The Case for Software that Doesn’t Follow the ‘Rules’
The exponential potential of unpredictable software.?
Why would anyone want a product that doesn’t work the same way every time? At first, it might sound counterintuitive—reliability has always been a cornerstone of good products. But as generative AI reshapes what’s possible, the concept of a “predictable” product may start to feel outdated. Rather than rigid tools limited to pre-set functions, we’re beginning to see the potential of adaptable, flexible software that learns, responds, and even surprises us.
Today’s newsletter is a bit different from our usual actionable insights and AI guidance. This time, I wanted to dive into a more conceptual, forward-looking topic on the evolving role of software and generative AI. If this shift in focus resonates with you—or if you’d prefer to keep the content focused on practical takeaways—please let me know. I’m considering whether to introduce a separate newsletter with content like this, or perhaps include it as a feature once a month. Your feedback will help shape where we take this!
Software products are inherently limited. For anyone who hasn’t worked in the software space—or even those who have—it can be easy to miss just how restricted these products are. Every feature you use is shaped by budget constraints, the ideas of the people in the room, and the need to reach as broad an audience as possible. There’s a formula to all of this, with decisions determined by the number of users, costs, and anticipated returns. The result? Products that confine you, the user, within a rigid box of pre-defined features. Sure, some software offers customisation options, but anyone who’s worked through complex ERP or CRM setups knows that this can often make things even more frustrating.
Generative AI, however, is challenging this whole approach, dismantling the notion of a product as something fixed, predictable, and finite. The traditional mindset says people want reliability, but I’d argue that this assumption is already becoming outdated. We’re beginning to value tools that aren’t boxed in by predictable rules but that adapt to our needs—even if they’re sometimes a little messy.
Breaking Out of the Box
Generative AI is fundamentally changing what products can be. Think about how text generation alone has evolved: AI can now generate anything from coherent essays to complex code, and even functioning software with a single prompt, with the potential to assist, automate, or even inspire. The possibilities are practically limitless. And this capability is going to go far beyond text. Products will soon do more than perform fixed functions; they will help organise your thoughts, filter your context, and make data connections that unlock an almost boundless variety of outputs. No longer limited to “writing a report” or “generating code,” text becomes a way to create actions, instructions, even solutions to scientific problems. In short, generative AI takes text beyond its traditional limits.
With AI, we’re moving away from products that perform a static set of actions toward ones that create genuinely unique responses in the form of text, images, video, code, and instructions. Consider an accounting tool that goes beyond crunching numbers. Instead of offering a predictable list of common accounting issues to “check,” imagine it could analyse your specific context, evaluate unusual trends, and give unique insights into why certain numbers might look the way they do. The tool isn’t offering canned responses; it’s giving you something fresh, tailored to your specific needs. Historically, software couldn’t operate this way—predictability was built into its DNA. But generative AI is shifting that paradigm by unlocking creative and personalised potential within each product.
This same potential is exemplified in tools like Claude, with its “artifacts” feature. Imagine being able to create entirely new products within the product itself, a tool that can generate custom code on the spot to expand its own capabilities, a tool that each time you give it more or better information, does something totally unique to your context. This isn’t about adding more options to a menu—it’s about creating functionality that never existed before, all in response to your unique requirements. Generative AI allows software to be something more fluid and flexible, shaped around you and your goals in real time.
Unbounded Potential in a Traditionally Bounded World
So, what does this mean for products and how we use them? Consider the possibilities of a product that isn’t restricted by the intentions of designers, developers, or investors. These products will perform traditional tasks in new ways, responding to your requests with dynamic, personalised output. With tools like Claude that can develop features on the fly, or platforms like Replit where agents generate code in real time, products are finally becoming tools that meet users where they are—not the other way around.
This flexibility creates a new category of software that isn’t confined to what it was initially programmed to do. Products like an accounting tool that not only calculates but also analyses, anticipates, and recommends new strategies based on your real-time data. Tools that don’t just respond to pre-set inputs but actively adapt to each unique situation. This is what it means for a product to be “unbounded”—not limited to a menu of features but capable of evolving with its user.
