Building with AI
A few things I've built, automated and figured out with AI.
01
Onchain Trading Wrap
A consumer-style annual wrap for onchain traders, inspired by products like Spotify Wrapped and YouTube Recap.
What I built
- Took the idea from concept to a live product.
- Built the frontend in Lovable and connected it to live 0xPPL APIs.
- Designed the core logic around trader personas, using wallet activity to determine which persona best fit each trader.
- Used Claude to help build, test and refine the product and the underlying logic.
Outcome
~1,000 wraps generated with zero paid spend.
See it live →
02
AI Content Engine
A system I built to make my day-to-day content work faster and more consistent.
How it works
- Connected Claude to the live X feed through MCP so it could work with current information.
- Gave it our positioning, audience, and voice, plus the accounts I wanted it to learn from, so drafts start in context.
- Set up a competitor tracker that turns their best posts into a weekly set of ideas mapped to formats.
- Used Claude to generate variations, then reviewed and edited the final output myself.
- Added guardrails around common AI phrasing and writing patterns.
Outcome
12.8K followers and 1.8M+ impressions, alongside the broader growth work I was doing at 0xPPL.
03
Feed Scoring Diagnosis
Investigating why our trending-token feed was missing some of the activity users cared about.
What I did
- Compared our feed against competitors to identify where we were consistently late.
- Used AI to analyze the differences and find patterns across the tokens we were missing.
- Dug into the scoring logic behind our feed to understand why those tokens weren't being picked up earlier.
- Tested different scenarios against the existing logic to identify where the biggest gaps were.
- Proposed changes to the scoring thresholds and worked with engineering on the improvements.
Outcome
The changes improved the feed and made the feature more useful for users.
04
Slack + Email Daily Brief
A daily workflow that pulled the information I needed from Slack and email into one place.
How it works
- Connected Slack and email through MCP.
- Pulled together the conversations, messages and tasks that needed attention.
- Worked out what context Claude actually needed to understand what mattered.
- Structured the output in priority order: what needed a reply, what could wait, and what to keep in mind, with a short summary.
Outcome
45 to 60 minutes saved every morning.