Hero Apps are evolving. Those apps that time, effort and more effort have been poured into over months and years are facing a challenge driven by the relentless march of progress currently spurred by AI advances. The headlines proclaim SaaS is dead, and, perhaps, there is a grain of truth underpinning the hyperbole. Over dinner, a good friend of mine suggested the title for this blog post - and I think its apt.

In the beginning, we kept trying to fit AI into the world we knew. We are perhaps now starting to think of doing different things instead of doing things differently. It reminds me of the stories about the steam engine and how, when the steam engine first showed up, people couldn’t help trying to jam it into the world they already knew. Engineers built these early steam-powered road vehicles that looked and acted a lot like the old coaches with big, boxy bodies on wheels, sometimes even with fake horse-like fronts or tiller steering, all while hissing and clanking along. They spooked the living daylights out of actual horses and got slapped with sky-high tolls from turnpike trusts that wanted to protect the horse-and-carriage business.

Fake Horse and Carriage

Fake Horse and Carriage

Today we’re doing something similar with AI. We’re still trying to cram this powerful new tech into the familiar shape of bloated, always-on hero apps piling on more and more features chasing super-app status. We slap generative interfaces and agents on top of those heavy, monolithic UIs instead of letting go and building truly fluid, intent-driven experiences that pop up when you need them, get the job done, and disappear.

The real breakthrough back then didn’t come from forcing steam into carriage designs, it came when people finally ditched the roads for rails and built systems around what steam actually did best. I think we are starting to get there with AI. What if we don’t need to create apps that live forever? What if apps evolve to be transient, personal and fit for specific purposes? Short-lived and useful to a very small audience at a given moment in time.

Cracks in the Monolith

Most apps today can be considered “hero apps” or even “monoliths” in the sense that once they are built, we keep expanding them with features and fixes and they live on pretty much forever. They solve some key problems, they bring value to a broad swath of users etc. Features get added (or changed) often based on enough user demand and the scope of what they are useful for (or proclaim to be useful for) keeps expanding. A couple of examples come to mind: Uber - started as a ride-sharing app, now offers food delivery and also freight matching too. Airbnb - once a simple short-term home rental platform, has now expanded into a travel super app, encompassing experiences, luxury tiers, celebrity homes and even exploring payments and flights.

Persistent,g monolithic platforms, WeChat-style ecosystems and SaaS giants like Salesforce and Uber, have built empires on centralization with long-term data hoarding and always-on interfaces. AI, however, is starting to challenge this status quo and perhaps a new paradigm is emerging - the anti-hero app.

Disposable by Design

What if, instead of building apps with a view that they live forever and are always expanding, we build apps that are short-lived, transient and solve a single purpose. AI enables you now to build apps fast - sure they are not perfect or solving thousands of different things BUT if I have a problem today that I could use an app to solve, I could tap into AI to help me build that quickly. It moves apps into a world where everyone can augment themselves with exactly what they need, at the right time they need it and also be ok, once the need is solved, letting the app go away because the time invested to build these apps is now so much shorter.

This is super powerful, because it now really enhances personal productivity and ownership in way that hasn’t been done before.

This new app paradigm has 4 key qualities:

  • Ephemeral by design - They are created via prompt, voice, or context (location, calendar event, etc), handle the job, then dissolve. Once the need is fulfilled the app dies.
  • Hyper-personalized - Built on your exact needs, mood, or data snapshot. The more context they have, the more personalized. They fit exactly what you need, when you need it.
  • Agentic and on-demand - AI agents orchestrate across APIs; the “app” is often just a generated UI or workflow that appears and recedes.
  • Disposable economics - Pay-per-use, micro-transactions, or free (vibe-coded in seconds) - no need for licenses or subscriptions.

Technological Foundations

We saw some precursors to the concept of anti-hero apps with custom GPTs and we are seeing this accelerate now with platforms such as Claude Code, Copilot Studio, Github Copilot, Replit, Loveable and others that let anyone describe an app in plain English and watch it materialize. These are all helping accelerate the move from persistent hero apps to transient, intent-driven experiences.

Vibe coding (as much as I don’t love this term, its sticking) lets anyone describe “a quick keto meal planner for this week” and get a functional prototype. Its enabling things that were once out of reach, to be created and created fast! You can try this out immediately with GitHub Spark - describe an app you want and within minutes you can have a front-end deployed. You can then start to tap into Github Copilot to add in back-end features (e.g. payment integration with Stripe, data storage etc.) to build out some key components.

Throw into this mix protocols like MCP (the USB-C for AI Tools) which allows builders to plug quickly into existing tooling, APIs and ecosystems to rapidly connect their apps to key systems and agent to agent frameworks to enhance cross-agent collaboration and you have a rich ecosystem that can enable app creators to connect to key parts of the ecosystem for the pieces they need at any given moment.

You may ask, where do AI Agents fit into all this? AI agents are the core engine and primary manifestation of the anti-hero app evolution, serving as the autonomous “brains” that make all this possible in the first place. They are lightweight, task-specific interfaces that appear contextually, execute, and vanish. Some agents can create UI on the fly as needed to adapt to evolving needs. If we think of agents more tied to each of us where they have a broader context of our likes, interests, needs etc. they can help create the right apps for the relevant challenges you may face.

Not All Heroes Wear Capes

Look, I don’t think hero apps will go away. They will persist through network effects, trust and the need for regulated/complex use cases. Places where social interaction has critical mass, where long term identity is tied and high-touch experiences are important, hero apps will persist. But they will also evolve. We will see more and more AI integration and, as more AI becomes empowering to the end user around app and agent creation, these hero apps will adjust to enable interactions and connectivity into the hyper personal world of apps and agents you own and build.

In the end, the rise of the anti-hero app isn’t about killing the hero, it’s about freeing us from the tyranny of the permanent. There is a vision of the world where software no longer hangs around, clinging like unwanted guests, but arrives precisely when needed, works its magic, and gracefully disappears, leaving only the memory of a task perfectly done. Like a certain wizard perhaps.

Gandalf the Grey

Gandalf the Grey