No CS degree, no network, from Bangladesh, full-time job + Uber. I built 2 AI products solo and just got into the ChatGPT App Directory.
I want to share this honestly, the wins and the parts that still scare me, because I never see stories from people who started where I did.
I grew up in Bangladesh. Did my BBA and MBA there. I've always been ambitious, the restless kind. After my MBA I started a small distribution business. Then in 2022 I moved to the US on an H1B, completely new country, new culture, new everything. Nobody around me works in SaaS. I have no CS degree, no connection to the tech world, no network in this industry. I didn't even know the community existed.
But I kept digging, always looking for a way to build something of my own.
Then AI happened.
At first I just used it to write applications and answer small questions. Then I slowly started to feel the real power of it. Before Claude Code existed, I was writing code by pasting between ChatGPT and Gemini, copying errors back and forth, fixing bugs by hand, half the time not knowing what I was doing. It was messy. But I shipped my first prototype that way, a beauty AI app called GlowyAura.
Then I found Claude Code, and that changed everything. Suddenly one person with no CS background could actually build real software. So I did.
I built two products:
HookMafia (mid-March 2026), a web app that helps creators write viral hooks and scripts and analyze their own TikTok performance.
HookLayer (May 14), the same engine but as an MCP server that runs directly inside Claude and ChatGPT. Score a hook, analyze a creator, predict virality, all inside the AI assistant.
I built all of it solo. Nights and weekends. While working a full-time job. While driving Uber for extra income. No co-founder, no team, no funding.
And it wasn't just the code. With AI, I became the whole team. I was the developer, the product analyst, the SEO and AEO specialist, the growth person, and the support agent. I built my own tracking dashboard on PostHog and Google Analytics to see which pages and channels actually lead to activation and payment, then optimized around what the data showed. Every role a startup normally hires for, I did alone, with AI as my leverage.
Where it is now:
1,227 users, 578 in the last 30 days
On August 14, HookLayer got approved into the ChatGPT App Directory. Signups went from ~7/day to 50+/day, all organic, zero ad spend
1,578 HookLayer tool calls, 798 in the last 7 days alone, more than doubling week over week
MRR grew $38 (June) to $115 (July) to $174 (August)
12 paying customers, 6 active subscriptions
Claude connector in final review with Anthropic
The part I'm proud of: when the ChatGPT wave hit, only 10% of people who connected actually used it. I dug into my own data, found the causes (ChatGPT needs a fresh chat to activate a connector, plus a permissions bug that was hiding my best tools), and shipped fixes the same night. Activation went from 10% to 39% in four days.
The part that scares me: HookLayer has a lot of usage but I haven't cracked usage-based monetization yet. 22 users hit the free credit wall and didn't buy. I'm shipping a "watch this creator, see what changed" retention loop to give people a reason to come back and keep using it. We'll see if it works.
People ask what my team looks like. It's me, at night, with AI doing the work of a developer, a data analyst, and a growth team.
I'm not writing this because I made it. I haven't. I'm writing it because a year ago I believed people like me, no technical background, no network, from a completely different world, couldn't build real software.
That's not true anymore. The tools changed. If you have the will and you're willing to learn in public, you can build.
Still a long way to go. But for the first time, the direction is clearly up.
Happy to answer anything, the stack, the ChatGPT App Directory approval process, the mistakes, how I use Claude Code, whatever helps.
Congrats - and the 10% to 39% activation jump in four days is quietly the best number in the post. Most people would have just counted the ChatGPT traffic and moved on.
The "22 users hit the free credit wall and didn't buy" part is the one I'd be obsessing over too. When they hit that wall, are they walking away because the credits feel too expensive, or because it's not clear what buying actually unlocks? Those need opposite fixes - and it's usually the second one.
