Everyone thought i was wasting my time. now i have 2 AI products and a ChatGPT App Directory approval.
When i started building, everyone around me thought i was playing. no CS degree. no network in tech. business background, not engineering. i genuinely had no idea what i was doing.
i started this journey about a year and a half ago, teaching myself to build with AI. early on i was pasting code between chatbots, fixing bugs by hand, no clue what i was doing. i shipped rough prototypes that way. then the tools got better and everything accelerated.
in the AI era you're a "solopreneur" on paper, but in practice i run a company of one. AI is my developer team, my marketing team, my product analyst, my finance team, my research team. i make the calls, AI helps me execute every function a startup normally hires for.
my two current products, HookMafia and HookLayer, are about 5 months old. here's where they are:
2 live AI products (HookMafia web app + HookLayer, an MCP server that runs inside ChatGPT and Claude)
1,500+ users across 40+ countries
revenue growing ~500% in 3 months (still small, ~$232 MRR, but the curve is real)
approved into the ChatGPT App Directory
the honest parts, because this is IH and you deserve the real version:
revenue is still tiny. this doesn't pay me yet. i still work a full time job.
most of my agent-product users use it heavily for free and don't convert. i'm slowly getting paid users and even repeat buyers, but monetizing an MCP tool inside ChatGPT is harder than a normal dashboard. people resist paying "inside" a tool they already pay for.
retention is improving but not proven long term.
so i haven't "made it." not close. but a year and a half ago i couldn't build anything, and now strangers in 40 countries use two things i made. that shift feels worth sharing.
if you're building something and everyone around you thinks you're crazy, keep going. the tools genuinely changed what one person can do.
happy to answer anything, the stack, the ChatGPT approval, the mistakes, how i use AI as my whole team.
Honestly, the fact that the same engine converts on your web dashboard but flops inside ChatGPT is super telling. It proves people are willing to pay, they just hate the current friction. Before you buy into building more paid features, I’d map out the exact user flow:
an MCP user saves something → opens the dashboard → comes back within 7 days.
Where exactly is the drop-off happening right now?
you nailed it, that's exactly the reframe i landed on. same engine, same value, people just hate the current friction of paying "inside" a tool they feel they already pay for.
interesting thing i've found: retention is actually stronger on the MCP side than the dashboard. the MCP solved my stickiness problem, people come back and use it. so my concern flipped, it's not "will they stick," it's "how do i convert that sticky MCP usage into the dashboard where paying feels natural."
the bridge is what i'm building now. the idea: everything a user does in the MCP gets saved to their account and shows up on the dashboard too. so when they move over, they land on a dashboard already full of their own work. it turns "pay inside a chat" into "here's your library, unlock more."
your user-flow question is the right one though: MCP user saves something → opens dashboard → returns within 7 days. mapping that exact drop-off is my next step before building more paid features. thanks for pushing on that.
Proving skeptics wrong by pushing through the initial noise is the ultimate validation for any solo builder, and securing official approval in a major ecosystem like ChatGPT’s app directory provides massive built-in leverage. Transforming persistent effort into a multi-product portfolio demonstrates that early vertical focus combined with strategic ecosystem integration can quickly turn ignored side projects into scalable distribution channels.
thanks, appreciate the kind words 🙏
500% growth off $232 MRR and a real ChatGPT App Directory approval is the part nobody can wave off, someone at OpenAI had to say yes. i use viewfy for finding threads where the buying question is allready typed out, and Reddit search for everything else. w
appreciate that, the directory approval is the one thing i can't talk myself out of ha.
and that's a smart tactic, finding threads where the buying question is already typed out. i've mostly been doing SEO/AEO and directory listings, but proactively catching high-intent threads is something i haven't done systematically. i'll look into viewfy, though i'm guessing reddit's own search gets me part of the way too. how are you making sure you're helping in those threads vs coming off as promo? that's the part i'd want to get right.
pasting code between chatbots and fixing bugs by hand is a real apprenticeship, even though it looks like flailing from the outside. you end up knowing where things break, which is knowledge people who started with good tooling often skip. the directory approval is the part that is hard to argue with, since somebody else had to agree.
this is a really good reframe, thank you. i never thought of it as an apprenticeship but you're right, all that pasting and manual bug-fixing meant i actually learned where things break and why. when Claude Code writes something now, i can usually tell when it's about to go wrong because i've hit that wall by hand a hundred times. people who started with clean tooling maybe skip that pain, but they also skip the intuition.
and yeah, the directory approval is the one thing i can't talk myself out of at 3am, ha. everything else i can rationalize as luck or hype, but someone at OpenAI actually had to look at it and say yes. that one keeps me going on the hard days.
appreciate you taking the time to write this, genuinely.
The adoption numbers are strong, but the monetization friction around MCP is the more interesting signal.
Curious whether the conversion problem is mainly that users don't see enough incremental value to pay for HookLayer, or that they don't yet perceive an MCP tool inside ChatGPT as something they should pay for separately.
you framed that better than i've been able to. it's genuinely both, but i'm increasingly convinced it's more the second one.
on the value side, the users who DO pay are the heavy ones who already saw clear incremental value, agencies and power users who burn through free credits fast and top up to keep going. so for that segment, value is proven.
but the broader pattern is the perception thing you nailed. people happily use it 100 times inside ChatGPT for free, then freeze at the paywall, not because it's too expensive, but because paying for a tool "inside" ChatGPT feels like paying twice for something they already subscribe to. the mental model is "this is a ChatGPT feature," not "this is a separate product with its own value."
what's interesting is the same person converts easily on my web dashboard, where they log in and "own" the product. same engine, same value, totally different willingness to pay. so the interface itself is shaping the perception.
my working theory is the fix isn't lower prices, it's making the value feel like it's mine not ChatGPT's, plus giving heavy users a reason to formalize (subscriptions, saved state, tracking over time) so paying feels like upgrading a relationship, not renting a chat feature.
still testing it though. how have you seen this play out, have you looked at MCP monetization or is this from a different angle?
You found the real insight and then buried it: people will not pay inside a surface they already pay for. That points to running HookLayer as free top of funnel and charging on your own account layer where saved work, history, and team access live, so the MCP acquires and the web app monetizes. At 1,500 users and $232 MRR the constraint is not traffic, it is that the paid thing is sitting in the wrong place.
this is exactly right, and you said it more clearly than i had it in my own head.
that's the direction i've landed on too: HookLayer as free top of funnel that acquires at basically zero cost, and the account layer, saved work, history, tracking, team access, as where the paying happens. the MCP's job is acquisition, the web app's job is monetization, and right now they're disconnected so the funnel leaks exactly where you'd expect.
the thing i'm building toward is a bridge, since both products already share one account system, so a HookLayer user can move into the web app with their existing account and saved state, no re-auth, no starting over. that turns "pay inside the chat" (which people resist) into "keep your work and unlock more on your dashboard" (which they don't).
you nailed the core of it though, the constraint was never traffic. it's that the paid thing is sitting on the wrong surface. that reframe is genuinely useful, thank you.
are you building in the MCP/agent space too, or seeing this pattern from the outside? would be curious how you got to this so cleanly.