Case snapshot

Who: idoubi (艾逗笔), indie maker; left Tencent around October 2023
Products: ThinkAny (AI search), ShipAny (AI SaaS boilerplate), and more
Public figures (his write-ups): ThinkAny launched ~2024-03-20, Product Hunt daily #4; ~170k users in ~3 months, peak monthly traffic in the hundreds of thousands; paid conversion ~0.03%, API costs hard to cover until cost cuts around Aug–Sep brought break-even. ShipAny Christmas 2024 preorder, top tier ~$299 at 50% off, ~$10k in ~4 hours; ~$1k MRR across products that December, with ShipAny’s first week beating half a year of the others.

Many AI builders collapse two metrics into one:

People use it ≈ the product succeeded.

ThinkAny shows those are not the same thing.

In his long BAAI Community post, traffic clearly worked; the cash breakthrough came later from ShipAny, a developer boilerplate. The question worth studying is not “why templates sell,” but:

Which kind of user value can become a durable business?

1. Traffic first—then unit economics

ThinkAny’s v1 shipped in a weekend around March 2024. The idea was clear: RAG over the web, then an LLM answer with citations; multilingual from day one. Users often loved the mind-map summary of results more than “yet another chat box.” Later versions added follow-ups and multi-model options (GPT-4, Claude, and so on)—every search still burned retrieval and token APIs.

Soon after launch it hit Product Hunt daily #4; creators on Twitter, YouTube, and TikTok amplified it across Egypt, Japan, India, and beyond.

About three months in: ~170k users, ~3k daily UV, ~6000 searches/day; SimilarWeb-style monthly visits around 580k. As acquisition, that is a strong proof.

May’s big release added paid upgrades. Then the ugly number: ~0.03% pay rate. Search and model APIs still dominated spend—smarter features often mean more real cost per use.

The growth loop looks clean:

Need to search → free use → more users → more searches

Unlike many classic web tools, each query can cost retrieval, crawl, tokens, and bandwidth:

More users → more queries → higher variable cost.

Plenty of internet products assume more users lower marginal cost. Early AI search often does the opposite. For a free user:

Contribution ≈ revenue − (APIs + model + infra)

Near-zero revenue and positive cost means free users are negative-margin. The business bets that enough of them convert. At ~0.03%, that bet fails—more traffic can mean more loss.

So ThinkAny actually answered two different questions:

Question	Result then
Is there demand?	Yes
Can you acquire users?	Yes
Will they pay enough to cover cost?	Weak

Demand proves usage. It does not prove the need is valuable enough for you.

2. When revenue won’t move, survive growth first

After acquisition worked, the break was acquire → pay.

Fundraising was hard. He later wrote that capital was cautious; scaling a general AI search on VC alone was unrealistic. Investors pushed vertical; he still chased global, general search—inside big-tech range on quality, latency, and retrieval. A solo builder rarely wins experience, scale, and margin at once.

When revenue could not jump, he cut cost: freemium limits, cheaper models/calls, less waste. Around Aug–Sep, spend came under control and the product reached break-even—with churn and softer traffic as trade-offs. Cost control often trims experience; growth wobbles.

Not a glamorous story, but a solo-realistic one:

If you cannot prove revenue yet, first prove growth will not kill you.

Break-even is not a high-margin business—it is bleeding → living. His year-end note: many products and big traffic should have meant money; monetization was weak, and ThinkAny ran negative for a stretch.

3. What changed was who pays

ShipAny did not “ThinkAny, but bigger.” It changed the buyer.

ThinkAny: people who want an answer.
ShipAny: builders shipping an AI SaaS who want to launch faster.

The gap is economic value of the pain, not age or geography.

A casual “look this up” is low value per hit. A developer still needs auth, payments, i18n, DB, deploy, account area—hours or days each.

ShipAny packages that repeated plumbing: ship an AI SaaS site quickly on the stack he had reused for a year (Next.js, auth, DB, i18n, Stripe-class payments, cloud deploy). Buyers get a base to build on, not a chat demo. He even finished the code in about a week after preorders—clear offer first, then delivery.

He was not really selling files. He was selling:

A compressor for development time, built from pits he had already fallen into.

Around Christmas 2024: landing page, pricing, payments live, preorder posts, 50% off before 2025. Public write-up: dozens bought the top tier (~$299 half off) and ~$10k landed in ~4 hours; one week beat half a year of other products.

4. Same maker, two businesses

	ThinkAny	ShipAny
Buyer	General web users	Developers
Job	Get information	Save build time
Usage	High frequency, mostly free	Low frequency, high value
Cost	Scales with each use	Mostly build & maintain
Why pay	“Slightly better search”	“Skip a pile of plumbing”
Price	Low, hard to push	Hundreds of dollars possible
Marginal cost	Relatively high	Relatively low

ThinkAny sells a search. ShipAny sells fewer wrong turns on a project. Value density differs.

If a boilerplate saves ~10 hours, $299 can still feel fair. If someone only wants a few queries, a tens-of-dollars subscription needs far stronger ongoing value. So:

High ticket is often about a more valuable problem—not a more complex product.

The four-hour spike is easy to misread as “AI templates just sell.” Better to unpack the deal stack:

1. Clear buyer — AI SaaS builders, not “everyone who likes AI.”
2. Priced pain — auth, pay, i18n, deploy skipped; buyers can estimate hours saved.
3. Clear delivery — a codebase to extend, not vague “AI power.”
4. Seller as proof — a year of shipping AI products; selling a process he had used, not a stranger’s zip.

Without trust and a crisp deliverable, a discount alone rarely explains that night’s order density.

5. Personal brand: weak for search, inventory for boilerplate

ThinkAny users leave after a query; they rarely care who built it. Brand barely moves pay rate.

ShipAny buyers ask: who are you? Have you shipped AI SaaS? Did you run this stack yourself? PH ranking, public postmortems, community, and a string of launches become trust.

Same influence, very different monetization:

Influence is not inherently valuable. It becomes an asset when it overlaps the people who pay.

Exposure earned in the traffic phase converts poorly on the wrong SKU—and can close a preorder on the right one.

6. Shipping a dozen products ≠ running a dozen business tests

In 2024 he shipped fast: viral tools, search, music, try-on, landing pages, podcasts, boilerplate… For AI windows, ship → learn → double down or drop is rational. A six-month build can miss the market.

It is also easy to misread. Fast launch ≠ fast commercial validation. If most products lack a clear buyer, ongoing promo, a pay path, and iteration, the ledger says “shipped ten,” not “tested ten businesses.”

Pagen is telling: strong SEO traffic, a ~$10k acquisition offer, MRR asked—he said 0 (little promo, little pay). The buyer still wanted it; he declined. Traffic can be priced by others; without a habit of paying customers, your own P&L still has no business.

He later admitted spreading too thin. Speed helps and distracts. The clean conflict that remains is ThinkAny (reach without profit) vs ShipAny (ticket + preorder).

You cannot copy ThinkAny’s PH moment or search-window luck; another AI SaaS template may not clear $10k in four hours either—timing, brand, and chance sit inside the result.

What travels better is the order of questions: what pain is worth paying for, then whether each new user adds revenue or cost. If “more users = more loss,” growth is not priority one. Versus ThinkAny, ShipAny’s lesson is less “sell boilerplates” than—

Price the work you have already proven and others refuse to redo—and sell it to people who already pay to save time.

A hundred high-intent buyers can beat ten thousand non-payers. Acquisition skill and monetization skill are different skills.

Product

• ThinkAny
• ShipAny