Malleable software is the next product category
Samsung’s smart fridge started showing ads in its users’ kitchens. Google was found legally guilty of intentionally degrading its own search engine. 76% of the websites analyzed by the FTC and its international counterparts use at least one dark pattern: designs deliberately built to manipulate the user. 56% of people have lost trust in a platform for exactly that reason.
Cory Doctorow has a term for this cycle: enshittification. The American Dialect Society chose it as word of the year in 2023. Australia’s Macquarie Dictionary did the same in 2024. In 2025, Doctorow published an entire book about it. The process always follows the same pattern: first a platform delivers real value, then it charges for it, then it extracts it. What’s notable isn’t that it’s happening. It’s that it can no longer be ignored.
For most users, this is just a growing frustration. For a builder with judgment, it’s something else: a market signal.
What Proton learned before everyone else
In 2014, Andy Yen was a physicist at CERN. Edward Snowden’s revelations about the NSA’s mass surveillance had been in the air for a year. Yen and two colleagues launched a crowdfunding campaign for an email service the provider couldn’t read. They raised $550,000 from more than 10,000 people.
The lesson wasn’t that there was demand for an email service. It was that there was demand for a philosophy. Privacy as a real promise: not a feature on a marketing bullet list, but the architectural principle of everything they built.
Today Proton has more than 100 million accounts, estimated revenue of over $500 million a year, and a valuation exceeding a billion dollars. Without traditional venture capital. Without monetizing data. Without advertising. Just subscriptions from users who trust that the product delivers what it promises.
Proton didn’t win because it made better email than Gmail. It won because it built a different category: software that doesn’t betray you. Every new product it launched extended that same promise: VPN, storage, calendar, passwords. The moat isn’t the features: it’s the philosophical coherence of the ecosystem.
The same bet, another dimension
There’s a direction nobody is building seriously yet, and it operates on the same logic as Proton.
It’s called malleable software. The idea is simple in concept and hard in execution: instead of making software every user uses in exactly the same way, make software each user can shape to their own needs. Not sealed apps. Living tools.
In June 2025, the research lab Ink & Switch published the year’s most influential essay on the topic. Its central argument: today’s apps are “avocado slicers,” designed for a use case so specific that they end up mediocre for everyone else. The doctor who can’t delete irrelevant fields from their records system. The accountant who needs a calculation no spreadsheet includes. The user with exactly one need nobody programmed for.
If you want more context on the idea and how LLMs changed the equation for the end user, we explore that angle here. For the builder, the question is different: not what it feels like to use it, but what it means to build it.
The moat nobody is seeing
A malleable product isn’t just more useful to the user. It’s harder to abandon.
When someone uses a generic product, the switching cost is low: find an alternative, learn the new interface, done. The case changes when a user has configured, extended, or built something on top of your product. When the tool already carries their history, their rules, their logic, the cost of leaving isn’t learning another product. It’s losing the product they built for themselves.
That’s a real moat. Not one of artificial lock-in or dark patterns. One of value genuinely created by the user inside your platform.
Mikhail Dubakov, founder of Fibery, puts it bluntly: by 2030, malleable AI platforms will make rigid vertical SaaS feel like a relic. Not because rigid software is bad per se. Malleable software can be everything rigid software is, and adapt on top of that.
The analogy is close for those who use Quantis: an exchange-rate platform can offer precise, up-to-date data as its core value, and let each user build on top of that the view they need: the calculator for their specific case, the alert for the threshold that matters to them, the chart for the period they monitor. Not because the team programmed each variant, but because the user described them and the product allowed them.
There’s no map. That’s good.
The malleable software category has no established playbook. There’s serious academic research, there are products that come close without naming it that way, and there’s an enormous space between what exists and what could exist.
For the builder looking for real opportunities, that blank map isn’t a warning. It’s exactly the signal that it’s worth exploring.
Proton didn’t invent privacy. Ink & Switch didn’t invent the idea of moldable software. The product opportunity doesn’t require inventing the concept. It requires executing it with coherence, with philosophy, and with enough patience for users to understand that this is different.
Tech is rotting in ways that no longer have a way back. And when an industry rots, something always replaces it. The question isn’t whether it’s going to happen. It’s who is going to build it.