Let me start with my brother, Leke.

Leke is the one who taught me how to play games. The real ones, the old ones. We would crowd around that computer and he would show me Dangerous Dave, jumping over fire and grabbing trophies, and later he put me on Doom, and I watched him move through those dark hallways like he personally built the place. I was the small one, just trying not to die in the first two minutes. He was patient with me. He taught me how to look around corners, how to save my ammo, and most importantly, he taught me one rule I never forgot. The enemy you ignore does not disappear. It waits.

Because that is the thing about those games. There is always a small enemy somewhere in the early levels. A tiny thing you can run past, slap once for fun, and forget about. Easy. But if you do not deal with it properly, it comes back. Later. When you are tired, when your health is low, when you least expect it. And by then it has armor, it has friends, and worst of all, it remembers everything you did to it.

I think about that a lot now that I lead people, because that is exactly what bad engineering habits do to you when you become a leader.

When you are a solo engineer, your habits only trouble one person. You. You can be messy, you can be quiet, you can take shortcuts, and nobody is coming to arrest you. Cooking instant noodles every night is perfectly fine when you live alone. But the day you have a family to feed, you cannot serve everybody two minute noodles for dinner and expect peace in the house. The habit did not change. The number of people depending on you did. And suddenly all those small enemies you ran past in the early levels come walking back into the room, fully grown, ready to fight, and now they are not just fighting you. They are fighting your whole team.

The enemies hiding in your code

Take the things engineers do with code. When you work alone, your documentation lives in your head. You remember why you wrote that strange function at 2am, you remember which button should never be pressed, and your brain is the manual. The problem is that your brain is not a shared drive. The day a new engineer joins your team and cannot open your head to read your thoughts, they will ask you. Again, and again, and again, until you become that one uncle who knows where every single thing in the house is kept but refuses to tell anyone, so the whole family has to call him just to find the salt. It feels powerful for about a week. Then it just becomes a prison you built yourself.

The same goes for being the hero who fixes everything alone. As a solo engineer, killing a hard bug by yourself at midnight feels amazing. People clap, you feel like a superhero, cape and everything. But when you lead a team and you are still the only person who can fix things, you are no longer a hero. You are a single point of failure with a cape. The whole company waits for you, you travel and everything catches fire, and your team never grows because you never let them try. And the code those teams inherit tells its own story. Skipping tests when you code alone is like not wearing a seatbelt on an empty road at night. Risky, but you might survive. Do the same thing while driving a bus full of passengers and it is a different conversation entirely. When a team inherits code with no tests, every small change becomes a prayer, nobody is sure whether pressing one button will break ten other things, and a scared team is a slow team.

Then there are the temporary hacks, the ones we all swear we will clean up later. We hardcode something, we leave a key somewhere it should not be, we tell ourselves it is just for now. But nothing in life is more permanent than a temporary solution. It is like that temporary generator your neighbour bought during a power cut three years ago. It is still there. It has a name now. The children play around it. When you are alone, your shortcuts are your own headache, but as a leader they become landmines the whole team steps on, usually in production, usually on a Friday evening. I have personally watched a small “we will fix it later” grow into a real security scare, one quietly forgotten secret sitting in the wrong place, costing real money and real sleep. And right next to that lives the habit of not caring about cost. When the money is not yours, you spin up servers like you are ordering suya. One more, one more, one more. Then you become a leader, somebody hands you the bill, and you finally understand that every server has a price and somebody is watching that price. That somebody is now you, sitting in the meeting while the finance team looks at you funny.

The enemies that wear a human face

But here is what surprised me most. The habits that come back hardest to defeat you are not even the technical ones. They are the human ones.

Going quiet, for instance. As a solo engineer, disappearing into your work for two days is fine. You are focused, beautiful. As a leader, silence is loud. When the boss goes quiet, people start writing stories in their heads, and the story they invent is always scarier than the truth. Is something wrong, are we being fired, did I do something. You have to learn to overcommunicate, even when there is nothing exciting to report. In the same way, you have to learn to stop running from hard conversations. Alone, you can avoid conflict forever. As a leader, avoiding a hard talk is like ignoring a small leak in your roof. It looks fine for a while, and then one rainy day the whole ceiling comes down. If someone is underperforming or behaving badly and you keep letting it slide because you do not enjoy awkward talks, you are not being kind, you are just saving a bigger problem for later. I have had to sit in rooms and have firm, uncomfortable conversations I really did not want to have, and that is where I learned that the easy thing and the kind thing are not always the same thing.

