ยท 3 min read

๐ŸŽน The Piano Inside the Torpedo

In 1941, Hedy Lamarr patented an unjammable torpedo. The Navy told her to sell war bonds instead. The insight was right. The wrapper was wrong. Most breakthroughs die this way โ€” and AI video has the same problem today.

๐ŸŽน The Piano Inside the Torpedo

In 1941, Hollywood's biggest star and an avant-garde pianist walked into a patent office with an idea that would take sixty years to change the world.

Hedy Lamarr had fled her first husband, an Austrian arms dealer who'd dragged her to Nazi weapons meetings. She understood munitions. George Antheil had spent years trying to synchronize sixteen player pianos for a single composition. He understood coordination.

Together, they designed a torpedo guidance system that hopped between 88 radio frequencies โ€” one for each piano key โ€” synchronized by matching perforated paper rolls inside the ship and the weapon. Patent No. 2,292,387: a "Secret Communication System."

The Navy's response: We think you're trying to put a piano inside a torpedo. They told Lamarr to go sell war bonds.

The concept was sound. The mechanism looked ridiculous. And the people evaluating it could not separate one from the other. Frequency-hopping spread spectrum, as it's now called, underpins WiFi, Bluetooth, and GPS. Lamarr never saw a dime.

Most breakthrough capabilities don't die because the concept fails. They die because the packaging doesn't match what the buyer expects to see.

Generative AI video is stuck in the same gap. Runway, Sora, and Kling can produce cinematic-quality motion in seconds. But the delivery mechanism for professionals โ€” type a prompt, hope for the best โ€” doesn't meet the bar. The capability is real. The wrapper around it isn't.

0:00
/0:11

Today's startup idea: a curated AI video marketplace that turns generative chaos into a reliable production tool. Structured shot specs in, vetted synthetic b-roll out. Getty meets Envato, rebuilt AI-native.

The wedge is going after shots that don't exist on any stock library: a macrophage engulfing bacteria, data flowing through a CPU cache, a yield curve inverting in real time. Technical b-roll too niche to film and too expensive to animate traditionally.

A focused two-person team running one vertical can hit $35K+ in monthly recurring revenue within nine months โ€” with three revenue layers stacked so that a single delivered clip generates income at the point of creation, again as a library asset, and again as a regeneration template.

Read the full playbook here:

Stock libraries can't keep up with demand for technical explainer footage. A new AI-native model turns structured video bounties into a regeneratable clip library with compounding subscription revenue.

Full Playbook

From the Vault:

The technology scouting software market is headed toward $600M+ but every tool tracks companies after they raise. This AI-powered micro-SaaS idea tracks what builders are shipping now and sells the gaps to founders and VCs.

Full Playbook

Billion-dollar sleep apps sell audio content. A startup idea in wearable health tech that reads your physiology, adapts sound in real time, and proves it downshifted your nervous system could own the outcome layer they ignore.

Full Playbook

Read next

๐Ÿ“บ Netflix Built the Wrong Button

๐Ÿ“บ Netflix Built the Wrong Button

Netflix spent four years building a button to solve decision fatigue. In 2023, they killed it. The problem was real โ€” the solution was wrong. A button says "surprise me." A channel says "sit down, we've got you." The startup opportunity hiding in that gap is worth stealing.

Startup Heist | Briefings
Startup Heist | Briefings
ยท 3 min read
๐Ÿ’ Chaos Monkey for AI

๐Ÿ’ Chaos Monkey for AI

Netflix didn't write a postmortem. They built a program that killed their own servers every day on purpose. They called it Chaos Monkey. Now thousands of companies ship AI into real workflows with zero stress testing. That gap is a startup idea worth $300Kโ€“$650K for a solo founder.

Startup Heist | Briefings
Startup Heist | Briefings
ยท 3 min read
๐Ÿชž ELIZA

๐Ÿชž ELIZA

In 1966, a secretary watched a professor build the world's first chatbot from scratch. She knew it was a trick. Then she asked him to leave the room so she could talk to it alone. What she revealed about human nature is now a billion-dollar blind spot in AI.

Startup Heist | Briefings
Startup Heist | Briefings
ยท 3 min read
New startup opportunities, ideas and insights right in your inbox.