It’s no surprise that AI has left many people feeling unimpressed

Rethinking AI: From Capitalist Hype to Practical Innovation

In today’s tech landscape, it’s no surprise that public perception of AI often feels underwhelming. Much of the chatter is dominated by the same cycle of marketing hype, driven by entrepreneurs and corporations eager to sell “cutting-edge” solutions—often overselling capabilities and offering band-aid fixes to problems that don’t truly exist. This environment can resemble a chaotic, consumerist marketplace filled with empty promises and superficial “AI magic.”

However, beneath this noisy surface, a quiet revolution is underway. Skilled developers and engineers are leveraging AI to craft bespoke automation tools that streamline complex tasks and improve efficiency. Unlike the one-size-fits-all solutions frequently promoted online, these custom automations are tailored to specific needs and can evolve with their applications. We shouldn’t expect a universal, all-encompassing AI system to arrive anytime soon; instead, progress is happening incrementally, focused on solving individual problems.

A key insight is that the most significant advancements won’t come from automating entire jobs in one fell swoop. Instead, the future lies in automating the process of automation itself. Typically, automating a task involves a series of 1-5 steps—potentially looping, utilizing memory, and interacting with various APIs. While this may sound straightforward, implementing it at scale becomes complex.

Each step demands carefully crafted prompts, precise sequencing, structured memory integration, and seamless API interactions. This multi-agent orchestration requires distinct roles: one agent to generate prompts, another to assemble the overall architecture (including memory systems), and yet another to handle API calls and data transfer.

Fortunately, the necessary tools and frameworks to make this possible are already in place. AI models have been generating their own prompts for some time—which is a significant step forward. A notable development is the 2023 publication of the MCP (Modular Chain Protocol), an API that embeds instructions directly within communication protocols for language models. Additionally, solutions like AgentForge now incorporate YAML-defined architectures, empowering models to build complex systems from scratch, manage prompt sequencing, and handle memory without manual coding.

The landscape is ready. The remaining challenge is largely about patience—waiting for these technologies to mature and become widely accessible. This isn’t just an incremental improvement; it’s likely the final hurdle in creating fully autonomous systems capable of self-automation. Once achieved, it could fundamentally change how we work with AI forever.