In the era of generative AI, building a startup that endures requires more than cutting-edge models; it needs defensible moats—sustainable competitive advantages that protect market position over time. Introducing the "4 Defensible Moats" framework we propose, we identify eight critical qualities that GenAI ventures must cultivate. Each pair of qualities aligns with one of the four moats—product flywheels and network effects, growth flywheels and social engagement, post-training barriers, and efficient monetization—to guide founders toward robust, long-term value creation.

Product Flywheels and Network Effect

Characteristic 1: Cultivating a Data-Driven Product Flywheel
A truly defensible product begins with building a relentless data-driven flywheel. At its core, this means instrumenting every feature, interaction, and decision point so that real usage feeds back into continuous improvement. Early on, you ship a minimum-viable product, collect signals on how customers engage, then leverage that data to train models or refine heuristics. As you improve recommendations, workflows, or automation, usage naturally climbs, which drives yet more data—and the cycle accelerates. Crucially, this loop doesn’t happen by magic. You need to invest in instrumentation from day one, design experiments to validate hypotheses, and ensure you can store, query, and process ever-growing volumes of information. What makes it defensible is that over time your dataset becomes unique and increasingly hard for competitors to replicate. If your flywheel truly powers a better user experience—with faster, more accurate answers or personalized content—users won’t switch to a new entrant that has to start its own data collection from zero. Moreover, as your model or algorithmic performance improves, you create “stickiness”: users become accustomed to a level of intelligence or customization they can’t find elsewhere. In practice, the product flywheel demands cross-functional coordination among engineering, analytics, and product teams to instrument effectively, iterate quickly, and keep the loop moving faster than any would-be imitator.

Characteristic 2: Building a Dual-Sided Network Platform
Complementing the product flywheel, a thriving GenAI startup must orchestrate a dual-sided network that brings creators and consumers into a shared ecosystem. This platform should facilitate content generation, such as automated article drafts, design assets, code snippets and others, fostering reciprocal value unlocks powerful network effects. Founders should architect attractive marketplace mechanics—revenue-share incentives, quality-based matchmaking, transparent reputation systems—that encourage professional creators to contribute their expertise while enabling end users to discover and consume high-value outputs. Early on, it can feel like a chicken-and-egg problem: you need content to attract users, and users to keep content coming. The solution lies in targeted seeding strategies—perhaps starting with a niche community of expert contributors—paired with lightweight incentives for early adopters to invite peers. Over time, the aggregate of creator contributions accelerates your model improvement as well (through richer training data) and diversifies use cases, turning the platform into a self-reinforcing engine of innovation, embedding a growing barrier to entry: new entrants face not only the technical challenge of model training but also the herculean task of building a vibrant, reciprocal network from scratch.

Growth Flywheels and Social Engagement

Characteristic 3: Engineering a Viral Growth Flywheel
Beyond product excellence, sustained expansion hinges on an engineered growth flywheel—a system of invites, referrals, and shared experiences that drives organic user acquisition. Effective viral loops blend frictionless onboarding with compelling incentives: time-limited premium trials for referrers, collaborative features that require inviting teammates, and built-in “share snapshots” that naturally broadcast usage to wider networks. Startups must obsess over the K-factor—ensuring that each active user generates more than one qualified invite that converts—while continuously optimizing invite copy, in-app reminders, and milestone celebrations. Equally important is integrating growth hooks directly into core workflows: think of an AI writing assistant that auto-formats invite-code watermarks on shared documents, or a design-generation tool that prompts users to tag collaborators. By aligning user delight with the mechanics of sharing, a founder implants growth at the DNA of the product. Over time, these viral loops compound, lowering CAC (Customer Acquisition Costs) and creating a moat around sustained expansion.

Characteristic 4: Fostering Social Engagement and Trust
Viral growth must be tempered with genuine social engagement and trust. Imagine runners sharing their today’s run route on social media via a map generated by their fitness tracking app: this not only projects their persona as fit and disciplined but also reinforces the fitness app’s brand identity. Platforms like Substack and Tencent IMA employ analogous strategies—paid knowledge sharing not only helps creators build an knowledgeable persona but also generates tangible income, further strengthening user recognition of the platform. This creates a win-win scenario that encourages sharing.These forms of social engagement are rooted in users’ personal identities and the trust they command within their communities. With user endorsements, a new app can penetrate communities far more easily. Ultimately, these social moats deter impersonators and reinforce a community-centered brand that newcomers find hard to replicate.

