Discover more from Posts in the Shell by Demren Sinik
Artificial Intelligence is a Threat to Founders
The future of software has never looked brighter – but mostly for buyers and incumbents
All founders, either explicitly or implicitly, believe the following to be true:
Customers will use my product instead of building it themselves
Customers will use my product instead of my competitors’ products
Fortunately, innovation has historically favored the founder. Despite deeper resources, customers and competitors are often too constrained to outcompete incredible new products. This cycle of innovation has historically enabled entrepreneurs to realize their visions and build valuable (and, at times, generational) companies. Today, entrepreneurs and venture capitalists believe artificial intelligence will drive the next wave of innovation. Unfortunately, most entrepreneurs and VCs are missing the bigger picture.
AI is poised to drastically lower the constraints that have previously enabled startups to succeed. Without these constraints, it will become harder for founders to build generational companies, especially in software.
This post explores the future of software – what some call “Software 3.0.” My goal is to illustrate why Software 3.0 poses unforeseen risks to founders. I believe that by keeping these risks in mind, founders can direct their incredible talents towards solutions and markets that deserve to be won by newcomers in the current age of AI.
What is Software 3.0?
Software 1.0 is how we typically think about software. The Software 1.0 programmer writes a set of instructions in a coding language, those instructions are converted into machine code by a compiler, and a computer reads and executes the machine code.
Software 2.0 is a term coined in a 2017 blog post by Andrej Karpathy. If code is the primitive for Software 1.0, then neural networks are the primitive for Software 2.0. To paraphrase Karpathy: The Software 2.0 programmer specifies some goal on the behavior of a desirable program, writes a rough skeleton of the code (i.e., a neural net architecture) that identifies a subset of program space to search, and uses the computational resources at their disposal to search this space for a program that works.
Software 3.0 is the future of software. The Software 3.0 programmer provides a set of instructions and a dataset, and an AI agent uses these inputs to generate a program. Though it will take a while to realize the full potential of Software 3.0, the building blocks are already here. Today, you can query ChatGPT for a piece of code, use GitHub Copilot to suggest code in real-time, or automatically debug your code with Codium. You can also give autonomous agents a goal written in natural language, and they will attempt to achieve it by breaking it down into subtasks. Sometime in the future, we may be able to prompt an agent to create an entire tool, workflow, application, and even a software platform with a high degree of accuracy and reliability.
For the purposes of this post, we won’t debate the details of Software 3.0 – the timeline, its precise capabilities, etc.
Instead, we will try to understand how Software 3.0 can impact the future of entrepreneurship. As mentioned, all founders share the same two beliefs: (1) customers will use my product instead of building it themselves and (2) customers will use my product instead of my competitors’ products. But how well will these beliefs hold up in the era of Software 3.0?
To Buy or To Build?
I believe customers of startups will increasingly become competitors to startups. Founders who have started companies in the past year or are in the process of starting companies need to be especially aware of this threat.
To understand why, let’s look at the primary reasons why companies have historically purchased software instead of building it themselves:
Figure I: Nine reasons why companies buy third-party software rather than build it in-house
To assess the impact of Software 3.0, it’s important to understand which of the above nine reasons are impacted by artificial intelligence. It stands to reason that if artificial intelligence can reduce some barriers to building software (without raising others), then companies will build more software than buy it in the future.
Figure II: Predicting how Software 3.0 will impact the build vs. buy decision tree
As we can see, the outcomes are mixed. Building internally with Software 3.0 isn’t necessarily better than buying third-party software. Most likely, companies will in-house simple software tools that can easily be built with Software 3.0 but will not drastically in-house more complex platforms that are purchased for other reasons beyond resource limitations.
But software founders should not start celebrating just yet. Figure II only measures a customer’s willingness to replace a mature software solution with an internal version. In reality, early-stage companies rarely demonstrate these nine characteristics. Early-stage startups don’t have network effects, they have nothing to bundle, their security and privacy standards are worse than mature alternatives, they may raise regulatory concerns, they certainly don’t have a brand, and they have very few contracts. Instead, they are usually more willing or able to direct resources to a specific problem. And in some cases, they might have better customer service and integrations.