This is the honest version of the solo-founder story people actually need. Rooting for the retention loop :3
thanks a lot, at this stage i really need this kind of feedback and encouragement.
you nailed the exact thing i keep going back and forth on. on the wall, i think you're right that it's more the second one, people arent sure what paying actually unlocks, not that its too expensive. the ones who DO pay are the heavy users who already saw the value, so the gap is making that value obvious before they hit the wall, not after.
one pattern i keep seeing: people resist paying for something inside an MCP/AI assistant more than they would for a normal dashboard, feels like a feature of the tool they already pay for. but i've started getting paid users and even repeat buyers on the agent side, so its improving. that repeat behavior is the signal im watching most.
and yeah, being solo means i can turn anything on a dime, no approval, no meetings. that speed is the one real advantage of doing this alone. appreciate you rooting for the retention loop, means a lot :)
It sounds like you've built quite an interesting tool with GlowyAura! Leveraging AI to analyze creators and predict virality is a compelling niche, especially in today’s content landscape.
When I first started my journey, I faced similar challenges no CS degree either, and building a network from scratch felt daunting. Focus on what you can control, which is creating a strong foundation for your product. Here are a few things that might help you as you continue to refine GlowyAura:
User Feedback: Since you’re covering a complex process (analysis and prediction), I recommend actively seeking feedback from early users. It might reveal what features are most valuable and where users get confused. You don’t always need a formal user testing group; even chatting with acquaintances can provide valuable insights.
SEO Optimization: Considering your tool’s focus on virality, make sure you’re implementing SEO best practices in your content marketing strategy. This will help you not only attract users but also create authority in your niche. When I was working on my own SEO tools, I noticed that publishing case studies of how my product improved metrics for users really helped with organic visibility.
Showcase Results: If your app provides specific metrics or results, showcase those! Whether it's an increase in engagement or improved reach, tangible results can catch the attention of your audience. When I shared metrics from beta users, it significantly boosted interest in my tool.
Content Strategy: Beyond your app's core capabilities, consider how to position GlowyAura in the content creation workflow. Can it integrate with existing strategies or platforms that creators already use? This alignment could strengthen its value proposition significantly.
Community Engagement: Join relevant forums or communities (like this one!) to establish connections and share insights. Building relationships can often lead to unexpected opportunities and partnerships that can help you scale your app.
It’s a challenging path, but staying focused on solving user pain points and leveraging feedback will serve you well. Wishing you the best with GlowyAura!
thanks for the thoughtful reply! small correction, the products are HookMafia (web) and HookLayer (the MCP tool) — GlowyAura was an earlier project of mine. but your points all land, especially on showcasing real metrics and case studies. that's actually something i've been leaning into. appreciate you taking the time.
Getting into the ChatGPT app directory as a solo dev without a network is a real distribution unlock; that's usually the hardest part once the product itself works. Curious how you're thinking about discoverability beyond the directory listing itself, since placement alone doesn't guarantee ongoing traffic once ranking or featured spots shift. Did you build in any way to track how much of your usage is actually coming from that channel vs. organic?
great question, and you're hitting the exact thing that keeps me up at night. directory placement got me the initial wave but you're right that it's not durable, rankings shift and featured spots move, so treating the listing as "distribution solved" would be a mistake.
on tracking channel attribution: yeah, i built my own dashboard on posthog + GA to see where usage actually originates. the honest answer is it's messier for the MCP product than for a normal web app, because when someone uses it inside ChatGPT or Claude, i don't get clean referrer data the way i would from a web visit. what i CAN see is the shape of it, spikes that line up with directory placement vs steady organic baseline, and i tag key creation and first-call events so i can tell "directory wave" cohorts from people who found me through SEO/content.
what i've found is the directory is more of an acquisition spike than a durable channel, so i'm treating it as top-of-funnel and building the durable stuff separately: SEO/AEO content that ranks, and a retention layer so users come back on their own instead of depending on me staying featured. the real unlock i think isn't ranking higher in one directory, it's being the thing the assistants recommend when someone asks a question, which is a different game than directory placement.
still very much figuring it out though, distribution is the part i'm least confident in. how did you approach the durability problem?