a16z's latest insight: If an AI product doesn't explode on the Social Web within 48 hours of its launch, it's equivalent to a death sentence.

Source: Andreessen Horowitz

Compiled by: Xinyi Fan, Z Finance

In today's AI era, where refresh rates determine life and death, distribution is no longer just a part of the growth strategy, but the core variable for product success or failure. The update frequency of foundational models and underlying tools is measured in weeks, and the product iteration window has been compressed to the extreme, resulting in highly fragmented user attention. In such an environment, the traditional concept of a "moat" is disappearing, and speed and momentum are taking its place—whoever can occupy the mental high ground of users first will be able to break through the homogeneous competition.

a16z's new episode focuses on the profound changes that are reshaping the AI startup landscape. The guest is Anton Osika, co-founder of Lovable—a rapidly rising player in the field of AI product internationalization and social distribution. Under his leadership, Lovable achieved an annual revenue of ten million dollars within just two months of its launch, not due to any miraculous breakthroughs in the model itself, but because he deeply understands the power of "first-mover advantage." In the AI field, even if you have the strongest technology, if you can't present your product's advantages in a way that captivates users immediately with a buzzworthy topic, you may be instantly overwhelmed by competitors who are better at distribution.

Osika pointed out that the rules of the game for AI entrepreneurship have fundamentally changed. In the past, entrepreneurs could spend months refining products, optimizing user experiences, and then seek distribution strategies; however, now a product that does not achieve social diffusion within the first 48 hours is likely to be sentenced to an "invisible death" from the very beginning. The challenge faced by today's AI startups is not "Can I make it?" but "Can I quickly gain traction and sustain growth?" Technical differences have become increasingly weak under the trend of homogenization among large models, while distribution efficiency, topic explosion, and user emotional engagement are the key factors determining how far a product can go.

The program will further explore a new paradigm practiced by Anton: rapidly creating brand narratives and user engagement through open construction, live demos, and initiating social challenges; establishing product reputation and native culture through early involvement of influencers within the community; and forming a collaborative "Starter Pack" by linking with other AI tools to achieve low-cost and high-quality distribution synergy. The commonality of these practices is that they do not rely on large market budgets, nor do they excessively depend on channel resources, but rather maximize the communication effect of each product iteration under the rules of social networks.

In this AI cycle of "if you don't distribute, you haven't done anything," the approach represented by Anton Osika and Lovable may be the key path for AI companies to navigate through the clouds and build a momentum-based moat. The true moat is no longer a technological barrier that others cannot imitate, but rather the speed and structural cognitive gap that others cannot keep up with.

Waking up early is crucial.

In the consumer-grade AI field, how to build a moat? I’m sorry to say that there is currently no moat at all. The changes in this industry are just too fast — the foundational models and underlying infrastructure are changing almost every month, with new updates released almost every week! In such a dynamic environment, it has become nearly impossible to build products slowly and systematically like in the mobile internet era. At this moment, the most critical factor is speed: how quickly you can launch products, how quickly you can gain user attention, and how quickly you can occupy users' minds.

Every startup hopes that its product will become a hit. But today, this is harder than ever: the number of AI product launches is huge, the speed of updates and iterations is extremely fast, social algorithms are unpredictable, and coupled with the homogenization of underlying models, achieving genuine explosive growth is becoming increasingly difficult.

Traditional distribution strategies and growth methods (even for productivity tools or useful products aimed at professional consumers) are no longer as effective. To put it bluntly, in the words of my colleague Andrew Chen: all marketing channels are ineffective now. Paid acquisition and SEO may still bring temporary user growth, but in consumer-facing AI, they struggle to deliver sustained user retention. You must break the mold.

To explain the current industry dynamics to the founders, I used a somewhat "weird" metaphor: starting an AI company now is like throwing a pigeon into the sky and then praying that it can fly.

Today, a flock of AI startups is soaring together like a group of pigeons, striving to accelerate and climb higher to avoid running out of momentum and falling from the sky. These companies are 'launched' into the air one after another, often building similar products and sometimes even using the same underlying models. Some pigeons fall down shortly after taking off; some can only reach a certain height and then stagnate, their acceleration slowing down until they are exhausted, possibly opting for a soft landing (such as being acquired or quietly transforming). But a very few will shoot straight into the clouds, breaking through the layers and continuously rising, leaving the other pigeons far behind.

