Built in a Day — How Tony Dinh’s “Better UI for ChatGPT” TypingMind Reached $1M Lifetime Revenue in 20 Months by Building in Public

TypingMind is a power-user front-end for LLMs — a single interface for ChatGPT, Claude, Gemini and more. It’s faster than the official apps, lets you organize chats into folders, save and reuse prompts, and switch models in one click. Its defining trait is BYOK (bring your own key): users plug in their own API key and pay the AI providers directly for usage. So TypingMind carries no API cost of goods and charges purely for the UI and workflow value.

It was built by Tony Dinh (@tdinh_me), a Vietnamese solo developer. Formerly a $105K/year software engineer, he quit to go independent, and had already shipped multiple products — the Twitter analytics tool Black Magic, the macOS screenshot tool Xnapper, and the developer utility DevUtils.

How TypingMind was born is telling. In March 2023, right after OpenAI released the ChatGPT API, Tony built the first version in a single day. Search, folders, custom prompts, multi-model support, BYOK — features every vendor now ships as standard — it was one of the very first tools to offer them, ahead of the official UI. A March 11 Product Hunt launch took #1 (product of the day), and with a push from his Twitter and newsletter audience, revenue jumped from $10K to $22K in a day, hitting $22K in the first 7 days.

What followed reads less like a viral spike than compounding. It reached $500K in cumulative revenue within a year, and crossed $1M by month 20 (November 2024) — with zero outside funding and a small team throughout. TypingMind is a flagship example of the most reproducible indie playbook: not ‘build frontier AI yourself,’ but ‘translate clunky official AI into a delightful product experience.’

From the founder (primary source)

The repeatable playbook

1. 1Pick a pain you feel yourself every day — make yourself user #1
2. 2Ship a minimal product in the first *days* after a new API/model launches; speed over polish
3. 3Use BYOK (customer plugs in their own API key) to push AI cost and usage-based spend onto the customer and keep margins high
4. 4Grow an X/newsletter audience via building in public *before* launch, then self-start velocity alongside Product Hunt
5. 5Publish revenue, internals, and failures continuously to compound trust and distribution
6. 6Take fast cash and validation with one-time pricing, then shift to subscriptions (Team plan) once demand is proven
7. 7Sell off products whose cost structure breaks or that stop growing, and concentrate resources on the one that’s growing

Deep dive

【Deep dive】TypingMind is a teaching case for indie builders because the *order and structure* are copyable, not just the size of the numbers. Five patterns.

■ 1. Stand at the trend’s starting line ‘in a day’ — timing is the biggest lever. The ChatGPT API launch was the moment heavy users worldwide simultaneously realized ‘the official UI is inconvenient.’ Tony’s edge wasn’t technical sophistication but placing a minimal product there *fastest*. He could build it in a day because his hands were already fluent in his own React/Next.js/Tailwind stack from prior Twitter tools — no time lost researching from scratch. The lesson: in the first *days* after a new API or model ships, speed beats feature count. Wait for polish and someone else who noticed the same pain gets there first.

■ 2. BYOK erases both cost-of-goods and scaling risk at once. Bring-your-own-key isn’t an implementation shortcut — it’s the core of the business model. Because users pay the AI providers directly, (1) TypingMind carries almost no token cost and keeps high margins, (2) more users don’t inflate its costs linearly via API billing, and (3) the price is justified purely by UI value. The biggest enemy of running an AI product solo is API cost and usage-based spend; BYOK shifts that to the customer by design. That’s the foundation that let a small team stay high-margin.

■ 3. Ignite the launch only *after* building the audience. Product Hunt #1 and $22K in the first 7 days weren’t luck — they were the payoff of prep. Tony had grown a Twitter and newsletter audience by building in public well before launch, so on launch day he had a crowd he could mobilize — driving both the PH votes and day-one sales himself. Reverse it — launch on PH as a nobody — and you often get zero initial velocity because the algorithm never picks you up. Audience first, launch second.

■ 4. The compounding of building in public — sharing revenue *is* the marketing. Tony grew from ~100 followers to 180K+ by continuously sharing revenue, internals, and failures. Publishing the numbers itself compounds into (a) trust (a real, selling product), (b) distribution (milestone posts spread), and (c) a built-in early audience for the next product. TypingMind’s acquisition channel is effectively ‘his own X’ because he spent years stacking that asset instead of buying ads. For indie builders, building in public isn’t a side hobby — it can be the highest-ROI distribution network you own.

■ 5. Shift quietly from one-time to subscription. TypingMind began as a one-time license. There’s a standard critique that ‘buy-once cannibalizes future subscription revenue,’ but Tony’s real data refutes it — license purchases rose for 20 straight months and never dipped. On top of that he worked to migrate customers to the Team (subscription) plan, and by 2024 revenue was roughly a 50/50 split of one-time and subscription. Take fast cash and validation with buy-once, then, once demand is proven, layer in recurring revenue — a clean way to resolve the cash-vs-recurring tension by sequencing.

■ Knowing when to cut and concentrate. Easy to miss: Tony didn’t grow purely by addition. When Twitter’s steep API price hike broke the cost structure of his former flagship Black Magic, he didn’t force it alive — he sold it and concentrated resources on the one product that was growing, TypingMind. For a solopreneur juggling multiple products, deciding *which to fold and which to bet on* shapes growth as much as the ability to build new things.