections at once and visually curate our collected objects. By pulling data from Shopify, Mr. Oyster assists with thematic sourcing, recommending relevant items to add based on the collection's theme. Meanwhile, the "Ideas" tab gives me a dedicated space to paste images of objects I want to source later.
IMPACT
Unified data system
Consolidated our scattered data into a single engine. By forming a data-informed recommendation that gets pushed as a task in TickTick, we don't need to jump between tools to form an insight anymore. It only gets better over time as it learns.
Strategic budgeting
Shifted from impulse buying to intentional, thematic sourcing. The system now automatically aligns every potential object with our active budget and creative goals.
LEARNINGS
Working in large organizations taught me to design at scale: aligning cross-org initiatives, connected features, and scaling across platforms globally. While that foundational systems thinking drove this work, building this agent pushed that skill to a another level. By wiring our isolated data sources together, I learned how to design holistically at the data layer that improves through feedback loops. Removing those technical blockers taught me that when you build the system to handle the busywork, you actually get your time back for the fun parts of running a business, like sourcing and creative direction.
WHAT'S NEXT
I'm building out the dashboard to measure how these recommended actions directly drive brand growth. I'm also curious about experimenting with different models, like Astra and Grok Bot, to see how running a small business can be truly effortless.
Big thanks to
<3
@sashaveryovka
Set design
@aleckvenegas
Editorial photography