Opinions by an entrepreneur and creative innovator who’s seen a few “deaths” before
Source: Tony Dieste
I’ve spent the better part of three decades as an entrepreneur, building my own companies and helping businesses build their own brands through innovation and creativity. Creativity was at the heart of many innovations, and software was always there right by our side, the silent enabler. Adobe, Figma, Canva, Microsoft, CapCut, ScreenDragon, Oracle, to name just a few. The use and scale of software services and products have been an integral part of our world. So, when I started hearing the drumbeat, “software is dead,” “AI killed the developer,” “SaaSmageddon”, “code is over,” I paid attention. Not because I believed it, but because I’ve learned that when smart people start declaring something dead, there’s usually a real signal buried under the noise.
Here’s my honest take: software isn’t dead. But the way we think about software, who builds it, how it’s built, and what it even means to “build” is undergoing the most fundamental shift since the cloud ate the data center.
The Case for “Dead”
Let’s steelman the argument. The people saying software is dead aren’t entirely wrong. They’re pointing at real phenomena.
First, the marginal cost of producing software is collapsing toward zero. An idea that took a team of five engineers three months in 2020 can now be prototyped by a single person in a weekend using AI-assisted development tools. When the barrier to creating something drops that fast, the value migrates away from the creation and toward the curation, distribution, and integration of what gets created.
Second, AI agents are beginning to replace discrete software applications. You don’t need a separate expense-tracking app if your AI assistant can parse receipts, categorize spending, and file reports by talking to your existing systems. The “app for everything” model starts looking bloated when a general-purpose intelligence layer can orchestrate across all of them.
Third, and this is the one that actually keeps me up at night, we’re watching the interface layer dissolve. The human doesn’t navigate software; the software (via agents) navigates for the human. Software has always been, at its core, an interface between human intent and machine execution. If a user can simply describe what they want in natural language and have it happen, the traditional UI becomes overhead. Not dead, but increasingly optional. I mean, look at the growth of OpenClaw/ClawdBot. I spent the good part of last Saturday trying to figure out (if I should) or how I could, set up a VPN/VPS set-up with OpenClaw running on its own MacMini with its own accounts and credit card limit as my own “assistant/agent”. In the end, I decided not to and wait a little longer and let the community work the bugs out before running my own self-hosting and letting a few first movers get hacked and expose any security risks.
Why I Still Think Software Wins
But here’s where the “software is dead” crowd loses me.
Software is not a product category. It’s a medium. much like paper, film, paint, or electricity: a substrate or channel through which knowledge, processes, intent, and value flow and get expressed. Saying software is dead is like saying “writing is dead” because people stopped sending telegrams. The telegraph died. Writing didn’t. It evolved — into email, into tweets, into prompt engineering. The medium adapts. It’s what is happening in the creative businesses.
What’s actually happening is that software is becoming ambient. It’s dissolving into the infrastructure of everything. The AI models everyone is excited about? They’re software. The orchestration layers connecting agents to APIs? Software. The guardrails, the compliance systems, the data pipelines feeding those models? All software. We’re not building less software. We’re building software that’s increasingly invisible to the end user, and that’s a sign of maturity, not death.
Companies, especially American companies, are also very good at focusing on their core competency, and for that reason alone, is why I believe they will let the domain experts be experts and continue to rely heavily on partners to integrate with their data to find solutions to their problems. This is why I believe that those who are tightly tied to large corporate data sets like Omnicom and Oracle will have data moats and can set up their own orchestration platforms or platforms-as-a-service PaaS models and others will thrive and survive in this new post-SaaS environment. Palantir has successfully used their Foundry and Gotham models to enable customers to integrate, analyze, and operationalize their own large, complex, often siloed datasets, turning raw data into actionable intelligence, decisions, and efficiency gains.
I’d also push back on the idea that AI eliminates the need for software engineers. What it eliminates is the scarcity of code production. But code production was never the hard part. The hard part was and remains understanding a messy, ambiguous human problem well enough to design an elegant system that solves it reliably at scale. Very similar to creativity and true innovation. AI is an extraordinary accelerant for people who can think in systems. For everyone else, it produces plausible-looking software that breaks in production.
What’s Actually Changing
If I had to bet on what the next five years look like, here’s where I’d put my chips:
The product manager and the engineer are merging into a single role. When anyone can generate code, the differentiator is taste, knowing what to build, not just how. The best builders I interact and work with today are hybrids: part strategist, part architect, part designer, with AI doing the heavy lifting on implementation. Think of the young entrepreneurs hauling in the Bitcoin today, they’re creatively using prompts to create and distribute new AI agents to solve specific problems and work for them in the background. At its most basic level, “coding” will be the domain of AI and “developing” (true creativity/innovation/strategy) will be the domain of talented humans.
Distribution beats creation. With the cost of building approaching zero, competitive advantage shifts entirely to go-to-market, brand trust, network effects, and data moats. The next generation of great software companies won’t win because their code is better. They’ll win because their understanding of the problem is better.
Infrastructure becomes the new application layer. The companies building the picks and shovels the model providers, the orchestration platforms, the identity and security layers will capture most of the value. Sound familiar? It should. It’s the same pattern we saw with cloud computing, and with mobile before that.
The Bottom Line
Software isn’t dead, just like creativity isn’t dead. But mediocre software is on life support, and that’s a good thing. The bar for what constitutes a meaningful software product just got dramatically higher, and the moat around “we wrote some code” just got dramatically shallower.
For those of us in the innovation and creative spaces, this is the most exciting moment in a generation. The tools are better, the cycle times are shorter, and the aperture of who can participate in building technology is wider than it’s ever been. But the core discipline understanding humans, designing systems, making hard tradeoffs, that’s not going anywhere.
The software era as we knew it. Maybe that’s ending. But what’s replacing it isn’t “no software.” It’s better software, with talented developers, whatever we call it, built by fewer people, solving harder problems, for more of the world.
And frankly, that sounds a lot like progress.
