Patents: how and why to get them

June 2nd, 2018

I'm going to discuss 3 very basic things about patents:

Some of my points are a bit naughty. But I maintain that they're based in fact and fairly widely known. So well-known, in fact, that I'm surprised to have never read it somewhere else.

My explanation is that the hatred of patents in the tech world is such that nothing except "HATE! HATE! HATE!" can be said on the subject in polite society. In this atmosphere, "Patents: how and why to get them" reads like "Humans: how and why to cook them."

If you can make yourself read this human-cooking manual, however, I think you'll find both amusing and useful things. I have more experience with patents than I've ever asked for, having worked on this stuff with lawyers from the smallest law firms to the largest ones, including lawyers who personally handled the most famous lawsuits for the most famous tech clients. I'm not an authority on patents, but I have good stories.

What patents give you

Some companies pay you money per patent. But it's rarely enough to make it worth your while, unless it's all you're doing. Patents look good on your CV, but reactions might be negative as well (you might appear "overqualified," "an expert in an unrelated field," etc.)

What's the one thing a patent undeniably buys you? A right to legally and publicly discuss your work – which you often can't get in any other way. This is not a side-effect of patent law, but its whole stated point. Patent law prompts companies to publish their ideas, in exchange for a time-limited monopoly right to use the ideas.

Note that publishing ideas in patents is easy, and the benefit for the author is certain. But getting and enforcing a monopoly for said ideas is not easy, so the benefit for the proprietor is not at all certain. Here's why.

What patents give (and don't give) your employer

Some problems with patents are so obvious that even patent lawyers will honestly discuss them with their clients:

But there's another, potentially far bigger problem, that patent lawyers will rarely mention, let alone admit its extent:

So what's the big deal, you might ask? The spec describes some device or method. The claims describe the supposedly new ideas used in this device or method. All you have to do is write a spec such that nothing of value is disclosed that is not covered by the claims.

However, in reality, the published spec is often quite close to the actual spec used by engineers, with all the details. That's simply the path of least resistance:

Ultimately, there's a conflict of interest between your employer and their patent lawyer, and a surprisingly perfect alignment of interests between the lawyer and yourself:

Funnily enough, this too is largely in line with the lawmaker's stated intent – prompting companies to publish ideas instead of keeping them secret. But why do companies file patents?

The answer is that patents are never read – they're counted. More precisely, a company's goal is to acquire enough patents so that they can only be counted – but not read and understood in a reasonable amount of time.

If you have too many patents to read and understand (hundreds, thousands or more), then investors and competitors alike assume you "own your domain" – you can counter-attack if sued. You're as well-defended legally as you can possibly be. But if you have few patents, someone might read and understand most of them – and create a narrative about some legal weakness. Such narratives are bad for the stock price.

This situation must be avoided. And that's all there is to it – at least in the computing industry. And I know it might sound too dismissive to be convincing. But the fact is that the content of patents is just too complex to drive business decisions. The feasible thing for a decision-maker is to pick the bucket to put you in, out of "no patents, some patents, a shitton of patents." For more information, see the seminal work "Pulling Decisions out of One's Ass: Fast and Slow," keeping in mind that decision-makers have a lot of decisions to make, so they must be Fast.

Why filing patents isn't a crime on par with cannibalism

Considering the above, I don't think that a product company employee filing patents pollutes the tech environment as badly as people believe.

Product companies file patents largely for self-defense. Some occasionally attack startups, but how many startups were destroyed by a patent lawsuit vs the number of those destroyed by a badly managed acquisition (with the original investors doing just fine)? And there are examples of big companies buying startups already attacked by a lawsuit filed by a bigger product company, confident that between two big companies, the legal result will be a stalemate. Thus for a big company genuinely fearing your product, it's much safer to buy you than sue you and have you bought by a big competitor.

The real trouble is patent trolls, who cannot be counter-attacked. But the only way a product company's patents will land in a troll's hands is if the company goes bankrupt and sells the patents. Well, guess what – in these cases, other product companies are eager to outbid the trolls. For example, when MIPS Technologies was sold to Imagination for ~$60 million many years ago, its patents, sold separately, fetched ~$500 million from some CPU cartel involving various big name CPU companies. Alternatively, a failing company can turn into a troll and sue a successful product company (MicroUnity comes to mind.)

Thus patents of failing product companies result in a weird form of socialism, where profit is spread more evenly between investors, with losers getting a chunk of the winners' profits. I don't think this chunk is nearly large enough on average to substantially reduce the incentive to work hard for the win, which is supposedly "the" trouble with laws subsidizing losers.

My point is that patent trolls and product companies seem to live in largely parallel universes. There are patents filed with the intention to be used by a patent troll, and there are patents filed by product companies, and the latter cause far less damage.

How to get a patent

I've lost count of the number of times I've heard the words "The Black Swan." It rather aggrieves me, but you gotta hand it to Taleb. Everyone is trying to pollute our language by needlessly coining catchphrases in a quest to be memorable, but he succeeded more than most. Surely I wouldn't hear this nonsense as often if he called the book "The Unforeseen Event."

Getting patents is a lot like branding. The trick is to call old things new names.

Why does it take a patent lawsuit and at least $1 million in legal fees to find out if a patent really is a patent – or to see it invalidated by the court? Because searching prior art is hard. "Prior art" includes everything published prior to the patent – older patents, academic papers, and everything else, really. Strictly speaking, you never know if you're done.

How does the patent office examiner examine prior art, at a cost much lower than $1 million? Some equivalent of quick googling. The input of search engines is words and short phrases. If you use words and phrases which are uncommon in your domain, the search will come up blank, or it will find things so obviously unrelated to your work that even a patent examiner will get it.

If you're extending the concept of a thread, don't call the result "extended threads", call it "hypercontexts." If you're calculating a histogram, call it a "distribution estimator." And so on. Again, I know this sounds too dismissive of the system to be believable. Well, try it. File a patent application full of "distribution estimators" and another one written in plain English. See which gets approved more smoothly.

Note that you might be tempted to conduct a prior art search yourself before filing the patent application, as a matter of due diligence. Yet some lawyers actually recommend against it, since if you do find prior art, you're now willfully infringing on it, and should cease and desist. My advice is to come up with a bunch of Black Swans/Distribution Estimators describing your idea, and pick the ones with the fewest Google results (patent search and otherwise).

And don't actually read any patent you accidentally find – don't willfully infringe, it's illegal. Just count them. Patents are never read, only counted – sounds familiar?

The other very important thing – which you mostly get to worry about in smaller companies – is to get the right kind of lawyer. Patent lawyers are fallen engineers, with engineering degrees, and sometimes actual engineering experience. The underlying engineer who's morphed into a lawyer ought to have specialized in your domain. No compromise is acceptable here. If you're doing optics, don't work with a guy who did chip design, and if you do chip design, don't work with a guy who did optics.

It doesn't matter if the lawyer is a Partner ($900/hour), an Associate ($450/hour), or some lesser life form in the law firm. It doesn't matter whether the firm is the biggest name in the industry or completely unknown. What matters is engineering knowledge. Don't expect a patent lawyer to honestly tell you he doesn't know your domain. He'll always accept the work, and you'll pay $truckload/hour trying to explain the most basic things to him, and failing. You need to actively ask about his education and experience.


Like most annoying things in life, patents aren't evil as much as they're absurd. Use them to your advantage.