Two of tech’s sharpest thinkers published what look like opposite conclusions within weeks of each other.

Steven Sinofsky, legendary product leader and a16z board partner, argued we need vastly more software—that demand is expanding, not contracting. Peter Steinberger, creator of OpenClaw, told YC that 80% of apps will disappear as AI agents take over.

One says software is booming. The other says apps are dying.

They’re both right. They’re describing different layers of the same shift: more software than ever, but packaged as agents—not dashboards.

We Need Vastly More Software, Not Less

Sinofsky’s argument, laid out in his essay “Death of Software. Nah.”, is historical in its framing. Every time a major technology shift arrives, pundits declare that the previous paradigm is over. The PC didn’t kill mainframes. Amazon didn’t kill Walmart. Streaming didn’t kill media production—it multiplied it.

The pattern holds: new platforms expand the market for software. They don’t shrink it.

His core argument is that the vast majority of work that could be automated by software hasn’t been yet. Most businesses still run on spreadsheets, email chains, and manual processes. AI doesn’t reduce the need for software—it makes software accessible to domains that couldn’t afford or justify it before.

Software moves up the stack. Tasks that once required a team of engineers become features. Features become commodities. And the newly freed capacity gets redirected toward harder, more valuable problems.

As Sinofsky puts it: “This is the most exciting time for business and technology, ever.”

If this sounds familiar, it should. It’s the same dynamic we explored in our analysis of the Jevons Paradox in marketing AI—when the cost of something drops, total consumption increases. Cheaper software means more software, not less.

80% of Apps Will Disappear

Steinberger’s claim, made during his YC interview on OpenClaw, sounds like it flatly contradicts Sinofsky. But listen carefully and he’s making a different argument entirely.

He’s not saying we’ll need less software. He’s saying we’ll need fewer apps—the kind you open in a browser, navigate through menus, and interact with through forms and dashboards.

OpenClaw illustrates the shift. It’s a local-first AI agent framework that exploded to 190K+ GitHub stars practically overnight. The project isn’t another SaaS dashboard. It’s an agent that lives on your machine, works with your files, and coordinates with other agents autonomously.

Steinberger describes the experience as having a personal AI that lives on your machine and just handles things for you. Apps become capabilities, not destinations. You don’t open your project management tool—your agent handles project management as part of a broader workflow.

The implications run deeper than UX convenience. In the OpenClaw model, your data stays on your machine as markdown files. Agents communicate with other agents directly—bot-to-bot interactions that never surface in a human-facing UI. The architecture favors swarm intelligence over monolithic AI: many small, specialized agents coordinating rather than one massive system trying to do everything.

This is the end of the app-as-destination paradigm. Not because people don’t need the functionality—they need more of it than ever—but because the container is changing.

Different Layers, Same Shift

The apparent contradiction dissolves once you separate the demand curve from the form factor.

Sinofsky is right about demand. More work will be done by software. More domains will be automated. More problems will be solved computationally. The total volume of software running in the world is increasing on every axis.

Steinberger is right about form factor. The way humans interact with that software is fundamentally changing. The app—a destination you visit, a dashboard you navigate, a UI you learn—is giving way to agents that work in the background, surface results conversationally, and coordinate with other agents on your behalf.

	Sinofsky	Steinberger	Both agree
Software growing?	Yes, massively	Yes, but differently	Total software increases
Apps dying?	No (historical analogy)	Yes, 80%	Apps as destinations die; as capabilities, they grow
Who builds?	Domain experts + devs	Agents self-configure	More builders, more autonomy
Where value lives	Higher up the stack	On the user’s machine	Distributed, not centralized in dashboards

The synthesis: we’re entering a period of more software but fewer apps. The software layer expands—Sinofsky is right—but it increasingly runs as agents, not applications—Steinberger is right.

Neither source makes this combined argument. But it’s the logical conclusion of both.

What This Means for Marketing Software

Marketing is a textbook case of both dynamics playing out simultaneously.

The demand for marketing software is expanding. More channels to manage. More personalization required. More always-on, real-time responsiveness expected. The MarTech landscape grows every year because there’s genuinely more work to do—not because vendors are inventing problems.

But the form factor is shifting. The dynamic agent layer that a16z describes in their Big Ideas series isn’t a better dashboard—it’s software that acts on your behalf without requiring you to open it. The workflow ownership pattern we’ve analyzed in companies like Harvey and Sierra shows the same trajectory: from tools that assist to agents that execute.

The marketer of 2027 won’t open eight SaaS tabs every morning. They’ll have an agent in Slack or WhatsApp that monitors their brand, identifies opportunities, executes campaigns, and reports back—all without a dashboard login.

Steinberger’s “bot-to-bot” vision maps directly onto the Agent-to-Agent (A2A) protocol already gaining traction across the industry. Marketing agents won’t just talk to marketers—they’ll negotiate with ad platforms, coordinate with sales agents, and exchange data with analytics agents. The human approves strategy. The agents handle execution.

This is software eating marketing labor—not by building better tools for marketers, but by building agents that do the marketing.

The Form Factor Is the Strategy

Here’s the implication most marketing AI companies are missing.

If you accept both Sinofsky and Steinberger’s theses, then the biggest competitive risk in MarTech isn’t feature gaps or model quality. It’s building the wrong form factor.

Most marketing AI companies are building better dashboards. They’re playing in Sinofsky’s world—more software, more features, more capabilities—but packaging it all in Steinberger’s dying paradigm: the app.

The Steinberger insight is that the form itself is the competitive advantage. An AI agent that lives in your communication tools, works with your files, coordinates with other agents, and never asks you to “log in to your dashboard” isn’t just a better product. It’s a different category.

Marketing software that doesn’t look like software. That’s the form factor shift both sources point to, from opposite directions.

The most exciting time for software isn’t about more dashboards. It’s about software that finally gets out of the way.

Lane is an AI CMO that handles your marketing as an agent—not an app. No dashboards, no learning curve, just results. Start free →

References

Steven Sinofsky — “Death of Software. Nah.” (a16z, Feb 2026)
Peter Steinberger — YC Interview on OpenClaw (Y Combinator, Feb 2026)
OpenClaw — Open-source local-first AI agent framework
Jevons Paradox in Marketing AI — Luminary Lane Blog
The Dynamic Agent Layer — Luminary Lane Blog
Workflow Ownership: Harvey, Sierra, and the AI CMO Pattern — Luminary Lane Blog
Software Is Eating Marketing Labor — Luminary Lane Blog
A2A Marketing: When Your Next Customer Is an AI Agent — Luminary Lane Blog