Post-Training Barriers

Characteristic 5: Establishing Post-Training Barriers with Domain Data
After the initial model rollout, post-training barriers relying on exclusive access to vertical-specific and user-specific datasets become vital. Generalist foundation models can solve most daily tasks, but only domain-refined models can master complex work tasks. These models, however, can only be trained on proprietary financial records, coding habits, or industry-specific logs, which demand specialized, hard-to-source inputs. Founders should invest early in acquring exclusive data from targeted niches, since building a unique corpus of domain-labeled data cements technical differentiation. This specialization not only enhances performance—driving superior accuracy in key use cases—but also raises switching costs: customers tied into a vertical-tuned AI ecosystem are reluctant to migrate to general-purpose alternatives. Against the backdrop of increasingly powerful foundation models, obtaining sufficient domain data—even user data—for post-training is more like a customized process. Only through continuous customization with domain data can the differentiation from foundation models be sustained.

Characteristic 6: Co-create Products and Co-tune Models with Your Users
Beyond acquiring domain-specific data for post-training, tuning model inputs and outputs based on users’ own usage habits is equally critical. While asking users to tweak technical parameters may be overly complex, enabling them to start with a set of best-practice templates and make visual, intuitive adjustments is feasible. This approach not only delivers a customized model experience but also allows startups to accumulate preference data, laying the groundwork for future product optimizations. More importantly, this is a process of co-creating the product with users. Once users invest effort in configuring their personalized tools, they may fall into confirmation bias due to sunk costs—continuously reinforcing positive feedback while ignoring negatives. They might even recall their invested effort before abandoning the tool, making resurrection more likely. Balancing user-friendly model configuration without creating experience friction is a delicate challenge, requiring a blend of product managers’ creativity, experience, and data-driven insights.

Efficient Monetization

Characteristic 7: Achieving Efficient Monetization through Upsell
The final moat—efficient monetization—relies on selling more to existing customers while controlling acquisition and support costs. Churn rate will always exist, but good SaaS companies can consistently achieve a Net Dollar Retention (NDR) rate exceeding 100%. The magic lies in upselling. GenAI businesses should embed tiered offerings, modular add-ons, and value-based pricing structures that allow for smooth expandability. Upsell paths may include higher-quality model access, increased usage quotas, premium support, or collaboration features, all presented contextually when users hit a pain threshold. The aim is to escalate average revenue per account (ARPA) without disrupting the user’s workflow or triggering sticker shock. Pricing teams must balance “input-based” metrics (tokens consumed, API calls made) with “output-based” metrics (value delivered, revenue generated for clients) to capture a fair share of the upside. Targeting incremental markets that were previously unserved by AI tools is also an effective strategy to command higher pricing. By architecting a smooth, scalable revenue engine, startups transform individual successes into predictable, repeatable financial growth, reinforcing their commercialization moat.

Characteristic 8: Considering Marginal Cost Optimization from the Beginning
To complement margin expansion, GenAI startups must also relentlessly drive down marginal costs—another dimension of the efficient commercialization moat. It is a new and challenging reality for SaaS companies that the marginal cost of using AI tools is not zero due to the currently relatively high token costs. While token costs for foundation models can be significant, caching of common queries and result reuse across similar customer profiles can slash per-request expenses. Special offers may be provided to additional users to acquire cached content, so that extra free money can be generated by the original request to offset token expenses. Shifting less sensitive workloads to open-source or smaller-footprint models may also help lower costs. Crucially, these optimizations must be planned from day one, ensuring that as scale increases, cost per user falls—creating another flywheel for R&D and growth. Such operational muscle is a defensible barrier: new entrants face the choice of absorbing higher costs or racing to build similar engineering sophistication. By threading cost efficiency into every layer—model selection, serving infrastructure, data pipelines—a startup solidifies its position both on the balance sheet and in the market.

