I Stopped Using Frameworks — AI Agents Do It All Now

AI Coding Agents Are Making Frameworks Obsolete

Something big is happening in software development, and a viral Hacker News post with nearly 300 comments just put it into words: software engineering is back. The post, written by veteran developer Alain DiChiappari, argues that AI coding agents are eliminating the need for the bloated framework ecosystem that has dominated web development for the past decade.

His core argument? Frameworks were always a compromise. They existed because writing boilerplate was tedious, hiring was expensive, and nobody wanted to think from first principles. But now that AI agents can generate, refactor, and maintain code autonomously, the entire rationale for adopting someone else's opinionated architecture is collapsing.

This is not a fringe opinion. By early 2026, 57% of companies are running AI agents in production. Gartner predicts that by the end of the year, 40% of enterprise applications will have task-specific AI agents built in. The shift from "developer writes code inside a framework" to "developer orchestrates agents that write code" is happening right now.

Why Frameworks Existed (and Why They Are Losing Ground)

DiChiappari identifies three reasons frameworks became dominant:

1. Simplification: Developers were afraid to design systems themselves, so they adopted someone else's blueprints. This was not real simplification — it was intellectual surrender.
2. Automation: Boilerplate code is boring. ORMs, CRUD generators, and API docs frameworks handled the grunt work nobody wanted to do manually.
3. Labor cost reduction: Companies preferred hiring "React developers" over "software engineers." Frameworks made developers interchangeable cogs.

The automation argument was always the strongest. But in 2026, AI coding agents handle boilerplate faster and more flexibly than any framework ever could. You do not need an ORM when your agent writes type-safe database queries in seconds. You do not need a CSS framework when your agent generates and maintains custom styles across your entire codebase.

As DiChiappari puts it: "I have been basically never writing twice the same line of code. I am instantly building small tools I need, purpose built, exactly shaped around the problem at hand."

The AI Coding Agents Leading the Charge

Not all AI coding tools are created equal. Here is how the major players stack up in February 2026:

What is remarkable is how these tools are converging on a similar pattern: agents that use bash, git, and basic Unix tools rather than framework-specific abstractions. As DiChiappari notes, "Bash was born in 1989. The most mediocre model running at this time knows bash better than any person in the world. The oldest tool turned out to be the most future proof."

For a deeper dive into how Claude Code's agent teams work, check out our complete guide to Claude Code agent teams.

From Autocomplete to Autonomous: The Evolution

The progression has been staggering:

• 2023-2024: Copilot-era autocomplete. Helpful but limited to line-level suggestions.
• Early 2025: Agent mode arrives. Cursor, Claude Code, and Copilot can now understand entire repositories and make multi-file changes.
• Late 2025-2026: Multi-agent orchestration. Steve Yegge's Beads system runs 20-30 parallel agents, producing 12,000 lines of code daily. Cursor agents built their own browser. Claude Code's agent teams let you spin up parallel workers for complex tasks.

The quality picture is more nuanced, however. Google's 2025 DORA Report found that 90% AI adoption correlates with a 9% increase in bug rates, 91% increase in code review time, and 154% increase in PR size. Speed is up, but human oversight remains critical.

What This Means for Developers

The Role Is Shifting from Coder to Architect

The developer's value is moving upstream. Instead of writing every line of code, you are now:

• Designing systems — deciding on architecture, trade-offs, and edge cases
• Orchestrating agents — choosing the right tool, providing context, reviewing output
• Quality assurance — catching the bugs that agents introduce, ensuring coherence across a codebase
• Product thinking — understanding what to build, not just how to build it

This is actually a return to real software engineering. As DiChiappari argues, the framework era turned engineers into operators. AI agents are giving them back the architect role.

Frameworks Will Not Disappear Overnight

Let us be realistic. React, Next.js, Django, and Rails are not going away tomorrow. Millions of applications are built on them, and AI agents are actually quite good at working within existing frameworks. The "popularity paradox" means AI tools reinforce mainstream frameworks because they have the most training data.

But new projects? Greenfield development? That is where the shift is most visible. Developers are increasingly starting with a Makefile and an AI agent rather than npx create-next-app.

The Skills That Matter Now

If you want to stay relevant, focus on:

1. System design and architecture — the thinking that AI cannot replicate yet
2. Prompt engineering and agent orchestration — getting the best output from AI tools
3. Code review and debugging — AI writes fast but introduces subtle bugs
4. Domain expertise — understanding the business problem deeply

If you are comparing AI coding tools, our Claude Code vs Codex CLI comparison breaks down the two leading CLI agents head to head.

The Serenities AI Perspective

At Serenities AI, we have been tracking this shift closely. Our coverage of AI coding tools, from Replit Agent reviews to deep dives on agent teams, is built on the same philosophy DiChiappari describes: use AI as a force multiplier, not a crutch.

The developers who thrive in this new era will be the ones who understand both the capabilities and the limitations of AI agents. They will use tools like Claude Code, Cursor, and Devin not to avoid thinking, but to spend more time on the thinking that actually matters.

Frequently Asked Questions

Are AI coding agents actually replacing frameworks in 2026?

Not entirely, but the trend is clear. AI agents are eliminating the need for framework boilerplate by generating custom code on demand. New projects increasingly start with agents instead of frameworks, while existing codebases still rely on traditional frameworks. The shift is most visible in greenfield development.

Which AI coding agent is best for replacing framework boilerplate?

Claude Code and Cursor are currently the strongest options. Claude Code excels at full-stack development with its agent teams and 1M context window, while Cursor's workspace context makes it ideal for multi-file refactoring. For fully autonomous development, Devin handles end-to-end task execution.

Will developers lose their jobs to AI coding agents?

The role is changing, not disappearing. Developers are shifting from writing code to architecting systems and orchestrating AI agents. Companies still need humans for system design, quality assurance, and product thinking. The demand is moving from "React developers" back to "software engineers" who can think from first principles.

What skills should developers learn for the AI agent era?

Focus on system design, prompt engineering, agent orchestration, and deep domain expertise. Understanding how to review AI-generated code and catch subtle bugs is becoming more valuable than writing code from scratch. The best developers in 2026 are architects who happen to use AI as their primary tool.

Is vibe coding the same as using AI coding agents?

Not exactly. Vibe coding typically refers to casual, prompt-driven development where you describe what you want and let AI build it. AI coding agents are more sophisticated — they understand your entire codebase, maintain context across sessions, and can execute multi-step tasks autonomously. Think of vibe coding as the entry point and agent-driven development as the professional evolution.