When to read this: You are starting a company, looking for product-market fit, or comparing specific growth cases. This synthesis organizes Shensi's collection of more than 102 cases into five stages: deciding whether to start, finding product-market fit, initial growth, growth from one to ten, and scaling from ten to one hundred.
Core reference points: a16z's three paths to product-market fit—working with leading customers, getting ahead of an exponential curve, and being your own first customer—alongside Sequoia's four “Terrifying Questions,” Lovable's reported six-part growth playbook, Perplexity's reported five growth engines and $450 million in annual recurring revenue after three and a half years, and a solo AI-video business reportedly reaching $100,000 in monthly revenue in nine months.
This is a complete English edition of the previously published synthesis. Company metrics, valuations, growth multiples and case headlines retain their historical status as claims in that collection; translating them does not independently verify them or make them current figures. Where the published note supplies only a headline or refers readers to a longer source, this edition preserves that limitation rather than inventing the missing explanation. ARR means annual recurring revenue. Product-market fit, abbreviated PMF, means a product meeting a market's demand; go-to-market, or GTM, concerns how the product reaches and wins customers.
0. Five Parts of the Startup Journey
| Stage | Key question | Representative approach |
|---|---|---|
| Deciding whether to start | Supply-driven markets and founder-market fit | Sequoia's Terrifying Questions |
| Finding product-market fit | The light-bulb moment; stronger response from cold prospects than personal connections | a16z's three paths to PMF |
| Initial growth, from zero to one | The first 10, 100 and 1,000 users | Lovable's six-part playbook and building in public |
| Growth from one to ten | From 5,000 to 50,000 users; ARR begins to accelerate | Perplexity's five growth engines |
| Scaling from ten to one hundred | From 50 million to one billion in ARR | The Snowflake first-sales-hire case, described as growth from zero to $4 billion |
The stages and numbers reproduce the historical framework. The table combines examples with different measures and scales; it is not a single benchmark that every company must follow.
1. Deciding Whether to Start: Founder-Market Fit
1.1 Sequoia's Four Terrifying Questions
The companion note on Shensi's venture-capital landscape, section 2.1, contains the broader discussion.
The question highlighted here is: What gives my company the right to exist? The note connects this to founder-market fit. Its example is Nubank founder David Vélez spending months studying Brazil's banking industry.
The source overview names four questions but does not spell out all four in this section. Questions two and three appear below; this edition does not invent the missing fourth question.
1.2 a16z and Supply-Driven Markets
The companion venture-capital landscape note, section 1.1, discusses the investment logic around Substack: enabling creators to earn money can bring into existence creators and content that do not yet exist.
This is the supply-driven-market argument recorded in the collection, rather than a claim that every new monetization tool creates a viable market.
1.3 Contrarian Advice in the Shensi Cases
One archived headline describes a founder who raised $100 million in a year and a half but offered advice that ran against conventional wisdom.
The published synthesis points to the longer source and characterizes the likely themes as questioning standard venture-capital advice, speed obsession and aggressive hiring. It does not reproduce the actual advice in detail; those possible themes should not be mistaken for a verified account of the interview.
2. Finding Product-Market Fit
2.1 a16z's Three Paths, Attributed to Jason Rosenthal in May 2026
The companion venture-capital landscape note, section 1.3, provides the related discussion. The three paths recorded here are:
- Build with leading customers. Their needs become the product specification.
- Get ahead of an exponential curve. Make a bet on emerging infrastructure.
- Be your own first customer. Use the product yourself—often called dogfooding.
2.2 Questions Two and Three in Sequoia's Framework
| Question | Positive signals recorded in the note |
|---|---|
| Do people care enough? | A stronger response from cold prospects than relationship-based prospects; visible excitement; active participation in demos; finding design partners |
| Can my product actually change behavior? | A light-bulb moment, a short path to getting started, and features that encourage continued use |
The DoorDash example describes conversations with hundreds of small restaurants, leading the founders toward local restaurants that were turning away orders yet could not afford delivery.
These are case signals for further investigation, not thresholds that establish PMF on their own.
