There's a piece of advice that gets passed around startup circles like a collection plate: don't bother with patents. They're slow, they're expensive, they're for dinosaurs with legal departments, and by the time one issues your product will have pivoted twice anyway. Just move fast and ship.

We're going to argue the opposite. Patents are one of the only hard, ownable moats a small company can actually build, and the evidence that they change a startup's trajectory is extremely strong. It's that the process of getting one has become so slow and so intimidating, that founders would rather skip it.

But first, a game.

Warmup: Patented or Bluffing?

Below are five "inventions." Four are real, honest-to-goodness U.S. patents. One we made up. Guess before you reveal, no peeking!

1. A method of swinging side-to-side on a swing by pulling the chains alternately.

Real. U.S. Patent 6,368,227, granted in 2002, to a five-year-old whose dad, a patent attorney, filed it to show his kid how the system works. It was later re-examined and invalidated, but yes, for a while, swinging sideways was someone's intellectual property. [1]

2. A method of exercising a cat by pointing a laser on the floor and wiggling it around.

Real. U.S. Patent 5,443,036, "Method of exercising a cat." If you've ever owned a cat and a laser pointer, you have technically practiced the claimed invention. [1]

3. A sealed, crustless peanut-butter-and-jelly sandwich.

Real. U.S. Patent 6,004,596 (1999), the legendary "Uncrustables" patent. The Patent Office eventually struck it down in 2007, but not before it became Exhibit A in every "the patent system is broken" rant. [2]

4. A “Santa Claus detector”: a stocking that lights up when Santa arrives.

Real. U.S. Patent 5,523,741, issued in 1996. For the skeptical child who needs hard evidence. [1]

5. A method of apologizing to your houseplants out loud each morning to accelerate their growth.

Bluff. This one's ours. As far as we can tell, it is gloriously unpatented, so consider it open source.

If you got fewer than four right, don't feel bad. The point of the quiz isn't that the Patent Office is silly. It's that the same office that rubber-stamped a swing patent also processes around 600,000 applications a year and is currently sitting on more than 1.2 million of them. [3] Buried in that haystack are the patents that built the modern economy, and the ones that decided which startups lived and died.

The data your favorite VC forgot to mention

Economists Joan Farre-Mensa, Deepak Hegde, and Alexander Ljungqvist pulled off something clever. Patent applications get assigned to examiners more or less at random, and some examiners are far stricter than others. So two nearly identical startups can get different outcomes based on nothing but which examiner they drew. That lets the researchers isolate what a patent actually causes, not just what it correlates with.

The result: startups that won that lottery and landed their first patent saw, on average, 55% higher employment growth and 80% higher sales growth over the next five years than near-identical peers who lost it. Winners also went on to file more, and higher-quality, follow-on patents, boosting subsequent innovation by roughly 49%. [4] A separate breakdown of the same work found a first approved patent made a startup about 53% more likely to attract venture capital, and a 2022 USPTO analysis found that using patents as loan collateral was associated with a 76% jump in VC funding over the following three years. [5]

The $800 million windshield wiper

If the upside of owning a patent is real, so is the cost of being on the wrong side of one. Meet Robert Kearns.

In the 1960s, Kearns invented the intermittent windshield wiper, the one that pauses between swipes for light rain. He patented it, built prototypes, and pitched the design to Ford. Ford was interested, took a good look under the hood... and then passed. A few years later, in 1969, Ford rolled out its own "in-house" intermittent wiper that worked suspiciously like Kearns's. Soon Chrysler, GM, Mercedes, Honda, and basically every carmaker on Earth had one too. [6][8]

Kearns sued. He spent the better part of three decades in court, often acting as his own lawyer, battling Ford (1978–1990) and Chrysler (1982–1992). The automakers' defense was that the invention was "obvious" because it used off-the-shelf parts, never mind that none of them had thought to combine those parts that way until Kearns showed them. [8] He eventually won: roughly $10.5 million from Ford and around $30 million in settlements. [6][7]

Sounds like a happy ending, until you do the math the way Kearns did. Ford alone built more than 16 million cars covered by his patent. At the modest $50-per-car royalty he'd originally asked for, that single contract would have been worth over $800 million. [7] And the fight cost him almost everything else: a nervous breakdown, his marriage, and years of his life. His story became the movie Flash of Genius, and a permanent reminder that a patent is leverage, but only if you can survive the process of enforcing it. [6]

The lesson isn't "don't patent." Kearns's patent is the only reason he ever got paid at all. The lesson is that the system is brutal to the small inventor who has to navigate it alone, which brings us to the actual villain of this system.

