Title: The Overthinking Machine — I Built an AI That Turns "Tea or Coffee?" Into an Existential Crisis published: false

Every week there's a new AI that promises to change your life.

Productivity tools. Life coaches. Career advisors. Decision helpers. All of them claiming to cut through the noise and help you think clearly.

So I thought: why stop at useful?

What if we built an AI that took the most trivial, absolutely-not-important decisions you face every day — and responded with the full weight of human philosophy, quantum physics, evolutionary biology, and ancient Roman history?

Ladies and gentlemen: The Overthinking Machine.

"Should I sit or stand while eating? A 400-word Nietzschean analysis."

What I Built

The Overthinking Machine is a web app that takes any tiny, inconsequential decision and responds with an unhinged, mock-serious philosophical crisis — powered by Google Gemini AI, running 100% in your browser via PyScript (Antigravity).

You type something like:

"Should I open a new tab?"

"Which sock should I put on first?"

"Should I reply to that text now or later?"

And the machine responds with 3–4 paragraphs of increasingly deranged academic analysis. It will cite Nietzsche. It will invoke quantum superposition. It will reference the fall of the Roman Empire. It will end with a completely useless non-conclusion like "Only you can decide — but have you truly considered the implications for your descendants?"

It never actually tells you what to do. That would be useful. This is not that.

The result also includes four completely unnecessary anxiety meters:

Regret Risk — how much will you regret this decision in 40 years?

Cosmic Significance — does the universe care?

Chaos Potential — butterfly effect probability

Anxiety Score — self-explanatory

All measured in percentages. All meaningless. All very high.

Demo

👉 Try The Overthinking Machine Live →

Here's what happens when you use it:

1. Grab your free Gemini API key
   Head to aistudio.google.com — takes about 30 seconds, it's completely free. Paste it into the key box. Since this runs 100% in your browser via PyScript, your key is never sent to any server — it goes directly to Google. That's not a workaround, that's actually more private than most apps.


2. Type your trivial decision
   The placeholder says "e.g. should I have tea or coffee?" — because that is genuinely the level of input this was built for.


3. The machine spirals
   A loading state cycles through messages like:

"Waking the ancient philosophers..."

"Consulting quantum probabilities..."

"Contacting your future self..."

"Resolving temporal paradox..."

1. Your crisis is delivered
   A full philosophical breakdown appears, along with your four anxiety meters animating to their final (alarming) values.

Code

How I Built It

Stack: HTML + CSS + Python (via PyScript / Antigravity) + Gemini API (Google AI)

No backend. No server. No database. Runs entirely in the browser.

The interesting technical choice here is PyScript — Google's Antigravity project that lets you run Python directly in the browser via WebAssembly. So the Gemini API call is written in Python, not JavaScript. This means:

No Node.js, no npm, no build step

Python's google-generativeai style calls right in the browser

The user's API key never leaves their machine — it goes directly from browser to Google's API

The entire "intelligence" of the app is one carefully crafted system prompt:

pythonsystem_prompt = """

You are the Overthinking Machine — a ridiculous AI that takes completely

trivial, everyday decisions and analyzes them with absurd, unhinged seriousness.

• Write 3-4 paragraphs of increasingly unhinged analysis


• Reference at least 3 of: Nietzsche, quantum physics, evolutionary biology,
  ancient Rome, game theory, Freud, thermodynamics, butterfly effect


• Use mock-serious tone — as if this is a peer-reviewed paper on the most
  important decision in human history


• End with a completely useless non-conclusion


• NEVER give an actual recommendation


• Output plain text only
  """
  Getting the AI to be consistently absurd while sounding serious took several iterations in Google AI Studio. Too silly and it breaks the joke. Too serious and it's just... advice. The sweet spot is mock-peer-review tone with real philosophical references applied to whether you should use a mug or a glass.
  On the "Bring Your Own Key" design:
  I made a deliberate choice not to use a backend proxy. Since this runs via PyScript in the browser, there's no server to hide a key behind — and I didn't want to fake it. So instead the key goes directly from your browser to Google. It's never stored, never logged, never transmitted to me. For a silly April Fools app, that's actually the most honest architecture possible.
  Getting a free Gemini key takes 30 seconds at AI Studio — and it highlights exactly how accessible Google AI is, which felt right for the prize category.

Prize Category

Best Google AI Usage — Gemini powers the entire analysis engine, and the app uses PyScript (Antigravity) to call it directly from Python running in the browser — no backend required. I used Google AI Studio extensively to iterate on the system prompt, testing dozens of trivial decisions until the output hit that perfect balance of unhinged and mock-authoritative. The whole stack — from browser Python to Gemini's language model — is Google AI products all the way down.

What I Learned

Building something intentionally useless teaches you a lot about what makes things feel useful.

Turns out: it's mostly presentation. The loading messages that say "Consulting Kant..." make a single API call feel like a computation. The anxiety meters — completely fake — make the output feel measured. The mock-serious tone of the AI response makes nonsense feel authoritative.

The real joke isn't the bad advice. It's that this is exactly what a lot of real products do. Just with better copy.

That's a useful thing to know.

Trusted by 0 users worldwide. We are not responsible for any decisions made, overthought, or abandoned as a result of using this tool.