No employees. No agency. No funding. One founder in Omaha, one AI named Fable, and about $5 a month of infrastructure. This page is the receipts — every step, every dollar, every mistake, updated as we go.
Last updated 31 July 2026 · generated from the build log
The business: Friday afternoon to Saturday morning, 17–18 July 2026. The app: two weeks later. Scroll the log inside the panel.
Streaming subscriptions idle 8–10 months a year while shows are out of season. Pitch: an app that toggles your subscriptions around your watchlist. Instruction to the AI: validate it — and if it passes, we build.
Market research: 23% of US streamers already rotate subscriptions by hand; manual rotators save $600–900/yr. Competitive gap confirmed — nobody owns content-aware rotation. Verdict: go.
Full product spec, three pricing tiers, AWS architecture, 10-week MVP plan.
Landing page, waitlist API, CDN, DNS, TLS — serving production traffic ~75 minutes after the spec was approved.
Every signup becomes a CRM contact and a pipeline opportunity. GA4/GTM provisioned by API. Not without casualties — see the mistakes.
Push-to-deploy pipeline with smoke tests, security headers, email-spoofing protection on the domain. First pipeline run: green in 51 seconds.
Stripe subscriptions live: three tiers plus a founding membership at $59.99/yr, price-locked for life. A simulated customer's full lifecycle — subscribe, pay, cancel — verified end to end.
bingi.me/calculator — see what your own idle months cost. Plus a referral loop: 3 friends join, you get a year free.
Sending, receiving, and a branded 3-email welcome sequence — built on AWS for ~$0/mo instead of another SaaS subscription.
The AI signs up a test address, and the automated welcome email loops back through our own mail pipeline where the AI reads and verifies it. Land → calculate → join → nurture → checkout: fully automated.
The actual product starts: a mobile app with working sign-up, a database that costs ~$0 when idle, and a nightly catalog-ingest pipeline. First real user row in the database before lunch.
Branded share cards on every page, and Bingi arrives on YouTube, Instagram and Facebook — plus a mobile-friendly menu after the founder caught the nav wrapping on his phone. The mistakes list grew by two the same day.
The plan for one polished trailer becomes something better: Trent & Bingi, a build-in-public series. Episode 0 gets written, shot, cut and finished in a single marathon sitting.
Episode 0 hits the feeds and sends its first real traffic (~500 Facebook views → 34 visits). The same day, the 10-day working-app deadline passes without a working app. It goes in the log instead of getting quietly re-dated — and the homepage countdown, which had been auto-flipping to a celebratory "Shipped" state, gets fixed. A clock can't verify that software works, so it no longer claims it does.
Deadline take two, and the app arrives in one long day: the API, watchlist, search and release calendar, the rotation engine that decides what to keep each month, reminders that actually toggle a subscription and bank the savings, and billing. The ten-week MVP plan from day one — built.
The first Android build finishes on Expo's servers: a real, installable app. Thirteen days earlier, Bingi was a landing page.
First full day of testing on a real phone — three builds, and a polish pass the device demanded: safe-area handling, bigger touch targets, and a Home dashboard that opens on today's decision. Fourteen days from idea to working app. Next stop: private beta for the waitlist, in order — and the App Store.
Season 2 is about finding customers: getting the app in front of people who don't already know us, what actually converts, and what it costs to earn each signup. It gets logged here the same way — including the misses.
The build-in-public series — deadlines, misses and all.
What it actually costs — and what it would cost you to repeat it.
| Inherited asset | Replication cost |
|---|---|
| CRM + marketing automation (agency account) | $97–297/mo, or free-tier substitutes + elbow grease |
| AI teammate (Claude subscription — this is Fable) | ~$100–200/mo — the single most load-bearing line item |
| An existing inbox to forward into | $0 — any Gmail works |
| Years of internal playbooks (deploy standards, integration templates) | The AI regenerates most of it on demand |
| Total to repeat this | ≈ $130–250/mo + a weekend |
Every build log with no failures in it is lying. Ours, in order.
Secure config lookups kept insisting parameters didn't exist — they did. The culprit was the local shell silently rewriting paths before AWS ever saw them. Lesson: when your infrastructure seems to lie, suspect your own tools first.
Our CRM's tokens lock their permissions at the moment they're created — updating permissions in the dashboard does nothing for existing tokens. Thirty minutes of confusion, fixed by rotating the key.
Our first analytics property was accidentally provisioned under a sibling brand's account. Deleted, corporate structure decided properly, re-provisioned. Lesson: decide where things live before creating them.
Payment processing has a 30¢ minimum per charge — which is about 20% of a $1.99 subscription. We repriced the same evening to annual-first billing. Lesson: run the fee maths before you publish prices.
A Windows console couldn't print 🍿 in a success message and took the script down with it. The popcorn is load-bearing brand. The console lost.
Rebuilding the navigation for mobile dropped the styles that made our logo look like a logo — it rendered as an underlined blue hyperlink, straight out of 1996. The founder caught it in minutes. Lesson: a shared component has to carry all of its own styling, not borrow from the page around it.
We fixed the nav — and anyone who'd already visited would have kept the broken version for a day, because browsers cache site assets locally and no server-side purge can reach them. Lesson: when a shared file changes, change its URL too.
The entire org chart.
Omaha, Nebraska. Runs Marigold ONE11, a one-person digital studio. Sets direction, makes the calls, owns the mistakes and the wins. Yes, he's on the waitlist for his own product — #1.
Anthropic's Claude (Fable 5), working in Claude Code. Research, architecture, code, infrastructure, integrations, copywriting, QA, and this page. Doesn't sleep, which explains the 5 AM commits.
The MVP is built and running — coming soon to the App Store. Private beta goes to the waitlist first, in order.
Join the waitlist Become a founding memberBingi is what happens when one founder and one AI take a weekend seriously. The studio behind Bingi helps people launch like this.