Questgen & Supermeme Today: 2.5 Million Signups, $6,500–7,000 a Month — Three AI Tools Down From a $100K Peak Year

Registered users across the portfolio exceed 2.5 million. Monthly revenue is $6,500–7,000. That gap is the substance of this record.

Ramsri Goutham Golla of India runs three AI products while keeping his job as a data scientist. We read the interview he gave Indie Hackers.

The three products

The three together currently gross $6,500–7,000 a month. Questgen charges $10–15 at entry and $50 at the top tier, with 40% off annual plans. Gross margin after LLM costs is 60–70%, per Golla.

How 2.5 million people become only $7,000

Put 2.5 million signups against $7,000 a month and it works out to roughly $0.03 per user per year. In generative-AI tools with generous free tiers, the overwhelming majority of signups never approach payment.

The same shape exists in our archive: the Pomodoro timer with 1M monthly users, 1,000 paying, and 80% of revenue from ads published a 0.1% conversion rate. User counts look like a business metric but move on a different axis from money.

Golla’s three charge by subscription, not ads, so free users sit purely on the cost side — every LLM call burns inference fees. The 60–70% gross margin is what remains after that.

And it’s down from the peak

Do not miss that $6,500–7,000 is not a number on the way up. By his account, the 2025 peak was $100K a year, roughly $8,300 a month. The current run-rate of $78K–84K is about 20% off the peak.

Generative-AI tools lose their reason to exist when the base models absorb the same capability. Golla acknowledges competing with ChatGPT and Anthropic, and names his defense: “specific workflow formats you can’t get elsewhere”, structured quiz output you can use directly, rather than prompting a general chatbot.

Plateaus rarely get published. Records like the Airtable extension that printed its five months of completely flat MRR are a minority even here.

The channels that worked

Four channels, all unpaid: publishing open source on GitHub and routing the README to the paid web app. SEO articles about question generation. Daily posting on X and LinkedIn. Product Hunt (Supermeme took #1 of the day).

His advice is about ordering: “Ship a light demo, make mini-content on X or YouTube, measure how the idea lands. Lean into what resonates.” Look at reactions before building deep.

The same “run several, find the hit” pattern shows in Marc Lou’s portfolio at $1,032,000 a year. The differences are scale and that Golla keeps his day job while running all three.

Free at the door, paid behind it

Structurally, the most interesting channel is the open-source funnel: publish the repo on GitHub, route README readers to the paid web app. Hold the spot where developers first land when they search the problem, for free, then hand them to billing when they actually use it.

Pricing runs $10–15 entry, $50 top, 40% off annual. The stack is NextJS and Supabase, deliberately within what one person can maintain.

This “build an audience at a free entrance, charge behind it” order also echoes Aceternity UI, which started with 7 free components and reached $80,000 a month on the paid tier. The difference: there, the paid launch made it jump. Here, only the free side kept scaling.

Position among our cases

¥975,000 a month sits below the median of revenue-disclosing cases on this site (~¥1.5M). Restrict to side projects, though, and the median drops to about ¥150K, within the side-project group this is upper tier. The real reading material is the mismatch between 2.5M signups and this amount.

$150K across four years

Divide Questgen’s $150K lifetime by four years: $37,500 a year. That is the four-year total for a product with 150K signups.

Generative-AI stories usually quote launch velocity. Few records show what remains at the four-year mark. The lesson available here is exactly one: speed of takeoff and thickness of accumulation are different things.

Designing around a day job

Whether $7,000 a month sustains a life depends on who you ask. But Golla keeps his data-science job, so these three don’t need to pay the bills, which also means decline is not a reason to fold.

The reverse reading: it has not reached the level that justifies going full-time. Where he would switch judgment, he doesn’t say.

What can be read is the shape of generative-AI tools: fast to rise, hard to hold. A Product Hunt #1 and two million signups have not carried monthly revenue past the $10,000 wall. Not published: churn, paying-user counts, per-product revenue, and actual LLM costs. We know the total, which of the three carries it, we don’t.