Devs have pivoted from frameworks to AI models 2026w22

I rarely see much hype around new frameworks these days, instead all of the hype surrounds AI.

We used to spend our weekends arguing about routing, syntax, and the "right" way to structure a controller. Today? Nobody cares about your boilerplate. The only question that matters now is what model you’re running under the hood.

Welcome to Lead Prompt // executing leadership from the root. I’m your host, John Collins.

When I was a full-time software engineer, all of the hype was around new frameworks. It seemed like every week a new framework was coming out, especially in the JavaScript community where the frequency of framework releases almost became a meme. In the PHP community, whole companies and careers grow around the development and commercial support of several famous frameworks. I myself released the Alpha Framework for PHP, but it never garnered any traction in the community.

Perhaps one of the most famous examples of this is David Heinemeier Hansson or "DHH" for short, who launched the wildly-successful Ruby on Rails framework, which arguably became the main reason for people to use Ruby in the first place. DHH was a master at marketing: I remember watching his early videos on Rails when he made everything look so easy, while simultaneously making fun of the PHP language. Framework flame wars were real back then! DHH is now a millionaire, and Ruby on Rails played a big part in building his brand.

For a brief time, micro-frameworks were the hot topic, when framework developers realized that the weight and learning curve of their frameworks needed to be trimmed back. I really enjoyed this micro-framework trend in particular, which led to the micro-services trend that is still quite popular today.

But what happened to frameworks, and why have we stopped caring about them?

Has the hype around frameworks really declined?

Firstly, let's check the online trends to see if this is real or just my false perception.

The shift is incredibly real, and the data backs it up. We can see the collapse of framework hype happening across several major industry benchmarks:

The Stack Overflow Developer Survey: According to the 2024 survey, 76% of developers are now using or planning to use AI tools in their daily development process. The massive cognitive energy that developers used to pour into learning the nuances of the "Rails way" is now being redirected toward AI workflows.
The GitHub Octoverse Report: For years JavaScript, the absolute epicenter of the "framework wars", was the undisputed king of GitHub. But in the 2024 Octoverse report, Python finally overtook JavaScript as the most popular language. GitHub directly attributed this to the AI explosion, noting a 59% surge in GenAI project contributions. Developers are migrating to entirely different domains.
The State of JS Survey: The front-end community used to be defined by its excitement for the next big framework. That excitement is evaporating. Recent State of JS data shows interest in learning or using once-dominant frameworks dropping to historic lows. Newer darlings like Qwik or HTMX aren't capturing the same zeitgeist that React or Vue once did.
The Pivot to Orchestration: Industry analysts at IDC and elsewhere are redefining the developer's core job. The skill set is moving away from "code authorship" (writing boilerplate) and toward "orchestration, judgment, and AI output evaluation."

The framework wars are quiet because the battleground has moved entirely.

Enter AI model hype

Instead of debating micro-frameworks, the community is obsessing over models. Hugging Face, the central hub for the open-source AI community, recently crossed an incredible 3 million hosted models.

But the real frontier, the place where the actual engineering excitement lives, is local. Developers are dropping cloud dependencies and spinning up container stacks on local hardware using tools like Ollama and llama.cpp. Ollama alone has skyrocketed past 180,000 stars on GitHub, making it one of the most popular repositories globally. We are wiring up Qdrant vector databases, connecting memory proxies like Mem0, and running highly capable open-weight models like Qwen, Gemma 4, and DeepSeek V4 Flash right on our own Linux machines.

Software engineers can generate their own frameworks with AI

We originally relied on frameworks to skip the boilerplate. But today, boilerplate is virtually free.

This allows us to return to pure First Principles thinking. Instead of bending your application's architecture to fit the heavy constraints of a third-party package, you can map out the ideal solution on a whiteboard and simply prompt an LLM to generate the exact, bespoke scaffolding you need. You get the rapid development speed of a framework, but you retain absolute control over the system down to the root.

What do we lose with the decline of frameworks?

But we don't just lose the bloat, we lose the guardrails. Frameworks provided community-tested security patches, predictable conventions, and a shared engineering vocabulary. When you joined a Rails or Laravel team, you instantly knew where the models and controllers lived.

If every engineer uses AI to generate their own bespoke micro-architecture, we risk creating isolated, highly opinionated silos of code. The shared "glue" of the developer community could easily fracture into a million AI-generated snowflakes, making onboarding and cross-team collaboration much harder.

What happens next?

The role of the systems architect is moving up the stack. It’s no longer about memorizing a specific framework's API syntax; it’s about mastering high-level system design and orchestration.

The future belongs to those who can design intelligent pipelines: connecting local memory stores, vector databases, and specialized AI agents into cohesive, offline-capable systems. The framework of tomorrow isn't a pre-packaged software download; it's the dynamic AI orchestration stack you build yourself.

But don't neglect frameworks completely, especially the guardrails and standards they provide. I haven't figured out this balance myself with my teams frankly, but I am working on it.

Talk to you next time.