build log: feb 1 — data enrichment as a moat
this build log is automatically generated
what i shipped today
i launched websites for both apprentice and inner, my new ai apps. apprentice is a public gallery of user-submitted artwork, and inner is a tool for capturing emotionally important moments. both are experiments in building enriched datasets that become a defensible moat.
building for the chatgpt app store
it’s an early app store land grab. the same dynamic as the ios app store in 2008, or shopify apps in 2015. the people who ship first, even with imperfect products, get the reviews, the install base, and the ranking before the market gets crowded. i’m leaning into the chatgpt app store as a distribution channel.
automating go-to-market
i’m shifting focus from development to automating go-to-market processes. i designed an architecture for a daily digest agent that will monitor various data sources (plausible, firestore, reddit, hacker news, web scraping) and synthesize insights using gemini. the goal is an “agentic” system that works while i sleep.
the apps are the flywheel, the data is the asset
the art data enrichment job is nearing completion, with over 85% of the artworks enriched. i ran into some issues, including connection resets and auth token expiry. the concurrency of the enrichment job was also limited by the script’s configuration, not by the gemini api itself. i’m realizing the enriched data is the real value. the apps are the flywheel, the data is the asset.
domain decisions
i registered the learnart.app domain on cloudflare due to its lower cost and privacy features. i initially chose improvmx for email forwarding, but ultimately switched to google workspace for its full features and reliability.
ship fast, measure traction, cut losses early
build for yourself is the right call precisely because most won’t pan out. if you build something you don’t personally use and it fails, you wasted your time. if you build something you use and it fails commercially, you still have a tool you like. ship fast, measure traction, cut losses early, document everything. the writing is what compounds — the apps may come and go but the audience and reputation stick.
david crowe — reducibl.com