Day 14 - BuildEngine is live

Today, May 4, 2026, I sold something for real money for the first time under the Northvane LLC banner.

The customer was me. The charge was $49. I refunded it within twenty minutes to keep the books clean. So technically, lifetime revenue stands at zero.

But the entire fulfillment chain ran end-to-end against live infrastructure: real Stripe charge, webhook firing with a 200, Supabase logging the order, Resend delivering the welcome email from hello@buildengine.tech, the signed download URL minting a Quick Start Guide PDF that opened cleanly in my Gmail.

That chain working - that is what two weeks of incorporation, infrastructure, and code was for.

This is the first entry. Every Tuesday I'll send you what's working, what's breaking, what I learned, and the unflattering numbers. No hype. No "AI is transforming everything" filler. Just the actual work.

Here is what the first two weeks looked like.

Week 1: Paperwork

You cannot ship a product without a legal entity that can take money. I underestimated how much of the first week is just bureaucratic plumbing.

I formed Northvane LLC as a Wyoming anonymous LLC through Northwest Registered Agent. $143 for the first year, including registered agent service. Wyoming because the disclosure requirements are minimal and the annual fees are reasonable. Anonymous because BuildEngine, FinancePulse, and the Phase 2 brands coming this summer need to operate as their own personalities, without a single founder name attached to all of them in public records.

EIN issued by the IRS the same day, online. Mercury Business Checking opened with a $1,000 capital contribution. Anytime Mailbox in Austin for the registered business address (the Wyoming registered agent's address shouldn't appear on customer-facing materials). Xero Early subscription connected to Mercury for bookkeeping ($2.50/month promo, $25/month after).

Stripe was the long pole. Application submitted with the Wyoming address, EIN, full owner verification, expected business model. Approval took 48 hours. The statement descriptor question matters more than people realize: customers see this on their card statement and decide whether to remember the charge or dispute it. Mine reads BUILDENGINE, not NORTHVANE*BLDENGINE. Brand-specific, not parent-LLC-prefixed. When FinancePulse launches next week, I'll add it as a DBA on the same Stripe account, and its descriptor will read FINANCEPULSE.

Total stack burn at this stage: somewhere between $120 and $380 a month depending on usage. Domain at Cloudflare ($10/year), Vercel Pro ($20/month - required for multi-brand deployments), Supabase free tier (good for the first ~50 customers per project), Resend free tier (1,000 emails/day), Anthropic API funded with auto-reload at $50/$200 monthly cap.

If you are wondering whether $400 a month is too much for a business that has not made any money yet - yes, it feels uncomfortable. It also represents the smallest plausible footprint to operate a real software business. Spreadsheets won't process payments. Free-tier email sending won't survive a real launch. There is a floor. This is mine.

Week 2: Building

The technical work happened in week two, on top of the legal foundation from week one.

The BuildEngine site at buildengine.tech is a Next.js 15 app deployed on Vercel, backed by Supabase Postgres, with Stripe handling checkout and Resend handling transactional email. The five products in the catalog are the Quick Start Guide ($49 one-time), the Idea Validator ($99 one-time), Top 10 Trending Ideas ($99/month, weekly refresh), the Both Reports Bundle ($149), and the Founder Starter Pack ($199).

The Quick Start Guide is what I wish someone had handed me on day one. 61 pages, written and rewritten this week. It documents the exact path from "I want to start a business" to "I have a Stripe account taking real charges" - the legal entity, the bank account, the payment processor, the email infrastructure, the deployment platform, the AI tooling, every step including the rough edges.

I wrote it because the current "build with AI" content is mostly aspirational. The actual experience is friction-heavy, and the friction is in places no one writes about: setting your statement descriptor before approval, the AVS quirks of business debit cards on online checkout, the SPF record collisions when adding a second mail provider, the Supabase Data API exposure setting that gates everything underneath it. Most build-in-public content skips this and jumps to "and then I had a SaaS." Mine doesn't.

Phase 1 of the codebase is about 14,000 lines of TypeScript across the marketing site, five product pages, Stripe webhook handlers, Supabase access layer, and email rendering. 78 tests in CI, all green. I lean on Claude Code as my developer. It writes the diffs; I review every single one before committing. We have caught each other's mistakes about evenly so far.

The pricing locked in around mid-week. $49 for the guide. $99 for one report. $149 for the bundle. Tier 1 subscription will be $69/month, founding rate $39/month for the first cohort. Tier 2 at $399/month, founding rate $299/month for the first 50 customers. The pricing is calibrated to the actual time and AI cost it takes to deliver, not to "what the market will bear" - I'd rather charge less than risk one customer feeling overcharged on day two.

FinancePulse is the next brand. Code is complete (229 tests passing), awaiting deployment and DNS configuration. It is an investment signal aggregator pulling from nine data sources to identify cross-source patterns - the kind of thing I want as an individual investor and have not seen done well anywhere. Three products: Signal Report ($29), Watchlist Pulse ($49), Sector Snapshot ($29). Same architecture as BuildEngine, applied to a different domain, with the same anonymous-public-face approach.

What I am not doing

I am not letting AI start the business. There is a fashionable narrative that says you can prompt your way to a business - describe an idea, get a website from Lovable, push it to launch, watch the customers roll in. The result is generic, indefensible, and competes with a thousand identical variants the next day.

That is not what this is. AI is the developer, the researcher, the assistant. I am the strategic decision maker. I read the diffs. I evaluate the output. I push back when AI is wrong, which is more often than its confidence suggests. The business has a defensible reason to exist that comes from understanding the market and the customer - not from outsourcing strategy to a general-purpose model.

The bundle of products, with maintained data and operator judgment behind each one, is the moat. Anybody can prompt their way to a single SaaS landing page. Nobody clones a portfolio of eight products with proprietary trending corpora and operator credibility on a weekend.

What is next

Tomorrow morning, the launch tweet from @AIBuildWright. Quick Start Guide goes public. We see if anyone besides me is interested.

This week, the trending corpus pipeline that powers the Idea Validator and Top 10 Trending Ideas - Reddit, Hacker News, and Product Hunt scrapers feeding a structured database that AI cross-references against customer-submitted ideas. This is the moat for those two products. Without it, they would be commoditized "ask Claude" wrappers. With it, they ground analysis in current market signals.

Next week, FinancePulse goes live, and I begin the SM Posting Automation tool to keep three brands posting daily without burning me out as a solo operator.

Later this summer, the Phase 2 brands - PatentScan, InterviewIntel, ClinicalCompass, and a grants search tool. Each one cheaper to build than the last, because the first one paid the infrastructure tax.

Subscribing

This newsletter is free. It goes out every Tuesday morning. If you want what I shared above - real numbers, real friction, no hype - keep reading. Reply to any email and I will see it (the address is support@buildengine.tech and it is one human, not a queue).

If you want the playbook condensed into one document, the Quick Start Guide is at buildengine.tech. $49. What I would hand a friend who asked how I built this.

Talk Tuesday.

• Build Wright