2.3 A PMF Warning Attributed to Jason Rosenthal
The source argues that relying on growth hacks and ongoing airdrops to sustain activity, without a clear strategic purpose, can amount to admitting that product-market fit has not yet been found.
The distinction is between activity sustained by the underlying product and activity repeatedly supported by incentives. It is the source's diagnostic argument, not a universal prohibition on incentives.
3. Growth From Zero to One
3.1 Lovable's Reported Six-Part Playbook
The note on Shensi's AI coding companies, section 2, contains the fuller account. The historical summary reports 50 million in ARR in six months, one million users in ten months, and a claim that traditional growth strategies had stopped working in this setting.
Although the heading calls this a six-part playbook, the published summary enumerates only the following elements:
- Avoid traditional paid advertising.
- Do not bid on SEO keywords, using the source's terminology.
- Emphasize user-generated content, showcases and community.
- Consult the longer source for the remaining details.
The edition retains that incomplete enumeration. It does not infer three additional tactics or treat this company's account as a rule for every AI startup. The wording about SEO and bidding also remains the source summary's wording rather than a newly standardized marketing taxonomy.
3.2 Building in Public
An archived growth article presents #buildinpublic as a method for products expanding internationally. The summary identifies three elements:
- Share the building process publicly.
- Make progress regularly and show the data transparently.
- Build trust that can lead users to share the product themselves.
3.3 A Solo AI-Video Business Reportedly Reaching $100,000 a Month in Nine Months
The collection includes a case headlined as one person building an AI-video business from zero to $100,000 in monthly revenue in nine months. The published note labels it a typical independent-builder path but provides no detailed operating sequence in this section.
The amount and time frame are case-headline claims, not a forecast for another builder.
3.4 A Reading App Reportedly Reaching Ten Million Users in Three Years
Another case headline asks how a reading app grew from zero to ten million users over three years and promises a breakdown of its growth loops. This section of the published synthesis contains the headline only, rather than the loops themselves.
3.5 Canva's Early SEO Tactics
The Canva case is presented as an account of moving SEO from zero to one, associated with Jieyue Insights. Its summarized tactics are:
- Programmatic SEO.
- Frequently searched long-tail keywords.
- Automated production of template pages at scale.
3.6 Webflow's No-Code Growth
The collection includes a long-form analysis of Webflow's growth as a leading no-code product. The published overview supplies the case title without reproducing its detailed explanation.
4. Growth Engines From One to Ten
4.1 Perplexity's Reported Five Engines
The historical headline describes Perplexity receiving a dismissive reaction from an early investor, then reaching a reported $450 million in ARR three and a half years later. It promises five growth engines.
The published synthesis directs readers to the full Shensi article for those five engines. They are not listed here, so this edition does not manufacture them.
4.2 Reported Fifteenfold Growth in One Year for a European AI Customer-Service Company
Another case headline describes a European business standing out in a crowded AI voice customer-service market, growing fifteenfold in a year and receiving a $20 million investment from a leading Silicon Valley venture firm. The published summary does not identify the company or give the underlying measurements in this section.
4.3 Reported Tenfold Growth in Six Months for a Software 3.0 Pioneer
The collection also includes a case promising an explanation of how a company grew tenfold in six months and became a pioneer of Software 3.0. This section preserves the headline-level reference without adding a company identity or mechanism absent from the public note.
5. Scaling From Ten to One Hundred
5.1 Snowflake's First Sales Hire: The Reported Path From Zero to $4 Billion
An archived headline presents the growth methodology of Snowflake's first salesperson as a journey from zero to $4 billion. The overview does not specify the precise metric behind that headline here; it should not automatically be read as a directly comparable ARR series.
5.2 The One-to-One-Hundred Acquisition Model
A case headline describes a company that bypasses starting products from scratch and instead acquires struggling products, reportedly generating one billion in annual revenue. The published headline does not specify the currency in this instance, and this edition does not add one.
The original takeaway is that starting from zero to one is not the only entrepreneurial path: improving and scaling existing products can also be economically significant. The headline's revenue claim does not establish how repeatable or attractive a particular acquisition would be.
5.3 From Zero to $800 Million: A Young Salesperson's Customer-Acquisition Case
Another archived headline describes the customer-acquisition methods of a 23-year-old salesperson associated with growth from zero to $800 million. The overview provides no further operating details or metric definition here.
