Skip to content
AI Dev Cases
BoltAI

The Fourth Try After Three Failures — How Daniel Nguyen's BoltAI Turned “the Official ChatGPT UI Is Annoying” Into a Weekend MVP, a Pieter Levels Retweet, and a $30K/Month Portfolio

A native macOS AI client built by Daniel Nguyen in Ho Chi Minh City, Vietnam. With BYOK (bring your own API key), it brings multiple AI models into any app — Xcode, Apple Notes, you name it. After three earlier failures, a weekend MVP went viral off a Pieter Levels retweet, growing to 7,000+ paying customers and $30K/month, built mostly on one-time licenses — part of a small portfolio alongside PDF Pals and KTool.

Published: Sep 21, 20264 min readPrimary-source verified · 7
Monthly (est.)
$30k/mo
Time to grow
22 months
Users
7K
Launched
2023
Daniel NguyenDaniel Nguyen@daniel_nguyenxThe Fourth Try After Three Failures — How Daniel Nguyen's BoltAI Turned “the Official ChatGPT UI Is Annoying” Into a Weekend MVP, a Pieter Levels Retweet, and a $30K/Month Portfolio

Key takeaways

  • Treat past failures not as losses but as accumulated skill, a publishing habit, and community connections to carry forward
  • Pick a pain you feel yourself every day, and ship an MVP fast — a weekend is enough
  • Use BYOK (customers plug in their own API key) to carry no cost of goods, and keep margins high with a perpetual one-time license

The pain point, and how they found it

The pain wasn't discovered through user research — it came from Daniel's own daily annoyance. While working in Xcode or Apple Notes, he hated having to switch to a browser and open the official ChatGPT UI every time he had a question. He'd also already had three products (WordPress themes, the uptime-monitoring SaaS StatusBoard, and the Kindle-delivery tool KTool) fail to take off — so the turning point came from going back to the most immediate frustration he lived with every day, in the middle of feeling like he hadn't yet found a hit as an indie builder.

Background & product

BoltAI is a power-user client that brings multiple AI models — ChatGPT, Claude, Gemini and others — into one native macOS app. Beyond a chat window, its signature feature is calling AI "inline" inside any app, from Xcode to your notes app. It runs on BYOK (bring your own key), so usage fees go straight from the user to each AI provider.

It was built by Daniel Nguyen, based in Ho Chi Minh City, Vietnam. He earned his first online dollar roughly 15 years ago selling a plugin for open-source forum software, then dropped out of college to build a WordPress consulting business up to around $100K before shutting it down. He later worked as a product engineer at several startups while also trying an uptime-monitoring SaaS called StatusBoard, which he closed after six months — one of three attempts before BoltAI that didn't take off.

BoltAI's origin is unglamorous: to fix his own daily annoyance of switching between the official ChatGPT UI and native Mac apps, he built an MVP over a weekend and launched in May 2023. Growth was slow at first, until a demo video he posted on X got retweeted by Pieter Levels, sparking a viral moment and a rush of early-access signups.

From there, growth wasn't a straight line — it was compounded by continuously publishing revenue, customer counts and failures while building in public. One year into going independent (Nov 2023), his three tools — KTool, BoltAI and PDF Pals — had reached ~$32K combined. By September 2024, BoltAI alone hit $15K/month with 7,000+ paying customers, and by March 2025, $30K/month. BoltAI runs mainly on perpetual one-time licenses (starting around $19 and raised in stages), plus a subscription revenue stream added through a Setapp partnership. His own profile lists lifetime revenue at "$500K+."

Based in

Vietnam (Ho Chi Minh City)

From the founder (primary source)

Daniel Nguyen
Daniel Nguyen
@daniel_nguyenx
·

I made $1,162 so far with BoltAI v2. It’s not much but it’s honest work 😁

Image
Daniel Nguyen
Daniel Nguyen
@daniel_nguyenx

I made $0 from BoltAI v2 so far 😁 A few months ago, I decided to rebuild the product. It was at the stage where it’s hard to make any meaningful change. It would be impossible to fulfill my product vision with the old codebase. New users were churning because of performance

Image
Reply
Read 24 replies
BoltAI growth channels and tech stack

How it was built

  1. Pick the frustration felt every day — switching between the official ChatGPT UI and native Mac apps like Xcode or Apple Notes — as the pain point
  2. Build the MVP in a short window, a weekend, and launch in May 2023
  3. Make BYOK (users plug in their own API key) the core of the design, supporting models from multiple AI providers
  4. Anchor pricing on a perpetual one-time license (with a year of included updates) rather than a monthly subscription, raising it in stages from an initial ~$19
  5. Ship 100+ updates, continuously incorporating customer feedback
  6. Partner with Setapp to add a subscription revenue stream on top of direct one-time sales
What AI did
BoltAI builds no model of its own — it 'translates' existing AI from OpenAI, Anthropic, Google and others into a native macOS UI via BYOK, with AI cost pushed to the customer