Figure III: Predicting how Software 3.0 will impact the build vs. buy decision tree for early-stage startups
In other words, early-stage software companies have far fewer characteristics that drive customers to purchase their product. And of those characteristics, the majority are likely to lose out to Software 3.0.
For early-stage entrepreneurs, you need to be thinking several steps ahead. It’s not enough to have a good idea and the resources (people, knowledge, and time) to devote to it. Instead, you need to build a product that lends itself to some other form of defensibility to prevent your potential customers from building it internally.
The Rising Threat of Competition
Preventing your product from being in-housed is only half the battle. Once customers decide they need to purchase solutions like yours, you need to convince them to purchase yours vs. someone else’s.
Unfortunately, competitors will become far more dangerous in the Software 3.0 paradigm. This includes both early-stage and mature competitors:
Code generation will reduce IP and unique insights: LLMs are really good at distilling the corpus of human knowledge and making it available to a broad set of people. As artificial intelligence capabilities improve, it will be more difficult for founders to identify unique insights and then build unique products on top of those insights. Furthermore, unique insights, once identified, will become more easily replicated with fewer resources.
Venture capital is oversaturating the market: VC investment is flowing freely in AI, and that capital is rapidly funding many competitors in many categories. This trend is always present in tech but becomes particularly exacerbated during hype cycles.
Code generation also enables incumbents: Code gen doesn’t just make the early-stage market more competitive; it also unlocks massive efficiency gains among mature companies with large engineering teams. Every early-stage founder should imagine a world where their mature competitors become at least twice as efficient/innovative as they have been in the past.
Customers are seeking vendor consolidation: Customers are tired of software bloat. In a high interest rate environment, they’re also prioritizing cost savings and don’t want to pay for too many solutions. This pressures markets to consolidate around mature market leaders.
Incumbents are seeking the next wave of growth: Public growth companies (or growth companies seeking to go public) are seeking ways to maintain strong top-line growth to support a premium valuation, particularly in an environment where valuation multiples have fallen from prior highs. Expect mature companies to use artificial intelligence to re-imagine their core product or move horizontally into tangential product lines.
Stakeholder pressure to invest in AI capabilities is high: The hype around AI has impacted all companies, not only early-stage startups. Mature companies are being pressured by investors, board directors, public shareholders, and customers to implement AI technologies. In Q1 2023, more companies cited “artificial intelligence” in earnings calls than ever (nearly double the average in 2022).
Not all mature competitors look the same: The definition of incumbent has changed over time. In the late 20th century, an incumbent may have been a non-internet enabled competitor. In the early 21st century, an incumbent may have been a competitor not utilizing the cloud, or not utilizing mobile. Today, an incumbent may be technologically illiterate (this is rare), or a modern public technology company that has scaled in the last decade, or a younger technology company that has raised billions of dollars in private capital. Most early-stage entrepreneurs and VCs don’t give enough credit to the innovation capabilities of the latter two groups. Here’s a simple test – determine the revenue growth rate of your mature competitors in the market. If that number is >20%, you should treat them as extremely formidable adversaries.
As a techno-optimist, I believe technology creates value and generally makes the world a better place. But value is not always dispersed evenly.
Software 3.0 – a new paradigm where code is written through natural language prompts – is poised to deliver meaningful revenue and efficiency gains to both software users and mature software companies. However, the job of the founder, which has never been easy, will become even more difficult. Most technologies have been driving forces for innovation. AI, on the other hand, will often be a driving force for industry consolidation and market share concentration.
What is my edge? Who are my competitors? How easily can someone copy me? These are questions from a bygone era.
Instead, founders should consider these questions in the context of the future. Is my edge defensible in the Software 3.0 paradigm? How dangerous can my competitors become using Software 3.0? Will Software 3.0 give companies the ability to copy me more easily?
Successful founders will be the ones who can answer these questions and navigate these obstacles. If you want to chat about Software 3.0 and defensibility in AI, don’t hesitate to reach out!
Thanks for reading Posts in the Shell by Demren Sinik! Subscribe for free to receive new posts and support my work.