By leveraging these 4 defensible moats, GenAI startups can forge enduring competitive advantage. This framework ties together usage insights, community engagement, technical differentiation, and sustainable revenue, hopefully offering founders a clear blueprint for resilient, long-term growth. Startups that continually evolve their products through real-world feedback, empower users as partners, and build monetizable products around tangible value will not only thrive—they will define the future of the industry.

In the fundraising journey of AI application companies, each stage comes with its own set of risks and opportunities. Compared with the Pre-Seed/Seed round—when teams are still refining their prototypes—and the Series A round—when they’re already expanding into multiple niche markets—the Angel round uniquely combines low-cost exploring with rapid growth. This article examines five key dimensions—product maturity, business model validation, team efficiency, valuation cost-effectiveness, and strategy flexibility—to explain why Angel round represents the best stage for AI application companies to achieve a significant value leap.

vs. Pre-Seed / Seed Round

1. Product Maturity

   • During the Angel stage, the product that startup has been developing takes a leap from its initial Minimum Viable Product (MVP) stage to a fully shippable state, offering not only stable core functionality but also a clear, actionable roadmap. Unlike the Pre-Seed/Seed round—where only prototypes or version 0.1 existed—this stage sees faster iteration cycles, broader feature coverage, and richer use cases, delivering end-to-end solutions that significantly enhance early adopters’ experience and satisfaction. At the same time, by continually refining the user feedback loop, the team can roll out multiple versions in rapid succession, steadily improving the usability and accessibility of key features.

   • During the Angel stage, the user base experiences explosive growth within its niche market, leaping from just 50–100 seed users to 5,000–20,000 active users. This scale effect not only supplies large-sample data for product validation but also lays a solid foundation for quickly replicating the business model. As the customer base expands rapidly, the company can more precisely pinpoint niche market demands, further refine product features, and on that basis engage deeply with potential enterprise clients—thus creating a virtuous cycle of scaled promotion.

   • Overall, the significant improvement in product maturity and the exponential growth in customer base make the Angel round a key “sweet spot” bridging early exploration and scalable expansion.

2. Business Model Validation

   • During the Angel stage, business model validation reaches a substantive breakthrough: the company records Annual Recurring Revenue (ARR) for the first time—surging past the $500,000 mark from zero income during the Pre-Seed/Seed round—signaling the product’s transition from technical validation to commercial value realization. By deploying diversified revenue streams such as SaaS subscription fees and usage-based charges, the business has built an initial, self-sustaining cash-flow loop, and ARR is climbing steadily.

   • At the same time, the team has successfully achieved Product–Market Fit (PMF) in its initial niche market and validated both pricing strategies and customer Lifetime Value (LTV) in realistic scenarios using authentic user feedback. The larger sample size and more representative insights at this stage dramatically reduce the uncertainty around large-scale rollouts. Moreover, this stage is crucial for testing the product’s growth flywheel: optimizing the user journey with a truly frictionless onboarding experience, continuously boosting activation rates through A/B testing, and embedding referral loops—such as in-app invitations and share-to-use features—that leverage the seed user community to accelerate organic growth.

   • The dual validation of ARR and PMF not only boosts the team’s confidence in growth but also lays a solid foundation for future Series A fundraising.

3. Team Efficiency

   • During the Angel stage, the team grows from just 2–5 people to a compact workflow of 7–15 people, striking the right balance between scale and agility. At this size, members can maintain tight collaboration while also specializing in distinct roles. Beyond the early R&D team, the company brings on dedicated product managers and operations managers, significantly deepening expertise and laying the groundwork for community engagement and viral social-marketing initiatives.

   • Compared with the efficiency bottlenecks and decision-making delays caused by individuals wearing multiple hats in the earlier stage, the team at this stage retains startup agility while benefiting from the professionalism that comes with a larger headcount—and at the same time reduces cross-functional friction and duplicate work. Since a startup’s product and market directions often pivot, having a relatively complete organization structure helps the company navigate periods of rapid transformation. Moreover, the compact team size strengthens cultural alignment and lowers turnover, keeping everyone highly focused on the company’s objectives and continuously driving innovation.

