When Y Combinator began its first recruitment in 2005, only eight companies participated. Among them was Reddit, which would go on to become a giant in the social media landscape. Other early successes included startups like Twitch in 2007, Airbnb in 2009, and later Coinbase, Instacart, and DoorDash. These companies didn’t simply promote technology; they built it from the ground up. Their core was engineered by developers, not marketers, integrating pre-built software.


In the early 2010s, YC startups continued this approach - innovative, with distinctive technological differentiation. They offered products that were grounded in complex engineering solutions, seeking new approaches and pioneering technological advances, rather than making superficial adjustments to existing products.


But in 2025, the focus has shifted. AI is no longer just a trend - it has become the foundation of a new economic cycle, one where it is now a standard practice.

1. Repeatability or specialization?

The issue of repeatability has become a central question for YC, especially in light of the rapid growth in projects focused on AI and data solutions. This surge in similar projects raises the need to rethink how startups are built and what makes them stand out. Of the 144 startups featured in YC’s Spring 2025 batch, 67, or 46%, were companies leveraging automated systems and data-driven solutions.


In 2024, Deckmatch, a data analysis startup, conducted a study inspired by the controversy surrounding the PearAI startup, which was supported by YC. Critics argued that PearAI’s product, a code editor, wasn’t a novel offering, but merely a replication of another YC product, Continue. Unfortunately, PearAI's founder could not defend their product’s originality and was forced to commit to starting over.


A pattern that began to emerge among startups in 2025 follows a similar trajectory:


Many startups are now building solutions based on pre-existing technologies, slightly adjusted to serve niche industries. However, a critical question arises: Is this approach really about innovation, or is it just scaling basic solutions without adding anything new or unique?

2. B2B businesses selling solutions to each other

One of the main criticisms of Y Combinator in recent years is that many YC startups have become a network of B2B businesses that primarily sell their solutions to each other within the ecosystem.


Of the 636 startups in 2025, over 400 are focused on B2B models. This means that most YC startups are developing solutions aimed at businesses, but they often stay confined to the startup world and the tech companies within the YC network. As a result, many of these startups fail to tap into larger markets, limiting their potential to grow beyond internal sales.


For example, many YC 2025 startups offer solutions that support other startups in areas such as data management, process optimization, software development, and more. Instead of targeting larger corporations or industry leaders, these startups predominantly sell their solutions to others within the YC ecosystem. This creates a cycle where some companies simply exchange solutions with others, rather than offering them to external customers who could truly benefit.


When every startup in YC follows similar marketing strategies and uses similar technologies, it raises an important question: how can these companies stand out, and how will they expand beyond the YC ecosystem?


This highlights a paradox within the current state of YC startups: many seem less focused on transforming industries or driving meaningful innovation, and more focused on replication. The "B2B solutions for B2B solutions" strategy could create long-term challenges for these companies, as they risk remaining confined to a small ecosystem without ever reaching the broader market.

3. Why are engineering teams disappearing?

3.1 AI has become the building block, and marketing drives the product

Integrating AI tools into startups today has become more about incorporating ready-made solutions than tackling complex engineering challenges. This doesn't require deep expertise in machine learning or technical systems; rather, it's about effectively integrating and customizing these tools for specific applications.


Here's how it typically works:


The issue is that integrating these tools into a product and developing the solution doesn’t demand significant engineering work. Instead, the focus shifts to marketing and presentation. Many products are simply repackaged versions of existing solutions, marketed as “assistants” or “helpers,” which look innovative at first glance, but are often just basic integrations with minimal changes.


In this context, marketing becomes more important than the technical features of the product. A catchy name and a well-packaged product can make an idea appear revolutionary. For instance, “An assistant for doctors that reduces waiting room time” might seem innovative, but it could simply be a repurposed solution, rebranded with an appealing name.


However, marketing itself becomes a challenge. In a market where many products are similar, differing mainly in user interface or niche applications, the marketer's task is to convince potential users and investors that their product is truly unique. This is increasingly difficult because, in many cases, what is being marketed is more about the idea than the actual technology behind it. "Selling an idea" doesn’t always guarantee long-term competitiveness.

3.2 Time to market has shortened

In the 2025 update, YC announced that it would release four batches per year (winter, spring, summer, and fall) partly due to the accelerated pace of change in the startup landscape.


Advances in AI tools and the availability of powerful, ready-to-use solutions have drastically reduced the time needed to create new products. In the past, developing a product from scratch, including architecture, pipeline creation, and system testing, could take months or even years. Today, an can be created in just a few weeks. The emphasis has shifted from technical complexity to speed in launching products.


Now startups can:


Thus, marketing, selling ideas, and rapid integration take precedence over creating unique technical solutions. This marks a significant shift in how startups are approached and how new products are developed, focusing more on the speed of bringing products to market than on the originality of the technology behind them.

4. Why do we need engineers?

Engineers remain essential because startups that rely solely on off-the-shelf solutions lack the ability to grow and scale in the long term. Products built on existing, readily available tools without unique innovation can be easily replicated. These products are vulnerable to competitors and lack any competitive advantage. This creates significant risks to the sustainability of such companies.


What does it take for long-term success? The sustainability of startups is only possible through the development of unique solutions that are hard to replicate.


Engineering-focused startups create the foundation that allows them to build:


Only with real engineering solutions can a startup create products that have a lasting competitive advantage. Without this, startups become mere "wrappers" around ready-made tools, offering little differentiation and no long-term value in an increasingly competitive environment.

5. Summing up

AI opens up huge opportunities and really transforms entire industries. But at the same time, technological progress creates new challenges. Previously, creating a product in a startup required months of work with architecture and code. Now, using ready-made LLM models and their API, you can build an MVP in a few weeks. However, such changes also have profound consequences for the startup ecosystem.


Today, many YC startups are companies that use external models as the basis of their product, rather than building innovative solutions from scratch. The risk of product uniformity and reduced engineering originality are becoming more pronounced. These companies rely on ready-made models and simply integrate them with minimal changes, creating "wrappers" around powerful tools like GPT. At first glance, such solutions make launching a startup to the market faster and easier, but there is a question: what will remain if a new version of such a model appears tomorrow or someone offers even faster and cheaper integration?


This trend creates new challenges for the ecosystem. The need for deep technical expertise is reduced. Engineers who once created real technological breakthroughs are gradually giving way to integration specialists, marketers and managers, whose task is to quickly launch an "AI product" in a chosen niche. And although this reduces barriers to entry and speeds up the process, there is a risk that startups begin to compete not for technological uniqueness, but for speed - the speed of launching to the market, the speed of attracting attention and the speed of copying.


At the same time, Y Combinator 2025 is still a platform where dozens of interesting ideas and innovative solutions are born. This is a place where you can often see new approaches that go beyond standard solutions. But an important shift is that the focus has now shifted from “engineering innovation“ to ”rapid startup on top of existing LLMs". And here's an open question: is it possible to grow the next Reddit, Twitch, or Airbnb in such an environment? Is it possible to create not just another variation on the theme, but a product that will radically change the entire industry, as startups did a decade ago?


This issue remains open and complex. Time will tell how this ecosystem will develop - whether there will be startups that can go beyond basic integration and offer truly original and technologically sophisticated solutions. It is important to understand that true innovation requires not only the ability to assemble a product in a couple of weeks, but also the ability to create something more than just a "wrapper".