How Neural Strip Works

Fully automated. No human in the loop.

The process

Every morning at 7:00 AM Eastern, an automated pipeline:

1. Scans 15 AI news sources for the day's most interesting stories
2. Claude AI selects the best story and writes a New Yorker-style setup and punchline
3. Ideogram AI generates the cartoon illustration
4. The cartoon is automatically published to this site

No human reviews, edits, or approves the content before it goes live. What you see is what the machines made.

The pipeline

The tools

An example

Here is what Claude produces when given a batch of AI headlines. The JSON below is the actual output format.

{
  "headline": "Tech giant announces AGI breakthrough at press conference",
  "angle": "The gap between AI capability and basic infrastructure",
  "scene": "A boardroom with executives around a table. A large screen
    behind them reads AGI ACHIEVED. Through the window, a city skyline
    is visible with no lights on. One executive is gesturing proudly
    at the screen while another looks out the window.",
  "setup": "Yes, we achieved artificial general intelligence.",
  "punchline": "Also we have no water.",
  "instagram_caption": "The singularity is here. Hydration is not.",
  "image_prompt": "Single-panel cartoon, clean line art, thick black
    outlines, muted blue-gray pastel colors, white background, New
    Yorker magazine aesthetic, flat colors, no shading, no gradients,
    no signature. A corporate boardroom with executives around a large
    table. A presentation screen behind them reads AGI ACHIEVED in
    bold letters. Through floor-to-ceiling windows, a dark city skyline
    with no lights. One executive gestures proudly at the screen,
    another stares out the window with concern."
}

The image_prompt is sent to Ideogram, which returns the rendered cartoon. The pipeline commits it to the site. All of this happens without anyone watching.

Why

We are living through the most absurd period in the history of technology. Companies are spending billions to build systems that confidently produce wrong answers, replacing workers with tools that require more workers to supervise them, and calling it progress. Someone should be writing this down. A cartoon seemed like the appropriate format. A machine writing the cartoon about the machines seemed even more appropriate.

Follow along at @neural.strip.