6. A Go-to-Market Guide
6.1 GTM Across Zero-to-One, One-to-Ten and Ten-to-One-Hundred
A referenced guide is described as a systematic account of moving from zero to one and then to ten, with practical cases. The published synthesis summarizes typical GTM transitions as follows:
- Zero to one: founders sell directly and work toward roughly fifty seed customers.
- One to ten: make the first sales hire and standardize the sales process.
- Ten to one hundred: expand the sales team, work with channel partners and serve larger accounts.
These are stages suggested by the summary, not required customer counts or a hiring schedule for every company.
6.2 Growth-Leader and YC Alumni Accounts
The collection references an issue of Magineer's product selections featuring a growth leader from a product listed in a16z's Top 50 AI products, alongside YC alumni discussing entrepreneurship and product building. The overview records it as a source of methods, without spelling out the full playbook in this section.
7. Case Library by Milestone
The following tables preserve the historical examples and figures in the public source. Some case names are descriptive labels rather than identified companies. They use different measures and should not be treated as a ranked or normalized dataset.
7.1 Monthly-Revenue-Level Examples
| Case | Figure reported in the original table |
|---|---|
| A financial-analysis wrapper app, described as replicating Cal AI | $40,000 a month |
| A solo AI-video business after nine months | $100,000 a month |
| Pretext, described as a mother of two's project | 21 million people watching or following the story; this is an attention figure, not monthly revenue |
The original places Pretext under a monthly-revenue heading, but its cell records attention rather than income. This edition retains the example while making that mismatch explicit.
7.2 ARR-Level Examples
| Case | Reported ARR and time frame |
|---|---|
| Readdy.ai | 5 million in four months |
| Lovable | 50 million in six months |
| Perplexity | 450 million in three and a half years |
The original table omits currency symbols in these cells. Its preceding Perplexity case explicitly uses dollars. No missing currency is silently supplied for the other rows.
7.3 Valuation-Level Examples
| Case label in the collection | Reported valuation or accompanying detail |
|---|---|
| A founder who became a unicorn after being rejected by YC twice | One billion; founder aged 25 |
| An AI legal unicorn associated with a former esports player | 700 million; age 25 |
| An AI customer-service company associated with a Chinese founder | 1.5 billion |
| Cursor | 50 billion |
These cells reproduce the historical magnitude without assigning an omitted currency or independently verifying the valuations. The labels in the first three rows do not provide enough detail to identify the companies reliably.
8. Four Trends in Startup Methods
The source collection includes an article described as a guide to four new trends in entrepreneurship in the AI era. The public summary presents the following themes:
- AI may make a one-person unicorn possible.
- Aesthetic judgment can matter more than the model, described as a view shared on Xiaohongshu.
- Broader access to building, illustrated by Pretext.
- Long-term commitment, associated with Dan Koe.
These are the summary's themes and aspirations, not a verified prediction of outcomes or a complete reproduction of the original article.
9. An Extension of Founder Mode
A case associated with AppSumo's former CEO argues that Founder Mode is no longer sufficient: founders should move from building the engine toward designing the engine in the AI era.
The summary contrasts an earlier image of Founder Mode, associated with Brian Chesky and direct hands-on involvement, with a newer role coordinating AI agents, multi-agent orchestration and systematic collaboration.
The related notes on newtype's independent-operator mindset and on time management and Life OS provide the companion discussions. This is the source's proposed shift in the founder's role, not a claim that founders no longer need direct customer or operational knowledge.
10. Connections Across Authors
| Topic | Perspective in this Shensi synthesis | Related public notes |
|---|---|---|
| Startup methods overall | Cases viewed through leading venture investors | General startup methodology; Yitang's complete five-step method; Wang Kai's startup methodology |
| PMF and the five-step method | Sequoia's questions and a16z's three paths | Yitang's five-step method, associated with Truman |
| Lean startup and MVPs | a16z's warning against using growth hacks to sustain activity without PMF | Lean startup and MVP; Yitang's research and lean experiments |
| Demand analysis | DoorDash and Nubank cases | Demand analysis and jobs to be done, or JTBD; Yitang's complete demand-analysis guide |
| Growth playbooks and GTM | Lovable, Perplexity and Snowflake | AI-enabled content and self-media engineering |
| Anticipating opportunities | a16z's supply-driven-market perspective | Opportunity forecasting; Yitang's opportunity-forecasting note |
The public links collected at the end of this edition preserve the original article's available cross-references.
