Everyone says "patent it." In Tanzania, it's complicated.

Whenever someone shares a clever idea, we reach for the same line: “I hope you patented it.”

I felt that same pull with my own digital solution. A patent was never the plan, though: I had a hunch software is not really the kind of thing you patent. So I reached for the responsible-sounding alternative. I had just built something I believed in, and I wanted to protect it.

Mwalimu wa Mathe (Teacher Isaac) is a maths tutor for Tanzanian Form 4 students that lives primarily on WhatsApp, with a web app as the secondary door. A student sends a NECTA question and gets back a step-by-step solution narrated in Swahili, with an animation that draws the working out.

Since a patent was never on the table, I went the copyright route instead. I registered the code and content with COSOTA, the Copyright Society of Tanzania, and for a moment I felt like I had locked the vault. I had struck gold, and now it was mine.

Then, somewhere in the process of filling in the forms, I stopped to understand what I was actually registering. And it landed as a quiet reminder. I might feel I was first to the whole idea of a WhatsApp maths teacher that explains itself in Swahili, but feeling first is not the same as owning it. What I could register was my code and my content. The idea itself was never mine to keep. I had not protected the thing I thought I had protected.

That sent me back to the most famous idea story on the continent.

“Please Call Me”, and the myth of the patent

That instinct to patent is well meaning. It is also, most of the time, wrong. And “Please Call Me” shows it.

In 2000, a young Vodacom employee named Nkosana Makate came up with the free service that lets someone with no airtime nudge another person to call them back. It became part of how millions of people communicate. Makate then spent more than fifteen years fighting Vodacom for a share of what his idea was worth.

Here is the part most people get wrong. Makate did not win a patent case. He never had a patent. In 2016 the Constitutional Court found in his favour on a completely different basis: that Vodacom had made an agreement to negotiate compensation for his idea in good faith, and was bound to honour it. The court was careful to distinguish this from owning the idea itself. He won a contract, not a monopoly. And even now, more than two decades later, the two sides are still in court over how much he is actually owed.

The lesson is uncomfortable but important: you cannot own a bare idea. Ideas are free. I usually say it in Swahili: ideas ni kama machungwa, kila mtu ana lake na kufanana ni kawaida. Everyone has their own, and two looking alike is nothing unusual. What the law protects is something narrower, and if you are building software, it is worth knowing exactly what, because even the registration I did protects less than it feels like it does.

Why software is so hard to patent

Start with the myth: the software patent.

In Tanzania, the law does not ban it by name, and that is what makes it interesting. The Patents (Registration) Act, Cap. 217, defines an invention as “a solution to a specific problem in the field of technology”, then in section 7 lists what does not count. On that list sit scientific and mathematical theories, schemes and rules and methods for doing business or performing purely mental acts, and the mere presentation of information. A computer program is not named. But so much of what an app actually is, a method, a rule, a way of presenting information, walks straight into the exclusions that are.

This is not some local quirk. Most of the world treats software the same way. Patents are meant for novel technical inventions, and they demand novelty, an inventive step, industrial application, an expensive examination, years of process, and annual fees to keep them alive for their twenty-year life. A mobile app rarely fits that mould, and the parts that feel most clever, the idea and the method, are exactly the parts patents refuse to cover. The door is not locked so much as extremely hard to walk through.

So if you cannot patent it, and you cannot own the idea, what is left?

What actually protects a software founder

Four things, and none of them is a patent.

Copyright. This is the workhorse for software. Copyright protects the original expression of your work: your source code, your written content, your videos and animations, your artwork. It does not protect the idea behind them. It stops someone from lifting your actual code or content and passing it off as their own. In Tanzania, copyright exists automatically the moment you create the work, and you can register it for evidentiary weight with COSOTA.

Trademark. Your name and brand are protected as a trademark, registered through BRELA, not through copyright. If your product’s name is part of its value, this is where you protect it.

Trade secrets. The genuinely clever parts of how your system works, the parts you never publish, are protected simply by keeping them private and controlling who can see them.

Execution. The honest one. For most software, the real moat is not a legal document. It is speed, distribution, trust, and the quality of what you ship. Anyone can copy an idea. Very few can out-execute you on it.

So what did my COSOTA registration actually give me?

Let me be precise, because it is easy to oversell, and for a day or two I did oversell it to myself.

What it protects: my code, and more to the point, what my code produces. The real asset is not the code. It is the growing library of Swahili solutions and animations the app generates, each one drawing the maths out step by step.

The code leans on public building blocks anyone can pick up: a web framework, an animation engine, an AI model. The animations, in my own Swahili, taught the way I choose to teach, are the part that is genuinely mine. A rival can build their own tutor tomorrow. They cannot lift my videos and content and pass them off as theirs. For that, COSOTA gives me dated evidence that I made them first. That is real, and worth having.

What it does not protect: the idea of an AI maths tutor in Swahili, or the method my system uses to deliver it. And it was never going to be a patent, because patenting an app like this in Tanzania is a door almost nothing walks through.

The vault I thought I had locked turned out to be a receipt. A useful receipt, but a receipt.

And then there is AI

Which brings me to the elephant in the room. I built Teacher Isaac in about a week of spotty work, far faster than I could have managed a few years ago, because AI did a lot of the heavy lifting. That is what makes this sharper than it would have been back then: the barrier that used to protect a builder, the sheer difficulty of building the thing, is falling away for everyone at the same time.

If AI can help me build my idea in weeks, it can help someone else build their version of it just as fast. In a world where building is cheap, the idea is worth even less as a possession, and the code is not far behind. What cannot be copied by prompting a model is the boring, human part: knowing NECTA students, earning their trust, showing up every day, distributing to the people who need it, and being better next month than you were this month.

Registering with COSOTA was still worth doing. It just is not the moat. The moat is everything that happens after.

So the next time someone shows you something clever, do not ask if they patented it. Ask what they have shipped, and what they are shipping next.