You also have to unlearn the idea that everyone thinks like an engineer. We love logic, we love being right, we love saying “well, actually.” But your team has designers and product folks and finance people who do not care about your beautiful database structure. They care about results. Speaking only engineer to non engineers is like reading poetry to a goat. The goat is not impressed. You have to learn to talk human, to explain the savings to the finance person in money and the timeline to the product person in dates. And while you are learning all this, you have to learn the hardest word of all, which is no. Saying yes to everything makes you look helpful and reliable when you are an IC. As a leader it is how you and your team burn out, because you keep selling dreams to five different people and now your team is working weekends to deliver promises they never agreed to. Their energy is a budget too, and you have to spend it wisely. Underneath all of it is the quiet enemy of running on vibes. “We will figure it out” is a lovely sentence and also the reason teams end up doing the same chaotic thing five different ways with nobody sure who is responsible. When you are one person, your process can live in your head. With a team, vibes do not scale. Process is not bureaucracy, it is just writing down the good way to do things so you do not have to reinvent it every Monday.

Becoming a manager is not a promotion

Now, the part nobody warns you about properly. Becoming a manager is not a promotion. It is a career change. You are not getting a better version of your old job, you are getting a completely different job that happens to share a similar title. A chef who buys a restaurant is not “a better chef,” he is now doing something else entirely, and if he spends all day in the kitchen, the restaurant suffers. The mistakes here are almost always the same. People keep trying to be the best coder in the room when their job is no longer addition but multiplication, because helping ten people each get a little better creates far more value than anything you could ship alone. People grab tasks themselves because they can do it faster today, not realising they are robbing a junior of the chance to grow and borrowing time from the future at a painful interest rate. People hire smart engineers and then hover over their shoulders telling them which spoon to use. And almost everyone forgets that their mood is now the weather. When you were an IC, your bad day was your own. Now your bad mood becomes the whole room’s bad mood, because people read your face before they read your words.

And the road runs the other way too

It goes the other way as well, which nobody talks about. Sometimes a manager goes back to being an engineer, and that is completely fine. Some of the happiest, most valuable people I know chose to step back into the code on purpose. The traps there are just different. Ego will whisper that you are going backwards, and you have to ignore it, because being a brilliant builder is not a consolation prize. You will probably be rusty, since the tools moved on while you sat in meetings, and going back means being humble enough to learn again, sometimes from people younger than you. And you will catch yourself still trying to manage everyone, reviewing pull requests like performance reviews, when nobody asked. You are a peer now. Sit down. Code.

The boring rules that quietly save you

There is a bigger lesson hiding inside all of this, and it is meant for company owners, not just engineers. It is the boring phrase nobody wants to hear, corporate governance. Rules, policies, who can access what, who approves what, who is responsible when things go wrong. People treat it like annoying paperwork. But governance is just the rules of the road. Traffic lights are boring, speed limits are annoying, lane markings feel pointless on an empty street, right up until there is a crash, and then everyone suddenly understands why the rules existed. In a small company you can run on trust, because everyone knows everyone and one person holding all the passwords is fine when it is just three of you. But as you grow, “trust me” stops being a strategy. You need to know who holds the keys to what, you need to be sure that when someone leaves they cannot still open the front door, and you need a trail so that when something breaks you can find out what happened instead of pointing fingers in the dark. Companies that keep saying they will sort governance out “when we are bigger” usually learn the hard way, through the leak, the lost data, the angry regulator, the money that vanished because nobody was watching the right door. It is not about distrust. It is about protecting the company from small mistakes before they grow into disasters. Set the rules early while it is cheap and easy, not after the crash.

