In late 2024, Boris Cherny set out to run an experiment. The founder and creator of Claude Code wanted to see just how far AI could go in writing software—not as a coding assistant, but as the sole author. The result was startling: the model now writes 100% of his code. Every line, every file, every feature. This isn’t a glimpse of some distant future. It’s happening now, in the hands of the person who built the tool. Cherny’s interview offers a rare, unfiltered look at where AI-powered development is heading—and why the software industry may never look the same.
The implications extend far beyond productivity hacks. Cherny predicts a fundamental restructuring of what it means to be a software developer. Teams will shrink. Roles will blur. The line between “technical” and “non-technical” will dissolve as building software becomes as accessible as creating a spreadsheet. At the same time, he sees a coming disruption to the SaaS business model that has defined the last decade of tech. The interview touches on everything from multi-agent systems to the cloud-versus-local compute debate, but its core message is simple: the barrier to building software is collapsing, and anyone who isn’t paying attention is already behind.
The Experiment That Changed Everything
Cherny describes Claude Code’s origin as something closer to a research prototype than a product launch. Late 2024. A small team. A simple question: what happens when you remove the human from the coding loop entirely? The experiment exceeded expectations in ways that reshaped Cherny’s own workflow. He no longer writes code himself. The model does it all—taking requirements, generating implementation, debugging, refactoring. His role has shifted entirely to direction and review.
This isn’t theoretical. It’s his daily reality. He runs hundreds of agents through loops using the Claude app on mobile, orchestrating complex development tasks across multiple parallel workflows. The phone becomes a command center for a software factory that never sleeps. The radical shift in his own practice forms the foundation of his predictions for the industry: if this works for one person, it will work for many.
Beyond the Solo Developer
The future Cherny paints isn’t just about individual developers going solo. He envisions teams becoming smaller and more cross-disciplinary. The traditional hierarchy of engineers, designers, and product managers will give way to generalists who span those boundaries. The AI handles the execution; humans provide context, judgment, and direction. This mirrors broader trends in the tech industry toward leaner, more adaptable organizations—but accelerated dramatically by capable AI agents.
The democratization angle is critical. Cherny compares the coming shift to the rise of Microsoft Office. Once, only trained specialists could create professional documents or manage complex spreadsheets. Now, anyone with basic literacy can do both. Software building will follow the same trajectory. The ability to create custom tools, automate workflows, and build applications won’t require a computer science degree. It will be a baseline skill, like word processing is today.
The SaaS Apocalypse and Business Model Shifts
One of the most provocative claims in the interview is Cherny’s take on the “SaaS apocalypse.” If every organization can build its own software more easily, the value proposition of subscription-based tools erodes. Why pay for a generic CRM when you can instruct an AI to build exactly what you need? The incumbent advantage of SaaS vendors—network effects, integration depth, brand reliability—will face pressure from a new wave of custom-built, AI-generated solutions.
This doesn’t mean SaaS disappears. But the market consolidates around tools that are hard to replicate: deeply integrated platforms, industry-specific expertise, and services that require human judgment. The commoditization hits the middle ground hardest—horizontal tools that solve generic problems. Cherny’s view suggests a bifurcation: ultra-specialized, high-value services at one end, and DIY AI solutions at the other.
Internal Tools Versus Public Release
A revealing part of the conversation touches on the gap between what Anthropic builds for itself and what it releases publicly. The company’s internal tools for software development are more advanced than anything available outside. Cherny acknowledges this tension: the public version lags the internal cutting edge, sometimes significantly. The reason is cautious deployment. Racing to ship the most powerful features first means risking failure modes that are still being understood internally.
This gap is worth watching. It signals that the AI coding capability available today is already older than what leading labs are running privately. The public rollout likely follows a staged approach—releasing capabilities once they’re sufficiently understood and hardened. For developers outside Anthropic, the question becomes: how far behind am I working?
Multi-Agent Systems and Parallel Execution
Cherny is bullish on multi-agent architectures. Rather than relying on a single AI model to handle everything sequentially, his setup runs many agents simultaneously, each handling a piece of the problem. They coordinate through loops, sharing context and results. The key insight: parallel execution doesn’t just speed things up. It enables entirely new problem-solving approaches that a single agent can’t achieve.
This connects to his vision of orchestration. The developer becomes a conductor, not a performer. You define what needs to happen, break it into parts, assign agents to each, and synthesize the results. It’s a fundamentally different mental model than writing code line by line. And it points toward development environments where the UI is less about text editors and more about task management and outcome specification.
Cloud Versus Local: The Compute Debate
The interview also explores where AI compute happens. Cloud-based models offer scale and state-of-the-art performance. Local models run on personal hardware, giving privacy and reduces latency. Cherny’s view leans toward a hybrid: cloud for heavy lifting, local for privacy-sensitive or real-time tasks. Neither dominates. The choice depends on use case, not ideology.
What matters more than the compute location is the capability ceiling. As models improve, more workflows shift from cloud to local because local models become “good enough.” The trend runs in both directions simultaneously, and the market is still finding its equilibrium.
What This Means for the Industry
The picture that emerges from Cherny’s interview is one of accelerating transition. AI coding is no longer experimental. It’s productive, personal, and already reshaping how one of its creators works. The broader industry implications—smaller teams, democratized building, disrupted SaaS models, evolving compute norms—are all connected to a single underlying shift: software development is becoming something other than what it has been.
The pace of change means predictions have short half-lives. But the direction is clear. The question for developers, companies, and investors isn’t whether this transformation happens, but how fast, and who adapts fastest.
Key Takeaways
- Claude Code began as a late-2024 experiment; its creator now writes zero percent of his own code personally
- Software development teams will shrink and become more cross-disciplinary as AI handles execution
- Building software will democratize to become a baseline skill, similar to Microsoft Office proficiency
- The SaaS business model faces pressure as AI enables custom-built solutions for generic problems
- Internal Anthropic tools outpace public releases, creating a capability gap that users should understand
- Multi-agent parallelization enables new problem-solving approaches beyond single-model workflows
- Cloud and local compute will coexist based on use case, with the balance still stabilizing