How it got users

  1. Growth was slow after launch, but keep posting demo videos and progress on X regularly
  2. A posted demo video gets retweeted by Pieter Levels, sparking virality and a rush of early-access signups
  3. Keep publishing revenue, customer counts and failures via building in public, compounding trust and audience (three tools at $32K combined one year in → BoltAI alone at $15K/mo by Sep 2024 → $30K/mo by Mar 2025)
  4. Combine word of mouth, paid ads and Setapp distribution to keep adding paying customers (7,000+ by September 2024)
  5. Don't rely on BoltAI alone — run a portfolio of several small AI tools including PDF Pals and KTool in parallel
First users
Early-access signups driven by building in public on X and a retweet from Pieter Levels
Early ad spend
Mostly word of mouth and building in public early on (specific ad spend undisclosed); paid ads and Setapp distribution added later

The repeatable playbook

  1. 1Treat past failures not as losses but as accumulated skill, a publishing habit, and community connections to carry forward
  2. 2Pick a pain you feel yourself every day, and ship an MVP fast — a weekend is enough
  3. 3Use BYOK (customers plug in their own API key) to carry no cost of goods, and keep margins high with a perpetual one-time license
  4. 4Keep sharing progress and demos on X regularly, raising the odds an influencer notices you
  5. 5Don't rely solely on direct sales — add a platform like Setapp as a second distribution channel
  6. 6Don't bet everything on one product; run a portfolio of several small AI tools in parallel to spread risk
  7. 7Keep publishing numbers via building in public even when they're inconvenient, like a revenue dip after a v2 relaunch

Channels they actually used

The hard parts

Before BoltAI, Daniel had three attempts that didn't produce standout results: a WordPress consulting business (started from selling an open-source forum plugin) that he grew to around $100K before shutting down, the uptime-monitoring SaaS StatusBoard that he closed after six months, and the Kindle-delivery tool KTool, which also didn't take off. BoltAI itself hasn't been smooth sailing either — when he shipped v2 in 2026, he publicly restarted the revenue counter from "$1,162," and we have not been able to independently confirm whether the $30K/month run rate reported in March 2025 still holds. (The numeric field conservatively uses the March 2025 first-person report, and this uncertainty is disclosed in the body.)

Deep dive

【Deep dive】What makes BoltAI a useful case study isn't a flashy hockey-stick story — it's the reproducible structure of "someone who failed three times, and what they changed on the fourth try." Five patterns.

1. Each failure quietly banked an audience and a skill. Taken alone, WordPress consulting, StatusBoard and KTool each look like a project that didn't pan out. But along the way, Daniel accumulated real Swift/native-Mac-app skills, a habit of building in public on X, and real contact with the indie-hacker community. He could build BoltAI's MVP in a weekend precisely because KTool had already trained the muscle of "build small, ship fast." The failed projects weren't losses so much as the run-up to the shot that landed.

2. BYOK plus a perpetual one-time license is designed for zero cost of goods and maximum margin. Because users plug in their own API key, AI usage costs never hit Daniel's books. On top of that, he anchored pricing on a one-time purchase (with a year of included updates) instead of a monthly subscription. Like TypingMind, this combination is the classic indie playbook for staying high-margin and untethered from usage costs as a tiny team — but BoltAI then layered on a second channel, Setapp, riding another platform's review and promotion instead of relying solely on direct one-time sales, gaining new customer touchpoints without ad spend.

3. Going viral wasn't engineered — it was noticed because he posted every day. Pieter Levels' retweet can look like pure luck, but the precondition was that Daniel was already regularly sharing demo videos and product progress on X. Whether a well-known influencer notices you is largely chance, but this case makes an obvious truth unusually visible: luck only lands on people who keep showing up. Having an existing audience touchpoint before the viral moment is exactly what let him convert the spike into early-access signups.

4. Don't run on one leg — spread the risk across a portfolio. Daniel doesn't run on BoltAI alone; he also operates PDF Pals (AI chat with PDFs, 1,000+ paying customers) and KTool (sending articles to Kindle) in parallel. Each is niche on its own, but they share engineering setup, marketing know-how and customer base, so a slowdown in one is cushioned by the others. Even after BoltAI became the flagship, he kept the other products running instead of folding them — a form of risk management that itself came from having failed three times: don't bet everything on the one hit, keep several small bets running.

5. The honesty of continuing to publish numbers — including the inconvenient ones from the v2 transition. In November 2023, September 2024 and March 2025, Daniel consistently shared results in his own first-person Indie Hackers posts. But when he shipped BoltAI v2 in 2026, he also publicly posted on X that the app had made just "$1,162 so far" under v2 — a modest number next to the $30K/month run rate from March 2025, and one he chose not to hide. We likewise haven't been able to verify how monthly revenue has moved since the v2 transition relative to the March 2025 $30K figure, so we disclose that uncertainty here rather than assert a number we can't back up.

Cross-case “growth playbook” report (coming soon)

We're building a paid report that aggregates every case in this database: which acquisition tactics worked, in which categories, and how well — insights you can't see from a single story. Get notified first when it launches.

Related cases