Alp SuleimanGrowth Partner

AI growth cases

Taking apart the loop behind fast-growing companies

Each case opens a company's public growth story through a growth partner's eyes: which channel, which loop, which price, whether retention is proven. Then it asks the same question every time: is this speed real growth, or visible growth?

Primary sources first, secondary sources flagged · every figure is numbered · opinion sentences are marked (opinion)

Growth Cases

40 cases

  • Vertical B2B

    11x

    The AI sales-rep company reached a ten-million-dollar ARR claim on the back of two top-tier investor signatures; an investigation showed that ARR rested on contracts with exit clauses, unauthorized logos and high early churn.

    Perception levelCautionary case
    • Cautionary fall
    • Enterprise sales
    • +1
  • Vertical B2B

    Apilex

    A legal assistant that entered the lawyer market from below with Turkish case-law data and, with no outside funding, claimed eight million dollars of annual revenue in nine months; the number is not necessarily wrong, but nothing about it can be verified from outside.

    Perception level
    • Vertical depth
    • Self-serve PLG
    • +1
  • Code and agents

    Bolt.new (StackBlitz)

    Seven years of in-browser infrastructure met the right model threshold and exploded into a tens-of-millions revenue claim in five months; after that the revenue curve went dark.

    Gaps remain
    • Launch cadence
    • Viral output
    • +1
  • Code and agents

    Builder.ai

    A no-code app platform valued above a billion dollars on an AI label went from default to insolvency in a month, amid allegations that most of its revenue was unpaid invoices and customers with one-dollar deposits.

    Perception levelCautionary case
    • Cautionary fall
    • Enterprise sales
    • +1
  • Consumer

    Character.ai

    A chat app that reached a billion-dollar valuation on more than an hour of daily use never disclosed revenue; the founders left for Google, the CEO for Disney, and under-18 chat was shut down.

    Perception levelCautionary case
    • Community
    • Cautionary fall
    • +1
  • Vertical B2B

    Clay

    A company that built product for six years on almost no revenue focused on one use case, turned its support channel into a community, and reached one hundred million dollars in ARR in two years; it disclosed its enterprise retention and its own pricing mistake.

    Close to a real reading
    • Community
    • Self-serve PLG
    • +1
  • Consumer

    Cluely

    Millions of views from controversy, a founder-claimed revenue figure, and the retraction of that figure nine months later; the starkest form of the gap between visible and real growth.

    Perception levelCautionary case
    • Cautionary fall
    • Creator program
    • +1
  • A studio from Istanbul claiming more than four hundred million dollars in annual revenue with no outside investment, over sixty mobile apps and heavy paid advertising; the largest scale claim from Turkey, with the least independent evidence.

    Gaps remain
    • Aggressive pricing
    • Launch cadence
    • +1
  • The company selling Devin, an autonomous software engineer agent, doubled its annual revenue over one weekend by acquiring Windsurf; the growth is real, but half of it is inorganic and the product scored poorly in an independent test.

    Gaps remain
    • Enterprise sales
    • Agent distribution
    • +1
  • Code and agents

    Cursor (Anysphere)

    A code editor that spread among developers with near-zero switching cost; it went from $500M to $1B in annualized revenue in five months, paid for selling variable model costs under a fixed subscription with a pricing crisis the same year; the SpaceX sale rests on secondary sources only (secondary data).

    Gaps remain
    • Self-serve PLG
    • Developer channel
    • +1
  • Vertical B2B

    Decagon

    A customer-support agent that grew through a chain of references rather than virality; the logo and funding evidence is strong, the revenue and resolution-quality evidence is only secondary.

    Gaps remain
    • Enterprise sales
    • Vertical depth
    • +1
  • Generative media

    ElevenLabs

    A company that never left audio grew a product family through viral output, a developer API and a single credit wallet, reaching a half-billion-dollar annual revenue claim in four years; margin and retention are still undisclosed.

    Gaps remain
    • Viral output
    • Developer channel
    • +1

Why this archive

To read the mechanism, not the story

When an AI company announces hundreds of millions in annualized revenue eighteen months after launch, there are two reactions: admiration or suspicion. Neither teaches anything. What teaches is opening the loop behind that number node by node and seeing which node is evidenced and which is a claim.

This archive does that. In every case we look first for the founder's own words, the company's own announcement or a funding release; where a primary source could not be opened, secondary sources such as news reports or data providers are used with a caveat. The number next to a figure leads to the source at the bottom of the page. Sources that could not be opened are marked as such.

There are fallers among the cases too. Nothing explains that a wave brings revenue but not a company better than a wave that has already broken. Cautionary cases therefore carry their own badge and are not hidden.

Method

Every case goes through the same ten sections

Profile, timeline, growth mechanism, real reading, shadow side, lessons, pattern tags, loop drawing, sources, open questions. The order is fixed, so two cases can be read side by side.

The source rule: first the founder's own words, the company blog, a funding release, a public filing or an interview transcript. Where those could not be opened, a news report or data provider is used and the figure carries a '(secondary data)' caveat in the text. A source the researcher could not open at all is marked separately.

The verdict has three levels. 'Close to a real reading': a revenue or usage outcome disclosed in a statement whose source was opened and whose definition is clear, with at least one stability signal. 'Gaps remain': the speed looks real but there is no retention, margin or cohort evidence. 'Perception level': what is measured is not an outcome; user counts, downloads, a podcast title.

Viral output
The product's output gets shared; every share is an ad.
Creator program
Creators are paid or given a cut; acquisition cost belongs in the same table as revenue.
Self-serve PLG
Users try, pay and spread the product without meeting a salesperson.
Developer channel
API, integrations and the developer community are the distribution channel.
Community
Discord, forums, the founder's account: the conversation around the product carries growth.
Aggressive pricing
Low entry price or a generous free tier at market entry.
Credit model
Sells usage credits; annualized revenue is not the same as recurring revenue.
Launch cadence
Frequent releases work as a channel; every release is a social event.
Model orchestration
No model of its own; packages other companies' models. The moat is a question mark.
Agent distribution
The product is distributed through agents or as an agent.
Enterprise sales
Long contracts, a sales team, pilots; PLG deliberately rejected.
Vertical depth
Embedded in one industry's workflow; the moat is context and data.
Cautionary fall
Fast rise, then a fall or a correction. The most instructive cases.
Pivot
A change of product, segment or channel started the growth.
Repricing
Pricing changed under margin pressure; the customer reaction is on record.

Frequently asked

Why is every figure caveated as a 'company claim'?

Because that is what it is. The annualized revenue of a private company is not an audited number; it is a founder's or an investor's statement. The archive does not hide this, it writes it next to the figure. The claim itself is data too: when it was said, under which definition and in which context all teach something.

Why are cautionary cases in the archive?

Because nothing shows that a wave brings revenue but not a company better than a wave that has broken. The same loop shape appears in rising and falling cases; the difference is made by a reason to return, margin and trust. Reading only the winners is survivorship bias.

How much of this applies to a product in Türkiye?

The mechanism applies, the scale does not. Viral output, creator programs and launch cadence work with low traffic too; but a dollar-priced model and a lira-paying customer build margin differently. Each case's 'what it teaches' section does that translation and says it is opinion.

Why doesn't a case page praise or condemn the company?

The aim is not to grade but to make the mechanism readable. The verdict is about the strength of the public evidence, not about the company. 'Perception level' does not mean the company is bad; it means an outsider has no outcome data in hand.