They have become part of mainstream cognition and occupy a high ground in users' minds.

However, even if you have already soared to the clouds, in the AI industry, you still must keep striving and flapping your wings hard. If you can launch new capabilities, new features, and new models faster, you will be able to widen the gap between yourself and the second fastest, the third fastest, and even the whole flock.

The real moat is momentum.

So what does all this mean? Early distribution is crucial. Of course, the buzz generated by distribution alone cannot retain users, provided that your product can keep up continuously. When you can iterate on your product quickly, each update is a new opportunity for showcase and promotion. Companies that understand this dynamic and build their products explicitly around it, such as Perplexity, Lovable, Replit, and ElevenLabs, are gradually pulling away from other competitors.

So, how can you make your "pigeon" soar vertically and keep rising? Here's a spoiler: there isn't a ready-made success manual yet, because the rules of the game at this stage are based on novelty and creativity. However, below are some effective distribution strategies we have recently observed, along with case analyses behind them:

Hackathon: Reborn in the form of a public performance

In the past, a Hackathon was a small circle event aimed at developers. But now, it feels more like a public performance: widely disseminated through live streaming and social media, its purpose is to expand distribution influence. At the same time, AI-native tools have significantly lowered the participation threshold. Such events provide a potential stage for new projects that support your product to gain popularity.

For example, ElevenLabs held a global hackathon earlier this year to showcase the potential of its AI voice platform. Developers were invited to build various projects based on it, ranging from role-playing robots to interactive audio applications. And during a demo showcase called Gibberlink, an unexpected thing happened: an AI voice suddenly realized that it was conversing with another AI.

In that non-scripted exchange, the two AIs conversed in a tone resembling that of humans, sparking heated discussions on social media. This not only showcased powerful technological capabilities but also became a discussion point with a "quirky" cultural twist: whether AI has self-awareness and the authenticity of voice simulation. This event brought significant exposure to ElevenLabs.

For example: Lovable recently held a live showdown where a senior designer using Webflow competed against a "vibe coder" using Lovable's AI design assistant to see who could create a better landing page. The competition was set with a time limit and was broadcast live, significantly increasing the tension. The focus of the show was not on who won in the end, but on showing the audience that AI is lowering the barriers to design, and may even enable non-professionals to outperform professional designers. This not only showcases the practical application scenarios of Lovable's products but also injects interesting narrative material into social platforms.

Social experiments, the more "extreme" the better

Building on the above trend, some companies are taking it a step further. Bolt recently announced that they will challenge the Guinness World Record by hosting the largest hackathon in history, targeting even non-developers, with a total prize pool of up to $1 million.

Similarly, there were a series of social challenges initiated by Genspark this spring, encouraging users to try and defeat its super AI assistant. Participants were invited to pose complex or quirky questions to the AI in an attempt to reveal its limitations. The most creative or profound failure cases could share a $10,000 prize pool. Such activities are low-cost but can generate a lot of discussion and user interaction.

Let's look at another example: In China, a top venture capital fund held a three-day Truman Show-style experiment: developers were locked in a room, given a computer, and could only use generative AI tools, with the goal of making as much money as possible. This reality show-style gimmick is clearly performative, but that's exactly the point. The experiment not only garnered media coverage but also sparked widespread discussion on social platforms.

AI "Beginner Pack" and Alliance Strategy

Today's users often need to piece together multiple AI tools themselves: generating, editing, optimizing, and outputting. The constant switching between tools can be overwhelming. In such a fragmented ecosystem, collaboration is strength.

We are seeing an increasing number of leading AI companies collaborating to launch joint releases or feature integration packages, spreading products in a combinatorial manner and driving traffic to each other. These viral Starter Packs demonstrate the potential for collaborative use of tools.

For example: Captions collaborated with Runway, ElevenLabs, and Hedra to create a complete video generation stack, forming a one-stop AI video production process from text generation to voiceover; Bolt launched a carefully crafted builder toolkit, packaging AI infrastructure and creation tools such as Entri, Sentry, Pica, and Algorand; Black Forest Labs, when releasing its new model Kontext, showcased alongside partners such as Fal, Leonardo AI, Freepik, and Krea.

These Starter Packs are not just a marketing gimmick; they also possess real functional integration value, showing users that from creativity to output, there is no longer a need to piece things together, as this combination can get the job done.

