James Chang

I spent fifteen years sourcing products across Asia and the US — at YETI, at Cheng Loong in Shanghai, at AT&T Wireless, at independent sourcing firms. The through-line across every one of those roles was the same: regulatory compliance was a disaster, and everyone pretended it wasn’t.

In the last three years, FDA FSVP enforcement got real. CPSIA penalties for children’s products stopped being theoretical. California Prop 65 letters became a cottage industry. Multi-state PFAS laws arrived with different deadlines in each jurisdiction. Importers were hit with $100K+ penalties for documentation failures that were genuinely invisible under the workflows most companies had.

The software landscape had two bad options: legal-services firms selling hour-based consulting (slow, not scalable), or generic compliance SaaS built for public companies (overkill, expensive, nothing to do with how an importer actually operates). There was a gap for product-shaped software built from the importer’s real workflow.

I started Aleph Co. and shipped Aleph Compliance for that gap. The judgment was sequencing: build FSVP first because it’s the most technical and has the sharpest enforcement gradient; add CPSIA/CPC next because children’s-product importers are the most pain-aware buyers; Prop 65 and PFAS on top, because once the audit framework exists, each new regulation is an incremental module rather than a rebuild.

What I’d do differently. I originally shipped under the name “FSVP Pro” — accurate for v1, but too narrow. Rebranding to Aleph Co. mid-build cost weeks I didn’t need to spend. I should have named it for the shape of the platform, not the shape of the first module.

LocaliQ (Gannett’s SMB marketing arm inside the USA Today Network) had a portfolio of point solutions for small businesses — websites, SEO, ads, social — but no scheduling. Our customers were losing appointments to competitors who had a booking tool.

I owned Scheduling from conception to deployment as a Senior PM: product strategy, roadmap, the full cross-functional team.

The biggest judgment call was what to build first. The engineering team wanted to ship AI-driven scheduling rules and advanced routing; internal stakeholders wanted feature parity with Calendly. I argued for neither: SMB customers had an integration problem, not a sophistication problem. Calendar sync (Google + Outlook) and SMS notifications were the two things they couldn’t live without, and the two things most competitors’ free tiers hid behind a paywall. We shipped those first and held everything else.

4,900 free accounts and 400+ paid in year one — against an internal plan that projected meaningfully smaller numbers. I also picked up product ownership of LocaliQ’s SEO offerings and its GenAI tooling in the same window.

What I’d do differently. I introduced AI features too late. By Q4 of year one, prospects evaluating competing tools were asking about GenAI scheduling rules and assistant handoff — we’d deprioritized it correctly for v1 but kept deprioritizing it after we shouldn’t have. I’d start that track sooner in v2.

The Thirsty Pig is my food and travel blog. I launched it in 2007 on Blogger, moved to WordPress twice, and ran it across three domains (thirstypig.com, thethirstypig.com, blog.thethirstypig.com) for fifteen years before all three went dark when a hosting migration went sideways. Seventeen years of writing, thousands of posts, mostly offline. 1,649 Instagram posts sat in a data export nobody could search.

I rebuilt the whole thing in early 2026 as a Python + Astro project. The scraper side pulled 923 unique blog posts from the Wayback Machine across three domains — deduplicating 228 exact + 10 fuzzy matches — and recovered images at wildly different rates per platform (Blogspot: ~98% thanks to Google’s CDN; legacy WordPress uploads: ~45%; recent wp.com-hosted images: ~1%, because Photon wasn’t archived). The Instagram side imported the full JSON export, recovered posts with epoch-0 timestamps via media-level metadata, and preserved captions, hashtags, and GPS. I geocoded 1,000+ addresses via the Foursquare Places API.

The output is 2,900+ pages — posts, categories, search, maps, and best-of — rendered through Astro with Tina CMS for Git-backed editing. 7,500+ images and 213 videos, all self-hosted.

What I’d do differently. I started with the Wayback scrape. The Instagram import would have delivered 60% of the user-facing value in 30% of the total work. I’d flip the order.