Patent

Anyone who works out a new method runs into a fork with no good side. Explain the method and whoever reads the explanation can build the same thing. Keep it quiet and the method lives only as long as the person who knows it. Both roads cost something.

The consequences of picking secrecy are on record. Medieval craft guilds never wrote down the mixture that produced a particular dye. They passed it from master to apprentice by mouth, which meant that when a guild broke up the technique broke up with it. A century later somebody had to work out the same thing from nothing.

A patent turns that fork into a trade. The inventor has to write the method down in enough detail that a competent person in the field could rebuild it, and the document becomes public. In exchange the state grants a fixed period during which nobody else may use it. The knowledge survives because anyone can read the filing, and the inventor is not punished for publishing it, because reading is allowed while copying is not.

That bargain is also why the right never appears on its own. Among the four kinds of intellectual property, only trademarks and patents have to be applied for and examined before they exist at all. A copyright attaches the moment the work is made and a trade secret exists as long as it stays hidden, but a patent starts as paperwork. Timing is half the work: an invention made public before the filing goes in can never be patented.

You file an application with the patent office, an examiner checks it against the statutory requirements, and if it passes the patent registers and stays in force for up to twenty years counted from the filing date rather than from registration.

The examiner is testing three things.

1. Novelty — the invention must not already be known anywhere in the world. That includes disclosures by the inventor. A detailed engineering blog post or a conference talk becomes prior art that blocks the applicant's own filing.
2. Non-obviousness — if someone skilled in the field could reach the invention by combining things already known, it is refused. This is the requirement most applications die on.
3. Industrial applicability — the invention has to be something that can actually be made and used. A principle with no described way to implement it stops here.

What the patent covers is not the invention as a whole but the claims, a numbered list inside the application stating exactly which combinations of features are being claimed. Claim drafting is a tug of war. Write the claims broadly and they are hard to design around, but they overlap more prior art and get refused. Write them narrowly and registration comes easily, while a competitor who changes one step of the process falls outside them entirely.

Cost and time are in a different class from the other rights too. A single domestic filing runs roughly $2,300 to $3,800 once attorney fees are counted, and one to three years passes between filing and registration. Patents are also granted country by country, so extending coverage to the United States and Japan pushes a single invention into tens of thousands of dollars. Nothing here is fast or cheap enough to do casually.

At a startup, patents tend to come up in four situations.

1. Fundraising — investors ask during due diligence for a list of granted patents and pending applications. A company that pitched itself as technology-led with an empty list has to support that claim some other way.
2. Enterprise contracts — large customers ask for a warranty that the product does not infringe anyone else's patents. Signing it means the vendor absorbs the damages if infringement surfaces later, so it is not a clause to wave through.
3. Competitive response — a granted patent gives a company something concrete to say when a rival ships a copy of its feature. The same mechanism runs in reverse: if the rival registered first, the letter arrives at your desk instead.
4. Grants and programs — public research funding and technology assessments count registered patents as scored line items.

Of the four, the ones that move the calendar are the first and the third. A round coming up pushes a filing forward, and word that a competitor is building the same feature pushes the claims wider.

"Let's get the application in before this goes on the blog. Publishing kills novelty."

An argument about sequence, not about whether to publish. Get the order wrong and there is no way back.

"It's granted, but the claims are narrow enough that we can't use it against their product."

Holding a right and being able to enforce it against a specific competitor are two different things.

"We got a demand letter. Let me pull the patent number and look for invalidity references first."

Before conceding infringement, the team will argue that the other side's patent should never have been granted.

Pagedeck ran into this shortly after its Series A. Theo had worked out a better way to reduce conflicts when several people edit the same document at once, and the team wanted to write it up on the engineering blog. Grace pushed publication back a week and booked a patent consultation first, because publishing would destroy novelty. The application went in and the post followed three weeks later.

When this comes up in a meeting, what actually gets decided is sequence and budget: file now or next quarter, domestic only or add the United States.

The most common assumption is that a patent means nobody else can build the same thing. The word monopoly invites that reading. In practice the blocked zone is narrow — it covers products that use the combination written into the claims — and a competitor who swaps one step for a different method walks past it. On top of that, no government enforces the right for you. Stopping infringement means sending a demand letter yourself and being prepared to litigate. A company that assumes the certificate itself is protection discovers the gap only when a copy ships. A patent is a legal instrument, not the kind of compounding advantage a moat describes.

The other frequent mistake is thinking an idea is enough to file on. The word invention sounds like a flash of insight. An application has to describe a concrete structure or process that implements the idea, and if that part is thin the examiner refuses it. Founders at the idea stage also explain the concept to investors and partners, and each conversation is a disclosure that eats into novelty. If the idea has to leave the room before filing, an NDA goes in place first, and reversing that order cannot be fixed later.