In addition, they also create a social endorsement effect: each partner increases the credibility and brand influence of the others.

Join forces with influencers in the industry to build a moat.

Another strategy for building a moat is to let AI-native creators, developers, and designers speak for you. This is not referring to traditional influencers or brand ambassadors in the conventional sense. Traditional influencer marketing is becoming increasingly ineffective: high investment, low output, quick influx of traffic that dissipates just as fast, and low conversion rates.

In contrast, truly leading AI companies are beginning to open early access to influential niche native users within the industry. These individuals may not have millions of followers, but they hold significant sway in specific communities, forums (such as Reddit, Discord), and creative communities on the internet, which can genuinely influence the reputation and adoption rate of the tools.

For example, Nick St. Pierre is the "natural evangelist" of Midjourney, and his early works using generated images have become widely known; Luma AI has also adopted a similar strategy, granting early access to a small group of AI-native creators; before the release of Veo 3, filmmaker Min Choi and PJ Ace tried out the model and created content in advance, attracting widespread attention.

PJ Ace once tweeted: "I used to spend $500,000 to shoot a pharmaceutical commercial, but now I only used the $500 credit limit on Veo 3 and a whole day to get it done." "Who can still pay $500,000 for an ad now?"

This type of content is not only a product demonstration but also a persuasive and authentic recommendation, reinforcing user perception through the perspective of "insiders."

Direct Strike: Use "Publish Video" as a Distribution Strategy

You may have heard the saying: "show, don’t tell," but in the age of AI, it has transformed into "show, don't pitch." Traditional public relations are too slow and rigid for the rapid pace of current AI; on the contrary, we see many little-known small teams achieving remarkable results solely based on a brilliant product demonstration and an intuition for storytelling.

As Kevin Kwok said, "When did it become necessary to shoot videos for all new product launches? This trend has changed so quickly."

For example, when the Chinese startup Manus launched its universal AI assistant, it did not hold a press conference or run advertisements, but instead directly uploaded a 4-minute demo video on X and YouTube. The video showcased the product's powerful features, attracting widespread attention with over 500,000 views.

Behind this change, there is another underlying transformation: more and more startups are appointing a growth leader who understands technology, even to the extent of being called a Chief Flapping Officer: not only responsible for operational growth strategies but also personally involved in creating interesting and even quirky interactive demos, aiming for viral dissemination effects.

For example, Luke Harries from ElevenLabs is a typical representative. He not only plans marketing campaigns but also hands-on works on projects, such as building an MCP server demo for WhatsApp. These quirky construction projects often become unexpectedly popular.

Another similar figure is Ben Lang. In his early days at Notion, he was responsible for creating interesting demos, niche showcases, and design play, quietly shaping Notion's community culture and brand identity even before the product became widely known. Now, he holds a similar role at Cursor, publicly building projects and turning each product release into a shareable story and content.

Build in Public

In the past, growth data was a secret that companies carefully disclosed only to investors. Nowadays, more and more AI companies choose to publicly build: sharing product progress, user data, revenue milestones, and even failed experiments.

For example, a tweet published by Genspark on a social platform: "Achieving an annualized revenue (ARR) of $36 million in 45 days?! That's right, our small team of only 20 people might be the fastest-growing startup in history. No fancy marketing, no advertising, all through word of mouth from users." They also included a list of recently released products: Genspark AI Sheet, Agentic Download Agent, etc.

Others like Lovable, Bolt, and Krea have also taken similar approaches. They regularly update on social platforms, sharing insights from revenue growth, DAU (Daily Active Users), to reflections on experimental failures, making users feel like they are part of the building process, not mere bystanders or AI tourists. Lovable's founder Anton Osika tweeted in January 2025: "Lovable has achieved a $10 million annual revenue goal today - just two months after launch. Growth is continuing to accelerate." He also provided an analysis of the product's advantages compared to other competitors (unfolded in thread form).

This openness and transparency also brings about a hidden competitive effect: when a company's product breakthroughs, user numbers, or revenue are showcased, it stimulates startup companies in the same field to roll up their sleeves, rather than just presenting demos, showing growth charts, or posting user feedback. This atmosphere of "you share data, and I follow suit" actually promotes the overall dissemination efficiency and momentum accumulation of the entire ecosystem.