   • Overall, a compact workflow of 7–15 people strikes the right balance between cost control and strong execution, serving as the key enabler for rapid product delivery and market expansion during the Angel stage.

vs. Series A Round

1. Valuation Cost-effectiveness

   • In terms of valuation cost-effectiveness, an Angel-stage company is valued at around $8 million—just a fraction of the tens of millions typical in Series A—meaning investors only need to commit a small portion of the capital. This relatively low valuation reduces investors' risk. By investing at this stage, once the company successfully reaches Series A and beyond, its valuation can typically achieve a 5× (or greater) premium, thereby greatly amplifying investment returns.

   • Meanwhile, the Angel stage also sees the fastest user growth: active users jump from an initial 50–100 seed customers to 5000-20000 in a very short period—an over‐100× increase. This explosive growth not only underscores the power of a Product‐Led Growth (PLG) strategy but also validates the success of achieving Product–Market Fit (PMF) within a niche market. Such a large user base lays a solid foundation for subsequent data‐driven optimization, word‐of‐mouth diffusion, and diversified monetization experiments, further driving up valuation.

   • Based on a typical projection, assuming a final Series B valuation of $200 million, Angel-round investors would see a 25× valuation uplift, whereas Series A investors would only achieve 4×. This “low entry valuation + high growth slope” combination makes it the steepest segment on the risk–return curve.

2. Strategy Flexibility

   • During the Angel stage, although the product has reached a shippable state, it hasn’t yet fully entered multiple niche markets or launched multi-channel marketing, so the team retains the flexibility to iterate and fine-tune the product/market direction quickly. By leveraging continuously gathered, authentic feedback from its core user base, the team can nimbly adjust technical approaches or business scenarios, promptly validate or invalidate product hypotheses, and avoid the resource waste that comes from premature commitments. In contrast, during the Series A stage companies typically run multiple growth initiatives in parallel across different niche markets; once they ramp up complex channels like social media and paid advertising, any strategic pivot incurs higher operating costs and greater cross-departmental coordination challenges, making it hard to respond efficiently to new demands, and harder for investors to intake.

   • During the Angel stage, startups lean heavily on a Product-Led Growth (PLG) model, with Customer Acquisition Costs (CAC) nearly zero. Without the need to fund paid ads, built-in product features enable users to refer and share organically, quickly validating growth tactics and refining the user journey. At this point, startups can rapidly test and tweak their growth flywheel, laying the groundwork for replication across additional niche markets. This “asset-light + low-cost + highly-flexible” exploring environment not only maximizes capital efficiency but also builds a robust data and operational foundation for future scalable expansion.

Inevitably, the Angel stage also carries risks. First, although the product is shippable and has achieved PMF in a single niche market, it hasn’t yet begun expanding into multiple niche markets or the broader general market. In the worst-case scenario, investors may find the company stuck in that one niche and unable to sustain growth. Second, moving into multiple niche markets and ultimately the general market brings a greater pool of competitors—from other AI applications to foundational models, and from startups to tech giants—forcing the company to carve out its own ecological niche in a fiercely competitive landscape. Finally, as headcount increases, the complexity of team management and cross-department collaboration rises in parallel, posing a significant challenge to the founders’ experience of running a company.

All in all, the Angel stage might be the “golden ratio” between exploration and breakout: the product is already shippable with a clear roadmap, the business model has been preliminarily validated with ARR in place, the team’s size and efficiency are optimized, valuation remains relatively low while user growth is explosive, and strategy flexibility is preserved. By making precise moves at this stage, one may not only maximize capital efficiency but also lay a solid foundation for later large-scale expansion—creating a win-win for both founders and investors.

Podcast version above is automatically generated by NotebookLM.

Last year, when Sand Hill Road funds went all-in on GenAI, I started to feel somewhat anxious. I was previously focused on Deep Tech investment, and I had to push myself to pivot into GenAI investing. From paying attention to a new sector to deal sourcing and researching, and then to hands-on investing, there are significant gaps to fill in. This year, after DeepSeek and Manus showing their disruptive capabilities, almost every Chinese investor started looking into AI. There was no longer any diverse opinion, and my anxiety deepened.

