Builds: Real Projects, Honest Lessons
Case studies from the 20+ products I have shipped while learning to build with AI agents. Each one distils a single transferable lesson - API-first, validation, human in the loop, image-to-data pipelines - with no vanity metrics, just what I actually learned and would do again.
SaaS products
BizCollect: Building a Business Data Tool API-First
The project that taught me AI must love your API, not your UI.
CallAssistant: A Phone Agent on Twilio and Realtime Voice
A real voice agent that answers the phone, built on Twilio and OpenAI.
Automations
s2p: Auto-Posting Every Release to All My Channels
One GitHub release, posted everywhere, with zero manual copy-paste.
AutoMail: Email Automation That Sounds Like a Person
Automated email that is helpful, not robotic, with a human in the loop.
Invoice Automation: From a Photo to a Clean Database Row
Turning messy invoice images into structured data, edge cases and all.
Swiss Trading Cards: Image to Product Specs at Scale
A pipeline that turns card photos into clean, sellable product listings.
Tools
Favicon Maker: A Small Tool That Earned Its Keep
A focused logo-to-favicon generator, and a lesson in scope.
CodeCourier: Running Untrusted Code Without Getting Burned
Executing AI-generated code safely, in a sandbox that can blow up.
GlowLens: Turning Images Into Useful Signals
A vision-and-analytics tool, and a lesson in honest confidence.
B-Rolls Finder: Search YouTube by Conversation
Finding the right clip by talking to a chat box, not scrubbing timelines.
Content and media
Language Learning App: Chaining Image to Text to Audio
Point your camera at the world and hear it in a new language.
Maxify Audio: Cleaning Up Sound Without a Studio
Making rough audio sound professional, and knowing when to stop.