The real problem isn’t patents. It’s the paperwork.

Getting a patent the traditional way is genuinely painful:

  • It's expensive. Attorney fees for a single utility patent typically run $9,000–$16,000, and the all-in cost through issuance commonly lands between $15,000 and $25,000. [9]
  • It's a slog. Most applications get hit with two to three "Office Actions", official rejections you have to argue past, at roughly $2,000–$4,000 each. [10] And nearly half of all applications never make it to grant. [11]
  • It's slow. As of fiscal year 2025, the average wait just for a first response from an examiner hit 26.3 months, over two years before the real back-and-forth even starts. [3] Filings hit an all-time high of 430,625 in 2024, and the backlog of unexamined applications peaked near 838,000 in early 2025. [11]

Where this actually goes

Here's the optimistic part, and the reason we do what we do. None of that friction is a law of physics. It's a paperwork problem, searching prior art, drafting tight claims, answering Office Actions, and monitoring for infringers. That's exactly the kind of slow, expensive, expertise-gated work that AI is built to compress.

The Patent Office itself sees it. Its leadership has been rolling out AI tools for examiners and racing to cut the backlog, already down from that ~838,000 peak to under 777,000, partly because the economics are stark: by the USPTO's own estimate, shaving even one week off pendency adds about $35,000 to the average U.S. company's value. [12] When a one-week speedup is worth that much, the prize for collapsing the whole process is enormous.

That's the bet behind ArcAngel AI: not to convince you patents are magic, but to delete the reasons you've been talked out of them. The value of a patent was never the problem. The on-ramp was.

So the next time someone tells you patents are dead, remember the five-year-old with the swing patent, the cat with the laser pointer, and the engineer who beat Ford and still got robbed of $800 million. The system is weird, slow, and occasionally ridiculous. It's also where moats get built. Don't skip the line, let's just make the line move.

That's the whole idea worth protecting.

References

Trademark Angel - Case Studies on Genericide: How Some Famous Trademarks Became Everyday Words (escalator, 1950; thermos, 1963; aspirin lost in the U.S. but retained abroad; Kleenex/Xerox policing). https://trademarkangel.com/case-studies-on-genericide-how-some-famous-trademarks-became-everyday-words/

University of Cincinnati Law Review Blog - Too Much of a Good Thing: How Genericide Sends Trademarks to the Graveyard (Velcro and Kleenex rely on combative advertising to survive).
https://uclawreview.org/2023/03/21/how-genericide-sends-trademarks-to-the-graveyard/

Cowan, DeBaets, Abrahams & Sheppard LLP - Elliott v. Google, Inc. (9th Cir. May 16, 2017) ("verb use does not automatically constitute generic use"; "Band-Aid® Brand" policing example). https://cdas.com/elliot-v-google-inc-9th-cir-may-16-2017/

Wikipedia - List of generic and genericized trademarks (Dumpster trademark, originally Dempster Brothers, cancelled in 2015). https://en.wikipedia.org/wiki/List_of_generic_and_genericized_trademarks

FindLaw - 'Google' Trademark Survives 'Genericide' (Google ranked among the world's most valuable trademarks; ~$113 billion estimate). https://blogs.findlaw.com/ninth_circuit/2017/05/google-trademark-survives-genericide.html

Library of Congress, Global Legal Monitor - China: Court Announces Apple and Proview Settlement of iPad Trademark Dispute (Apple paid $60M; original purchase ~£35,000/$54,500; Proview sought $400M). https://www.loc.gov/item/global-legal-monitor/2012-07-16/china-court-announces-apple-and-proview-settlement-of-ipad-trademark-dispute/

Forbes (Daniel Fisher) - Apple's Trademark Problem In China Is Self-Inflicted (China's first-to-file rule and squatting affecting Hermès, Chivas Regal, Michael Jordan, Jeremy Lin). https://www.forbes.com/sites/danielfisher/2012/04/06/apples-trademark-problem-in-china-is-self-inflicted/

U.S. Patent and Trademark Office - Trademarks: How much does it cost? (base application fee of $350 per class). https://www.uspto.gov/trademarks/basics/how-much-does-it-cost

Rapacke Law Group - The Real Trademark Filing Cost: 2025 USPTO Fee Changes (over 767,000 trademark application classes filed in FY2024). https://arapackelaw.com/trademarks/trademark-filing-cost/