When meeting with founders, I found out that many of them are Gen-Z people like me, but with more first-hand knowledge about GenAI than I have. This is a whole new founder persona, different from the Deep Tech founder that I am used to co-work with. I feel easy to empathize with these founders, since I have nearly the same age as them. But I still have encountered challenges as follows:

1. The industry is immense, spanning from AI infrastructure to all kinds of AI applications. From which sectors should I seek entry points?

2. Are there technological barriers inherent in AI applications? If technological barriers are absent, what alternative forms of competitive advantage should AI applications develop?

3. As large language models (LLMs) continue to advance in capability, will many currently popular products ultimately prove to be merely transitional in nature?

4. Should I invest in AI-Native applications or AI-Powered applications?

Click here to check out the previous article.

Third question:

As large language models (LLMs) continue to advance in capability, will many currently popular products ultimately prove to be merely transitional in nature?

• I consider this from several aspects.

• First, is it possible for a single model to dominate all scenarios? Actually, this seems to be the ultimate form of AGI. But in current stage, you can observe that in an AI Agent or an Agentic Workflow, models of various parameter sizes from different suppliers are used in parallel. Not every task requires the full-scale model—some 3B models can perform certain tasks better, faster, and cheaper. Even foundation model suppliers provide models in different sizes. Different tasks require trade-offs among quality, time, and cost, making it difficult for a single model to dominate all scenarios.

• Second, is it feasible for the orchestration of several models to dominate all scenarios? Many Agent frameworks orchestrate multiple Agents—some designated for planning, others for reflection, and some for execution—to accomplish a variety of tasks. I would like to introduce a coordinate: the horizontal axis represents the degree of creativity dependency (essentially reflecting human involvement), and the vertical axis represents the degree of task standardization (whether there is a SOP or repetitive execution).Utilizing this coordinate system, tasks can be divided into four quadrants.

• The tasks most suitable for direct use of foundation models or general purpose agents are those with low creativity dependency and relatively divergent needs (low standardization), such as DeepResearch, where the goal is clear (researching a specific topic). Even though sometimes a universal research methodology exists, the workflow remains relatively divergent and requires flexible extraction and processing of information.

• The adjacent quadrant, characterized by low creativity dependency and highly standardized tasks, is better suited for old-school programming rather than using foundation models.

• Another quadrant, involving high creativity dependency and divergent workflows, such as art creation or cutting-edge science exploration, still presents challenges at this time and is better done by humans.

• The final quadrant, with high creativity dependency and highly standardized work, such as formulaic writing, is well-suited for AI-Powered tools.

• However, this coordinate system is not static. The quadrant suitable for the direct use of foundation models and the one suitable for AI-Powered tools are constantly encroaching on the quadrant suitable only for humans. In other words, more and more tasks traditionally suited for humans can now be replaced or partially replaced.

• Even within the AI-Powered tool quadrant, much of the engineering effort may be diminished by the iteration of foundation models. Eventually, it resembles a pyramid shape.

• After considering which scenarios suit foundation models/agents, AI-Powered tools, or humans, does it mean everyone should focus on developing and investing foundational models? Not necessary. Here’s my radical view: the pyramid shape above is based on workload, but if we look at value generated, it might be the opposite—a reversed pyramid shape. Deployment of model is relatively simple and quickly becomes commoditized. In contrast, the deployment of humans, as in the case of newborns, is challenging and will become increasingly scarce in the future.

• So my radical view is to lean towards the user and humanity, where future value will be greater. In product development, one should continually move closer to the user, ultimately becoming the interface through which users understand and utilize AI. If most products are indeed transitional, this may be the key to securing an ecological niche.

This naturally leads to the fourth question:

Should I invest in AI-Native applications or AI-Powered applications?

