A first-time iPhone developer built a photo vault app with coding agents in about three weeks. He says it made about $41K in a month. No ads: an agent writes the content that brings people in.
In today's issue:
The Build: a privacy app, a daily SEO loop, and a $39.99 lifetime plan
Money Moves: OpenAI's Friday shutdowns, DeepSeek's off-peak rates, Claude fast mode, Replit's switch
Try This: sell a "Friday deadline" model check to anyone you've built automations for
🔦 The Build
A serial founder shipped his first iOS app with Codex and Claude. He says it makes about $41K a month.
Who: Rasty Turek (@synopsi), maker of Vaultaire, per Starter Story.
What: an encrypted vault for photos, videos and files on iPhone. No account. You open it with a drawn pattern. He says users store things like crypto wallet keys, IDs and birth certificates.
How he built it: about three weeks, with Codex and Claude. "I literally haven't written a single line either of code or the text on the web."
Who pays: users, through the App Store. The app is free. Pro is $1.99 a week, $19.99 a year or $39.99 for life, per the App Store listing.
Revenue: about $41K in a month, with just over 60,000 downloads. He also says recent days run around $2,000, which would be higher. We quote the lower number. Self-reported.
Lifetime is the plan people pick. He says it's about 60% of purchases and, he thinks, "north of 80" percent of revenue. So a monthly-revenue dashboard makes the app look smaller than it is.
Our take: The app isn't the clever part. The search engine behind it is, and you can copy that. Two caveats. He's no beginner. A security company he was part of was acquired by Google, an entertainment company by Amazon, and he sold PEX last year. And lifetime buyers pay once. When the traffic stops, so does the revenue.
🛠️ The Playbook
Sell a fix for a real fear. People worry about losing private photos and keys. A one-time $39.99 fix is an easy yes.
Let agents do the first build. It was his first iOS app. He says the coding agents did all of it.
Start with a plain site. List the features. Then write guides for long, specific searches, like "how do I store my photos on iPhones."
Automate the keyword hunt. Every morning, a command-line tool he calls Gum pulls Google Keyword Planner data. It finds searches where his site shows up but gets no clicks. An agent writes one guide a day for those.
Link new guides from your best pages. He says Google picks up a new page in about three to four days. Put the download button where readers finish reading.
No paid ads. He says the web is now the majority of his traffic.
🧰 The Stack
Tool | Job |
|---|---|
Swift + UIKit, iCloud | The native iPhone app |
Codex, Claude | All the code and all the website copy |
Gum (command-line tool) | Pulls Keyword Planner and Google data for the agent |
Cloudflare | Listed in his stack; use not described |
He didn't share costs. The app has no servers, so there's no server bill to grow with users.
⚡ Money Moves
OpenAI switches off a batch of older models this Friday, Oct 23. Its deprecations page lists gpt-4, gpt-4-turbo, gpt-4o-2024-05-13, o1, o1-pro and o3-mini, with gpt-5.6-sol as the replacement. gpt-3.5-turbo and o4-mini move to gpt-5.6-terra. Fine-tuned models on those bases go too. Google shuts down its deep-research-pro-preview-12-2025 agent the same day, per its changelog. If a client's automation calls one of these, it breaks Friday. Fixing it is billable work (see Try This).
DeepSeek's cheap model costs half outside peak hours. On its pricing page, deepseek-flash costs $0.15 per million input tokens and $0.60 per million output off-peak. Peak rates are double: 01:00–04:00 and 06:00–10:00 UTC on weekdays. Weekends are off-peak. Run batch jobs outside those windows. Tell clients before sending their data to a China-based API.
Claude's fast mode is now open to every API user. On Oct 15, Anthropic took fast mode for Opus 5.5 out of research preview, per its release notes. No access request needed. It costs $8 per million input tokens and $40 per million output, double the standard rate, and it doesn't work with Batch, per the pricing page. If speed is part of what you sell, like live chat, price it as its own tier.
Replit's CEO says OpenAI now powers its agent by default. Amjad Masad called it "a decision based on both cost, partnership, and performance," per TechCrunch. Users can still pick Claude. If you sell apps built on Replit, re-run a standard build before you quote a fixed price.
🧪 Try This
Sell a "Friday deadline" model check.
If you've built AI automations for anyone, some may call a model that stops working on Oct 23. Here's a small, billable job.
List every model name in their workflows: Zapier or Make steps, scripts, and the API usage dashboard.
Check each one against OpenAI's deprecations table. Anything dated Oct 23 stops Friday.
Save 20 real inputs and outputs from the current setup.
Run them on the listed replacement and on one cheaper model. The listed swap isn't always the cheapest one that passes.
Re-price. On OpenAI's pricing page, per million tokens: Sol is $4 in and $20 out. Terra is $2 and $12. Luna is $0.20 and $1.20. A workflow using 5 million input and 1 million output tokens a month costs about $40 on Sol, $22 on Terra or $2.20 on Luna (our math).
Charge a flat fee for the check and the switch. OpenAI says Sol's promotional pricing runs "at least through November 21, 2026," so put a price-review date in the quote.
You're not selling tokens. You're making sure Friday is a normal day for your client.
💬 Before you go
Built something with AI that earns money? Email newsletter@aimoneybrief.com with what it is and what it makes. We might feature you, with your numbers labeled honestly.
Know someone who'd find this helpful? Forward it to them.
Was this forwarded to you? Subscribe here.
Income figures are reported by the builders and aren't guarantees. We label every number Verified, Reported or Self-reported. Prices are as listed by each company on the date of this issue. AI Money Brief is operated by TFS LLC.
AI Money Brief
Get the next one in your inbox.