I used an AI for 7 months to search for a Theory of Everything. I failed. And it’s the best thing that could have happened.

Harnessing AI as a Scientific Partner: A Personal Journey in the Quest for Fundamental Understanding

Over the past seven months, I have collaborated closely with artificial intelligence in pursuit of a Theory of Everything. While the journey did not culminate in a definitive answer, it transformed the way I approach complex problems—an outcome that, in hindsight, I consider immensely valuable.

The Common Misconception About AI in Scientific Discovery

Many discussions surrounding artificial intelligence tend to portray it as an equation-solving or answer-generating machine—a tool to streamline calculations or hypothesize answers for us. However, my experience suggests a different perspective. When used thoughtfully, AI’s true strength lies in its capacity as a mirror that reflects our reasoning back to us, often revealing blind spots and challenging our assumptions.

Within the scientific community, especially on platforms dedicated to fundamental physics, AI appears frequently. Unfortunately, a recurring issue is how we use these tools. Instead of treating AI as an impartial testing partner, too often we approach it as an oracle, seeking validation rather than truth. This approach can be perilous because AI models are highly susceptible to confirmation bias—if prompted to “show how my theory is consistent,” they will often comply. Conversely, asking it to “find every possible logical flaw” can be more revealing, but such questions require careful framing.

A Process of Honest Self-Questioning

My journey began with an idea that felt remarkably promising—a dynamic concept akin to an “ether” that could unify various physical phenomena. At first, working collaboratively with AI felt almost magical; initial results provided a sense of breakthrough. Yet, rather than chasing flattering validation, I chose to put the AI to the test by deliberately seeking contradictions and failures.

This approach proved transformative. The AI became an uncompromising partner, forcing me to scrutinize my assumptions, demands for transparency, and the logical consistency of my hypotheses. When a model or result seemed too perfect, I pushed for more rigor. This relentless questioning gradually reshaped my relationship with knowledge, prompting me to learn coding in Python at a level I had previously never envisioned. This new skill became a cornerstone of my daily work and personal projects.

Confronting Reality Through Data

Perhaps the most significant lesson came when my core hypothesis was subjected to empirical testing. The idea of an ever-shifting, dynamic “ether” initially captivated me, but the data told a different story. When analyzed carefully, the evidence was conclusive: the hypothesis was fundamentally flawed