• Firstly, AI-Native applications, or direct use of models/agents, may easily become commoditized, but that does not mean they are not valuable. Instead, they will serve as infrastructure. Foundation models will be as essential as utilities or cell networks. Token bills look like phone bills anyway. Therefore, MaaS(Model as a Service) will be the battleground for the giants.

• For startups, the goal is not to compete head-on with the giants, but to leverage resources to break through from a small entry point. This inevitably raises a key question: What is the potential size of the market?

• In deep tech investments, I have summarized two points that may also apply to AI investments:

  • Existing markets are established and crowded with existing players. Startups must look for an incremental market, which is hard to estimate but has the potential to become larger. Incremental markets are vague and hard to estimate so that giants might miss them, giving startups the opportunity to deliver products first.

  • Second, in hard tech, I have invested in many upstream components. These components typically have a limited market size when used in a single application, but as platform technologies, they can be applied across various applications. For example, lasers can be utilized in lithography systems, optical communication transmitters, LiDARs, and more. As platform technologies, they can "eat from multiple bowls." When investing in AI, it is crucial to consider from the outset how to expand the user base and ultimately build a platform. Whether in deep tech or AI, what eventually leads a startup to an IPO often differs from initial expectations. This requires a flexible team, with founders and investors maintaining an open mindset.

Podcast version above is automatically generated by NotebookLM.

Last year, when Sand Hill Road funds went all-in on GenAI, I started to feel somewhat anxious. I was previously focused on Deep Tech investment, and I had to push myself to pivot into GenAI investing. From paying attention to a new sector to deal sourcing and researching, and then to hands-on investing, there are significant gaps to fill in. This year, after DeepSeek and Manus showing their disruptive capabilities, almost every Chinese investor started looking into AI. There was no longer any diverse opinion, and my anxiety deepened.

When meeting with founders, I found out that many of them are Gen-Z people like me, but with more first-hand knowledge about GenAI than I have. This is a whole new founder persona, different from the Deep Tech founder that I am used to co-work with. I feel easy to empathize with these founders, since I have nearly the same age as them. But I still have encountered challenges as follows:

1. The industry is immense, spanning from AI infrastructure to all kinds of AI applications. From which sectors should I seek entry points?

2. Are there technological barriers inherent in AI applications? If technological barriers are absent, what alternative forms of competitive advantage should AI applications develop?

3. As large language models (LLMs) continue to advance in capability, will many currently popular products ultimately prove to be merely transitional in nature?

4. Should I invest in AI-Native applications or AI-Powered applications?

First question:

The industry is immense, spanning from AI infrastructure to all kinds of AI applications. From which sectors should I seek entry points?

• We are presently in the nascent stage of the GenAI era, presenting significant opportunities to invest in leading players. These prominent companies typically exhibit distinct characteristics, such as focusing on consumer-oriented products and targeting the largest possible user base. Notably, there is currently a divergence in value consensus between China and the United States regarding AI. While American technology giants concentrate primarily on foundational models, numerous teams in China have begun to develop applications and solutions.

• The underlying secret conferring an advantage to Chinese AI companies is frequently attributed to two key factors. First, China possesses an abundance of engineers, which enables rapid product development and iteration. Second, there is also a substantial pool of highly experienced product managers within the country. If one were to speculate whether the next historically viral app, akin to TikTok or Pinduoduo, would emerge in China or the United States, the prevailing consensus would likely favor China. This perspective is rooted in the extensive product development expertise accumulated during the PC Era and Post-PC Era.The rationale for referencing TikTok and Pinduoduo, as opposed to WeChat or Alipay, lies in the discernible distinction between what I called, Post-PC Era 1.0 and 2.0. A tangible manifestation of this difference is that TikTok and Pinduoduo(Temu) have demonstrated the capability to achieve dominance in overseas markets, whereas WeChat and Alipay have not. As representatives of Post-PC Era 2.0, products such as TikTok and Pinduoduo exemplify how quantitative accumulations in product design can ultimately lead to qualitative transformations.

• Thus, a third meaningful point is to focus on industries where China already has an advantage, such as e-commerce (especially short-form video commerce and live commerce), and creative industries (short-form video, novels, comics, animation, gaming and so on). Should AI applications achieve widespread adoption in these sectors in China, it is anticipated that their international adoption will be comparatively straightforward. Essentially, this is about learning from leading players.