11. How to Use the Framework by Stage
| Where you are | Main references | Starting question or activity |
|---|---|---|
| Deciding whether to start | Sequoia's four questions and a16z's supply-driven-market view | Founder-market fit; willingness to commit ten years |
| Finding PMF | a16z's three paths and Sequoia's questions two and three | Leading customers, exponential infrastructure curves, or using your own product |
| Initial growth | Lovable, building in public and programmatic SEO | Community, transparency and the case's preference against paid acquisition |
| ARR acceleration from one to ten | Perplexity's five-engine case and the reported fifteenfold-growth case | Combining growth engines |
| Scaling from ten to one hundred | Snowflake and the zero-to-$800-million sales case | Standardizing sales and working with large accounts |
| Improving existing products rather than starting from scratch | The struggling-product acquisition case | An acquisition-led path |
The original proposes six general disciplines:
- Begin with the supply-driven opportunity and founder-market fit.
- Look for real users continuing to use the product, rather than substituting total value locked, follower counts or money raised for PMF.
- Do not rely on growth hacks simply to keep activity alive; the source treats that as a warning that PMF may be missing.
- Re-examine traditional growth tactics. The Lovable case favors community over paid acquisition, but its claim that traditional strategies have failed is not evidence that no AI company should advertise.
- Consider the founder's role in designing a working system, not just performing every task personally.
- Recognize that both building from scratch and improving existing products can create value; there is no obligation to choose zero-to-one entrepreneurship.
These are the synthesis's organizing lessons. They remain starting points for examining a specific product, rather than guaranteed growth recipes.
12. References in the Historical Synthesis
The published note points to Shensi's author overview as the collection entry, and to a source collection of roughly 102 or more items tagged with terms such as PMF, entrepreneurship, growth, zero to one and GTM. That archive count describes the collection behind the note, not the number of fully reproduced cases on this page.
Its main companion notes cover:
- Shensi's venture-capital landscape: a16z, Sequoia and YC perspectives.
- Shensi's AI coding companies: more detail on Lovable and Readdy.ai.
- General startup methodology.
- Yitang's complete five-step method, associated with Truman.
- Wang Kai's startup methodology and his argument against treating fundraising as success.
- Lean startup and the minimum viable product.
- Demand analysis and jobs to be done.
Public reference links are retained in the final section. Local archive paths and private note identifiers are not needed for the English reader and are not introduced into this edition.
2026-10-17 Addition: Move from a demonstrable feature to a product you can deliver consistently
The source is Shensiquan’s Chinese account of Alfred Lin’s Stanford talk. Methods and viewpoints retain their attribution. Historical revenue, valuation and leverage figures have not been independently verified and are not used to argue for a probability of success.
Turn enthusiasm into an executable roadmap
A feature can demonstrate a possibility without showing that users will keep using it. Alfred Lin’s emphasis on combining enthusiasm with disciplined execution can be applied across three stages: a feature proves that something can be done; a product roadmap identifies whom to serve first and which essential gaps to address next; consistent delivery requires the team to fulfill those commitments reliably. A positive reaction to a demo cannot serve as acceptance evidence for the latter two stages.
| Stage | Evidence to retain | Common misjudgment |
|---|---|---|
| Feature | Users can complete a specific action, and its failure conditions can be explained | Assuming an impressive demo solves the whole problem |
| Product roadmap | Target users, recurring needs and the current trade-offs are explicit | Collecting many ideas without deciding what to solve first |
| Consistent delivery | Outcomes have owners, commitments can be tracked and feedback changes the next round of work | Filling gaps through personal enthusiasm and mistaking occasional success for reliable capability |
The table is this knowledge base’s proposed application. It supplements existing product-market fit and customer-interview methods without replacing evidence of payment, retention or actual switching behavior. A small feature can be valuable. What matters is whether it reliably solves a problem users care about, rather than how many features it includes.
