Project updates · August 2026
Where Are They Now
Built with Claude in 28 Days documents four ventures and a daily AI comic strip
as they existed at the end of a four-week build. The book was a snapshot. The projects did not
stop when the manuscript did. This page is the running record of what each one has become,
plus the two new ventures that came out of the work: which grew, which was parked, and why.
From 11K products to 20K+ unique SKUs across 125+ retailers in 20 product categories
covering 140+ major consumer electronics brands. The platform now includes an MCP server
for AI agent integration, an AI-powered search tool, an enterprise API, and a
“Price Now” feature that solves the data staleness problem common to price
comparison sites. What started as a weekend project in the book is now becoming an
authoritative source of Canadian pricing intelligence.
Visit HomeGadgets.ca →
A new venture born from the HomeGadgets infrastructure. ShopButler is a framework for
consumer-driven agentic shopping: AI agents that can research, compare, and act on behalf
of shoppers. This is the thesis of the book’s Platform Layer (Tier 3) playing out in
real time. Claude as infrastructure inside a product, not just a tool used to build it.
Visit ShopButler.ai →
An open document standard for job applications. Candidates generate a single portable
.credentia.json file containing their complete professional record.
Employers receive structured, verified data that their systems can read without parsing
errors. The schema is open source under Apache 2.0, the candidate tool is free, and
employers pay per job posting. Built on the same observation that drove every project in
this book: a broken system that everyone accepts because fixing it used to be too
expensive.
Visit Credentia.one →
Parked. The grocery comparison model proved technically viable but the economics of
food and grocery price data do not justify the infrastructure cost at this stage. The
catalog normalization lessons were absorbed into HomeGadgets. The site remains live as
a reference.
Visit BasketBrain.ca →
Expanded beyond the original three studies. New signal types added. Sector-specific
testing across finance, insurance, and healthcare to validate whether the wide-market
hypothesis holds at the industry level. The research platform continues to grow.
Visit ContextQuant.com →
Still publishing daily. The automated pipeline described in the book continues to run
every morning at 8am UTC without intervention.
Visit Neural Strip →
Pre-pilot. Waiting for bank partner.
Visit MerchantLink →
The book captured the first 28 days. The projects kept going. That is the point —
capability unlocked is not a one-time event. It compounds.
← Back to Built with Claude in 28 Days