Déjà Vu: AI Agent Frameworks and the JavaScript Gold Rush Revisited

Having lived through the era of a "new JavaScript framework every week," we now find ourselves in the gold rush of the AI agent framework space.

New frameworks appear daily, each claiming to be the 'ultimate' solution for building AI agents, often backed by YouTubers enthusiastically promoting demoware and usually their own library, framework, or SaaS offering. Unfortunately, this enthusiasm can lead companies to uncritically adopt these tools without considering the long-term implications.

The recent AutoGen situation is a prime example of the chaos. Created by Microsoft, one might have expected stability, yet the framework has splintered, with founding contributors releasing their own fork, leaving developers confused and exhausted.

Similarly, LangChain presents numerous overlapping solutions, and occasional breaking changes can make it more of a hindrance than a help, as they require constant learning and frequent updates.

The lesson from the JS framework wars is clear: focus on understanding the core principles rather than tightly coupling your solution to any single framework.

The biggest takeaway we learnt back then was to think in terms of components and composition, which I covered in my "Build Components Not Walls" talk.

Some AI offerings will undoubtedly emerge as long-term winners, just as React did, but many will fade as early hype and quick profits give way to stability and sustainability.

How to build solutions during the age of the AI gold rush?

Adopt a "do one thing, and do it well" approach with framework-agnostic functions.

This will enable you to adapt, experiment, compose, and combine agentic components as the landscape shifts. Having such abstractions allows you to future-proof your approach. By breaking down elements, such as prompts and schemas, you can more easily adapt if a framework falls out of favour or is abandoned.

However, balance is crucial. Abstract too aggressively, and you may lose the genuine benefits that a framework provides.

In my upcoming posts, I'll share examples of how I've applied these principles.

For now, the key takeaway is simple:

Focus on core concepts over fleeting trends.