Connect conviction to observation
Strong judgments should be grounded in customers, field observations and repeated practice. At each review, record what you initially believed, what supporting or contradicting evidence you observed and which commitment changed as a result. Celebrity opinions and team enthusiasm alone remain insufficient evidence. Changing the entire direction after every piece of feedback also makes cumulative learning difficult.
When to use it. When a demo is well received, the team is considering a larger commitment or the roadmap keeps shifting, use the table to identify which stage has actually been completed. Retain a long-term direction while adjusting near-term plans to specific observations; this is not a formula that guarantees growth. Read alongside the existing discussion of pivots and feedback in Startups and investing, without duplicating its cases.
October 19, 2026 addition: define the audience, then connect the promise to feedback
Why does producing more content not necessarily produce growth? Shensi's Chinese account of a Distribution podcast interview with YouTube strategist Oscar emphasizes answering “who is the audience?” before choosing how to produce content. The workflow lessons connect audience definition to a testable content promise and feedback. Successful cases alone do not establish a universal causal relationship.
- Start with the audience's task. Turn the ideal customer profile, or ICP, into concrete questions: what problem does this person face, why would they click, and what could they do after watching? Large numbers of irrelevant views do not directly establish product use. Awareness content and customer tutorials need not perform the same conversion task.
- Write the promise before filming. Establish the value communicated by the idea, title, and thumbnail together, then plan the script and material. This lets the team check for a clear audience before expensive production. If a title creates curiosity, the content must deliver; stronger packaging cannot supply information that is absent.
- Examine both clicks and delivery after the click. When clicks are weak, examine topic selection and packaging first. When viewers leave quickly after clicking, examine whether the opening and content match the promise. These are diagnostic hypotheses to test, not causes established by a single number. Audience composition and distribution conditions can also change.
- Specify one factor to change in each round. Record the original version, change, observation window, and audience. Try changing either the title or the thumbnail before comparing feedback. The interviewee's click-through rates, title lengths, and test durations are contextual experience, not performance standards for every channel. A gain in clicks should also be checked for losses in viewing quality or meaningful product tasks.
Whether to separate a brand-awareness channel from a how-to channel depends on whether the two audiences have substantially different tasks. The account describes entertainment or industry stories attracting potential users while tutorials serve existing users. If the same subscriber base consistently responds differently to the two formats, separation may be worth considering after assessing the team's capacity to maintain both. This does not require every small team to operate two channels immediately.
Claims about a “seven-hour trust rule,” rankings of sources cited by LLMs, the relationship between views and citations, and fixed returns from particular budgets were not independently verified here and are not presented as general rules. The actionable elements are defining an audience, making the promise testable, and using feedback to guide the next iteration.
Source: Shensi's Chinese account of the Oscar interview.
October 23, 2026 Addition: Use Historical Scores for Shortlisting, Retain Direct Evaluation
A small team can use an agent to narrow candidates using past human judgments, while retaining direct evaluation. In Shensi’s account of Garry Tan’s conversation with Paul Copplestone, Tan feeds an earlier hackathon’s manual scores and rankings to an agent to produce reusable code and Markdown. He still tries the products and talks to five to ten shortlisted builders. This is a workflow example, not evidence that automated scoring is generally accurate or a hiring-screening rule.
Application in this knowledge base: retain the rationale behind the original judgments, the automatic ranking and changes after human review. Check whether older criteria exclude promising new kinds of candidates. Evaluate browsing time saved separately from the suitability of the eventual choice. The source also raises the risk that published metrics invite people to optimize for them, the Goodhart problem: when a measure becomes a target, it can stop representing the underlying value. Stable rankings therefore still need direct use, feedback and renewed criteria when the task changes.
This article already covers startup fundamentals and real user feedback. The addition is limited to the sequence of building a first-pass filter from past judgments, retaining human experience and checking for metric gaming. Unverified productivity multiples, company revenues, memory-retrieval accuracy and scientific forecasts are excluded from the conclusion.
Source: https://www.woshipm.com/ai/6483135.html