T M C M

The Next Frontier in AI Engineering is the Idea Itself

Nov 13, 2025 · 4 mins read

For the past few years, our relationship with AI has been a crash course in communication. For a time, the dominant skill was prompt engineering: the art of crafting the perfect instruction, almost like a magical incantation to unlock the desired output from the model.

The industry then shifted to context engineering. It became clear that the model’s performance depended less on the nuance of the instructions and more on the quality of the information provided. The focus moved from how to ask, to what to include. Success became a function of curating the right documents, style guides and data to ground the AI in a specific reality.

But this phase, too, is becoming commoditised. As models grow to understand vastly larger contexts and new techniques allow them to parse entire databases in an instant, the manual work of “context engineering” will become largely automated. When an AI can access and comprehend all available information, providing it with that information ceases to be the critical human skill.

This brings us to a new bottleneck, a new frontier. When the AI has the perfect instructions and access to all the context, what is left for the human to contribute?

The answer is the seed of the idea. The quest.

The Inherent Limit: AI’s Aversion to True Creativity

The next evolution in human-AI interaction is being shaped not by the AI’s capabilities, but by its fundamental limitations. The most significant of these is its difficulty with true, non-derivative creation.

Large Language Models are, by their very nature, masters of the average. They are trained on the vast corpus of human knowledge to predict the most probable, coherent and useful next word. This makes them incredibly powerful for synthesis, summary and execution. But it also anchors them firmly within the bounds of existing patterns.

Even when settings like “temperature” are increased to encourage randomness and variation, the model is not becoming truly inventive. It is simply exploring the outer edges of its standard deviation, remixing known concepts in more unusual ways, not generating a concept from a void.

This is a feature, not a bug. These models have been painstakingly aligned and fine-tuned to produce high-quality, reliable and logical outputs. True creativity, however, is often illogical, chaotic and unpredictable. It requires a tolerance for generating the useless or nonsense, for pursuing dead ends, for producing gibberish on the path to a breakthrough. To allow an LLM to steer into that “non-average-ness” would mean sacrificing the very output quality and reliability that makes it so useful. We can’t have true randomness within systems designed for alignment.

From Engineer to Visionary

This creative limitation is precisely what defines the next human role. As the “how” (the prompt) and the “what” (the context) become increasingly automated, the value shifts entirely to the “why” and the “what if”.

The critical skill is no longer instructing the machine or curating its library, but defining a quest worthy of its immense power. It’s about formulating the initial spark of a novel idea that the AI, by its very design, cannot generate on its own.

  • An AI can synthesise a thousand business plans, but it cannot originate the disruptive insight for a product no one has ever imagined.
  • An AI can write a beautiful story, but it cannot feel the creative restlessness that leads to a completely new genre.
  • An AI can design a system based on known principles, but it cannot ask the foundational and paradigm-shifting question that redefines those principles.

The evolution is clear: we moved from being instructors (prompt engineers) to librarians (context engineers). Now, the role is becoming that of the visionary, the strategist or the originator. The future of leveraging AI won’t be about who is best at talking to the machine, but about who has the most interesting thing to talk to it about. The ultimate value is and will remain the quality and originality of the human idea.