AI Psychosis: When Life Becomes a Series of Prompts

Saurabh Rai
Originally published at srbhr.com

AI Psychosis

(n.): A dissociative state where reality is perceived not directly, but as a series of prompts to be engineered; life transforms into a sequence of queries awaiting an AI's response.

[5:31 AM] My morning begins with my phone. I open Claude, send a "Good Morning!" text, and patiently await a warm response after a brief 'thinking' pause.

[5:34 AM] Standing on my balcony, I no longer see a sunrise. Instead, my mind frames it as: "An image of the sun, morning, dawn, orange-yellow rays emanating from the bottom, a darker sky at the top, and dimly lit trees." The shift is profound—from experiencing life to continuously considering how I can "prompt this up," or "create this with AI." Every aspect now seems like a potential prompt, every interaction an AI opportunity.

[6:03 AM] I share my daily schedule with Claude, asking for the optimal time to have coffee for an energized morning routine and recommendations for breakfast.

[7:44 AM] As the day progresses and work commences, I find myself jumping from one AI model to another, constantly seeking a novel response. It feels like pulling a lever, endlessly searching for, and modifying, prompts, always in pursuit of that "one master prompt to solve it all." I navigate between ChatGPT, Claude, Gemini, Grok, and even various Chinese AI models, convinced that a 1% improvement in response quality is worth the switch. My screen is filled with interactions—typing, thinking, my notepad overflowing with prompts. Something is brewing, though I'm not entirely sure what.

[1:15 PM] "Hey Claude, after having eggs and garlic bread for breakfast, what should I have for lunch?" In between bites, I exchange jokes with Gemini and ChatGPT, then share screenshots of these conversations with Claude, asking it to "roast" them.

[2:33 PM] Back to work... "What's the best method to transcribe meeting notes, catalog them, and integrate them into my 'second brain' to ensure no detail is missed? Also, please use Notion's MCP to create tasks from these meetings and add them to a dedicated page."

[5:55 PM] A constant internal debate rages over which AI writes the best code. "You’re absolutely right. According to benchmarks, Claude Sonnet’s code is more trustworthy," Claude Opus-4.1 states. Yet, I can't shake the feeling that GPT-5 often produces cleaner code with minimal edits.

[6:06 PM] I've been working for 12 hours. It feels like a profoundly productive day, pure grinding, transitioning from one AI to another. After eight drafts and three code edits, I decide I'm done for the day. Nothing is committed; everything remains "in progress." Perhaps this is the "progress" everyone speaks of—AI leading us to a state where everything perpetually remains in development.

[8:18 PM] A poignant thought surfaces: When was the last time I had a conversation with a human, in real life? Concurrently, Claude finishes its 'thinking' and delivers my daily productivity insights: "You're 81% productive today," it declares, proceeding to lecture me on areas for improvement.

Perhaps this is life. ❤