Hydraux: post-mortem of an 11-day indie bet
I shut down Hydraux on Day 11. Here's what I learned.
The bet
I spent the last 11 days building Hydraux — a paid React component pack for Shopify Hydrogen, Next.js Commerce, and Remix storefronts. The pitch: "production-ready commerce UX systems with the operational depth most kits skip." Subscriptions, bundles, partial fulfillment, the messy states real merchants hit but every component library glosses over.
Pricing was $149 individual / $399 team / $999 agency for the paid pack. A free Pack 0 served as the lead magnet. A preorder mechanic — $99 capped at 25 customers — was set to launch on Day 14 (today).
The thesis had teeth on paper: AI commoditizes generic UI components, but "judgment layer" docs (CONVERSION_NOTES, ANTI_PATTERNS, EDGE_CASES, INTEGRATION) embedded with each component would be defensible. Agencies pay for de-risked client delivery; indie devs pay for time compression.
I built fast: brand locked, domain live behind Cloudflare, full Astro + React landing page with a live React-island demo, Lemon Squeezy store with 3 products configured across multiple tiers, ConvertKit/Kit account with a 3-email welcome sequence wired through the native LS integration, GitHub org spun up, full strategic documentation.
What I didn't build
A single conversation with a real Hydrogen agency. Zero cold DMs. Zero validation calls. Zero outreach of any kind.
Eleven days. Zero outreach.
The Day 11 stress-test
I ran the bet through a structured multi-advisor council debate. The verdict was uncomfortable but unambiguous:
- 11 days of infrastructure without a sales call is "elaborate procrastination dressed as execution."
- The kill criterion I'd set ("<2 paying customers by Day 60") was pre-negotiating with reality.
- Pack 1 build math was fiction — 150+ hours of solo work in 26 days while also doing distribution and support.
- The "agency-vs-indie" funnel contradicted itself: strategy said "agencies buy on referral" while the plan was cold DMs.
- The preorder shape was wrong: $2,475 ceiling on 150+ hours = $16/hour with massive failure exposure.
Then I checked the question no advisor had directly asked: how many subscribers were on my Kit list right now?
Zero.
Not zero qualified subscribers. Zero subscribers, period. Eleven days of having a live site, a working free product, and a complete brand had attracted no one.
That's the test result that ended it.
What I got wrong
Building before validating. I treated infrastructure as evidence of progress. It wasn't. A working Lemon Squeezy checkout and a Kit automation are technical capabilities, not demand signals. Demand signals are people answering yes to specific questions about their actual pain. I asked nobody.
Mistaking AI velocity for product velocity. I could build with AI assistance at roughly 5x normal pace. That made it feel like I was running fast. I was — in the wrong direction. AI lets you outrun demand by 10x. That's a tax, not a feature. The faster you can build, the more careful you have to be about what you build.
Pricing a thing no one had said yes to. $149 / $399 / $999 tiers were anchored to competitor pricing (Tailwind UI Commerce at $349), not to anything a real customer had said they'd pay. Competitor anchoring without a track record is fantasy.
Setting a kill criterion that wasn't a real kill criterion. "<2 paying customers by Day 60" wasn't a hard test. It was a permission slip to ride hope for 60 days.
Assuming the wedge translated to demand. "Operational depth others skip" is a real insight about the commerce-UX market. Real insights aren't products. Products are insights someone has named a price for, in their own words, before you wrote any code.
What I learned
- Validation precedes production. Always. If I can't get 5 real conversations in 5 days, the product dies before I write a line of code.
- Preorders are not validation. Preorders are commitments. Sell them only after demand is proven.
- The infrastructure is the asset. The setup took 11 days and $30. It now takes me 2 days to ship a new landing page for the next bet. That's the actual return.
- Solo founders should sell first, build to spec second. Reverse order = burnout futures.
- "Judgment layer" as a moat is real but unprovable from zero. The moat becomes real only after shipped production deployments demonstrate it. Before that, it's marketing copy.
- Cost of killing fast is cheap. Cost of killing slow is the difference. $30 + 11 days vs. another 30 days of building things no one asked for + 25 refund-risk preorder customers.
What's next
I'm pivoting the infrastructure to a service-first bet: paid Hydrogen Commerce UX audits. $500–1500 per audit, ~5 days of work each, no product to build until I've done 5+ paid engagements and see what patterns repeat across real client codebases.
Same expertise — the 30-state PDP checklist I built for Hydraux IS the audit framework — but inverted order. Clients pay before any deliverable exists. Demand validation is built into the business model.
If five audits land in 30 days and patterns repeat clearly, a product pack might re-emerge — specced from real client engagements, not from what I imagined agencies needed.
What stays available
- AddToCartButton with the full 13-file doc layer. MIT-licensed. GitHub.
- The 30-state PDP self-audit checklist at /checklist. Use it on your own storefront.
- This post-mortem. Take whichever lesson is useful.
Final note
I'm not sad about the 11 days. I built real infrastructure, ran a real stress-test, killed cleanly, and learned exactly the pattern I needed to break. That's $30 + 11-day tuition for a transferable lesson. Cheap.
If you're a Hydrogen / Next.js Commerce / Remix agency or indie dev shipping commerce storefronts, and you'd want a paid audit of your PDP/checkout/account UX against the 30-state checklist — DM me or email hello@hydraux.dev. I start audits next week.
— Alex