Second question:

Are there technological barriers inherent in AI applications? If technological barriers are absent, what alternative forms of competitive advantage should AI applications develop?

• This has been a persistent confusion since I pivot into investing AI. Deep tech companies can establish technological barriers for a period, allowing first-mover to convert technological barriers into other advantages. In slower-iterating industries, technological barriers can be converted into economies of scale. And in faster-iterating industries, robust customer relationships and continuous product iteration can serve to sustain technological barriers.

• In the context of AI applications, there do not appear to be particularly evident technological barriers. This raises the question: what alternative forms of competitive advantage should AI applications develop? What core capabilities are companies fundamentally competing on?

• Companies differentiate themselves by the depth of their understanding of users. Achieving this requires first clearly identifying the core user group. Whether in software or hardware, great companies typically begin by developing a SOTA product tailored to a specific niche market, thereby capturing that segment comprehensively. As additional customers recognize the product’s applicability to adjacent niches, the market naturally expands and the product achieves broader adoption. Companies such as DJI and Shokz exemplify this approach. In the early stages, the core user base does not need to be large, but their persona must be clearly defined. Only with a precise user persona can a company distill the most streamlined product definition and deliver a best-fit, SOTA solution for its users.

• Once the user persona is established, the next thing is about finding user needs. Even with a clearly defined user profile, there will inevitably exist both urgent and less urgent user needs. A genuine and urgent need is often characterized by a misalignment between the tasks on which users spend most of their time and those that generate the greatest value—where substantial time is devoted to routine or menial activities, while comparatively little time is spent on value creation. If a product can enable users to focus their efforts entirely on value creation, delegating repetitive or low-value tasks to AI, the overall value generated by users will be significantly amplified. This principle is consistent with the philosophy behind Vibe Coding and can be extended to broader concepts such as Vibe Creating or even Vibe Anything, forming a fundamental logic for how AI is poised to reshape applications.

• Finding user needs is like cracking a safe—listening carefully, trying repeatedly, and finally unlocking it. Once needs are identified, These needs should be a guidance to product definition. Product definition involves determining what to prioritize and what to exclude. It becomes possible to streamline the product by eliminating unnecessary features and selectively introducing key enhancements, thereby leveraging critical advantages to expand the user base. Throughout this process, it is also essential to recognize the limitations of AI, understanding both its capabilities and constraints. By adopting a multidimensional approach to user interaction and incorporating diverse functionalities from an engineering perspective, one can effectively compensate for AI’s shortcomings and deliver a comprehensive product experience.

• If a company possesses the aforementioned capabilities and has established a system to ensure sustained excellence, it should be able to secure a distinct product-level advantage. AI applications iterate rapidly and are relatively difficult to replicate; without accurate insight into core user needs and continuous innovation, competitors cannot easily duplicate or deliver a comparable user experience. Beyond user insight, what additional qualities should startups possess? Can the success of TikTok and Pinduoduo be distilled into key characteristics of exemplary products?

• TikTok exemplifies strong user engagement, underpinned by two core features: personalized recommendations and a vibrant content community. Personalized recommendations continually enhance the user experience, making it increasingly enjoyable and precise, thereby creating a data-driven flywheel effect. The content community unites creators and consumers, establishing a dual-sided network effect in which growth on one side increases value for the other. This dynamic attracts additional users and generates a second reinforcing flywheel.

• Pinduoduo represents community marketing and viral growth. Campaigns like UGC content seeding and check-in marketing are also forms of community marketing. At a deeper level, these strategies are fundamentally about persona building and trust-based dissemination. For example, sharing one's running route image using Keep, a Chinese fitness tracking app similar to Strava, builds a persona of fit and disciplined, while also reinforcing the brand association with Keep, resulting in a win-win and encourage sharing. Platforms like Substack and Tencent IMA employ analogous strategies—paid knowledge sharing not only fosters a knowledgeable persona but also generates tangible income, further reinforcing user recognition of the platform.

To Be Continued...