The Economics of Mass Market Conformity
But this shift isn’t just about what’s technologically possible; it’s about what’s financially viable. Historically, software has been funded and shaped by venture capital. This system demands products that appeal to the largest possible market, often sacrificing specific features or depth for the sake of simplicity. Products designed for mass appeal are optimised for reliability and accessibility but often end up offering a diluted experience. Companies strip away niche features and remove personalisation options to reach the widest possible audience.
Generative AI changes this dynamic. Because AI reduces the cost of creating and running software, small teams—or even individuals—can now build and maintain highly customised tools without the need for major funding. It used to be impossible to justify creating niche products, but AI-driven efficiencies mean developers can now focus on specific user needs rather than diluting the experience for mass-market appeal. Niche markets are suddenly viable, and we’re seeing products emerge that are finely tuned to the unique requirements of specialised users.
Personally, I’ve seen these shifts firsthand. With generative AI, I’m able to accomplish tasks that used to require a team. In a given week, I’m effectively covering the roles of multiple full-time people, with my workload reduced by about 80 hours. Recently, a task that should have taken 35 hours took me less than four. This isn’t just increased productivity—it’s a demonstration of how AI opens doors for small teams to accomplish what once demanded significant capital, skills and manpower.
Moving Beyond Deterministic Software
Most people don’t realise how restricted they’ve been by software designed to cater to a broad user base. Products built to attract venture capital need to achieve sky-high valuations, meaning they must appeal to enormous user bases. Consequently, they deliver diluted, standardised experiences that rarely meet specific needs. Consider how even large-scale tools like Microsoft Word now integrate generative AI with limited functionality—an attempt to make complex AI features accessible to everyone. While these integrations are convenient, they lack the flexibility and potential of standalone AI tools.
While generative AI’s flexibility opens up exciting possibilities, there are still many scenarios where software needs to remain predictable. In fields where precision and consistency are paramount—such as accounting, compliance, or medical diagnostics—a fixed and predictable workflow is essential. Unfortunately, some companies are pushing generative AI features into products where unpredictability can actively hinder the user experience. This isn’t an argument to replace reliable, deterministic features with AI tools that “sometimes work”; rather, it’s an opportunity to think beyond traditional product boundaries and understand where AI can truly add value. Generative AI’s strength lies in areas that benefit from exploration and unique insights—not in replacing features that already work effectively with fixed, rule-based outcomes.
Generative AI, however, enables us to break free from this determinism. Instead of relying on software that was built to perform a rigid set of actions for the masses, we can start using products that adapt to our unique needs. With fewer companies relying on venture capital, creators are now freer to build specialised products focused on specific problems rather than creating generic tools that fit everyone’s use cases but serve no one perfectly.
The Shift Toward Adaptable, Unique Software
In the end, this shift isn’t just about making products more powerful or adaptable—it’s about redefining what software can be. Removing rigid determinism from software opens new possibilities, prompting us to rethink what we expect from the tools we use. When AI-driven software can evaluate our context, respond intelligently, and generate unique responses, we’re no longer confined to a predictable set of options. We’re opening the door to software that is constantly evolving, designed not only to function but to function differently and uniquely for each user.
As AI continues to develop, users may come to see dynamic, contextually aware software as more valuable than predictable software. We’re moving toward a world where the best products aren’t those that simply offer pre-set functionality but are those that adapt, evolve, and unlock new opportunities. This change isn’t just about more powerful software; it’s about a fundamental shift in how we think about tools. The days of rigid, one-size-fits-all products may soon give way to a world where software can be a dynamic, evolving partner in our work and lives.
Did you enjoy this? Let me know with a like, comment, share, or a DM. It helps me know what content is helpful or enjoyable, and whether it's worth the time to polish these ideas for public posting.