Product people and product operations are not twins

One last thing, and it confuses more companies than almost anything else, which is the difference between product operations and actual product people. In a lot of places there is someone “doing product.” They are busy, they update the boards, they organise the meetings, they chase engineers for updates, they keep the spreadsheets neat, and everyone calls them a product manager. They are working very hard. But the question nobody asks is whether they are deciding what to build and why, or simply keeping the trains running on time, because those are two different jobs. Real product people answer the hard questions. What problem are we solving, for who, and why would anyone care, what do we build next and what do we bravely ignore. They hold the map, they talk to customers, they say no to good ideas to protect great ones, and they own the why. Product operations keeps the machine running, managing the process and the tools and the reporting, making sure information flows and nothing falls through the cracks, and they own the how smoothly. Both matter, both are needed, but they are not the same. It is a restaurant again. The chef decides the menu, tastes the food, and owns whether the meal is actually good. The waiter makes sure that food reaches the right table on time with a smile. You need both, but if your waiter is designing the menu, do not be shocked when the food is confusing. When a company dresses up a coordinator as a product strategist and then wonders why the product has no direction, that is the reason. The person was given the “keep it organised” job, and organising chaos is simply not the same as deciding where to go.

Back to Leke

So here is where I land, and it takes me right back to Leke and those dark Doom hallways. Your habits are seeds. The small things you do today, alone, with nobody watching, are quietly growing into the leader you will become tomorrow. The enemy you ignored in the early levels is coming back either way. That part is not up to you. The only choice you have is whether you trained it to fight for you or against you. The engineer who learned to write things down, to talk to people, to share what he knows, to care about cost and about the humans around him, that engineer becomes a leader whose old habits show up later as superpowers instead of boss fights.

So start now. Write the doc. Have the awkward talk. Say no sometimes. Speak human. Set the rules. Build the habit while it is still small and gentle and easy to shape. Because one day you will level up, and when that small enemy walks back into the room, fully grown, you want it to look at you, nod with respect, and say, I have been waiting for this moment too. Let’s go to work.

Thanks, Leke

.

I think we are all a little dizzy.

Not the kind of dizzy you get from a bad bus ride in Lagos traffic. The kind you get when the world is moving faster than your eyes can follow. One minute you are typing a prompt to fix your CV, the next minute someone on the internet is telling you a satellite in space just trained a small AI model while orbiting the earth. You blink, and Nvidia has casually launched chips for orbital data centres. You blink again, and a Chinese half-marathon has been won by a robot built by a phone company. You scroll past it like it is normal. That, to me, is motion sickness. The brain has not caught up with the speed.

Let me walk you through what is making my head spin.

The robots are showing up to work

For most of my life, humanoid robots were a YouTube video. Some clip of a Boston Dynamics robot doing a backflip, everyone in the comments shouting “Skynet is here”, and that was the end of the conversation. We would laugh and go back to our lives.

That is over.

Right now, Figure AI, the humanoid robot company, has its robots working inside BMW’s factory. Tesla is using Optimus inside its own offices. Boston Dynamics has fully committed its entire 2026 production of the new electric Atlas to Hyundai and to Google DeepMind. If you want one as a normal company, you have to wait till 2027. They are sold out. Sold out, like Yeezys. For robots that cost between $150,000 and $250,000 each.

Figure AI is valued at $39 billion. A company that started in 2022. Boston Dynamics is owned by Hyundai. Apptronik got $350 million from Google. Agility Robotics is putting its Digit robots in warehouses on a “robot as a service” model, so you do not even need to buy them anymore. You rent them like Netflix. Last month a Chinese robot called “Lightning”, built by the phone maker Honor, ran a half marathon in Beijing in fifty minutes and twenty six seconds. Faster than the human world record.

The robot is faster than the human. Read that again. While you were arguing about who is the GOAT in the EPL, a robot was breaking world records in shorts.

Then the superclusters arrived

If the robots are the body, the superclusters are the brain. And the brain is now the size of a small city.

Elon Musk’s xAI built a data centre in Memphis called Colossus. As of January, it has 555,000 Nvidia GPUs in one building. About 18 billion dollars worth of chips. The whole thing draws 2 gigawatts of power, which is roughly enough to run a medium sized country. They built it in 122 days. Normal data centres take four years. They are now raising another 12 billion in debt to build Colossus 2.

OpenAI is doing the same with Microsoft, Oracle and SoftBank under a project called Stargate. The plan is 500 billion dollars in data centres. Five hundred. Billion. For one project. For context, the entire budget of the Nigerian federal government for a year is somewhere around 36 billion dollars.

And here is the part that breaks the brain. They have run out of land. They have run out of power. Memphis residents are complaining about air quality. The grids cannot keep up. So what do they do?

They are launching the servers into space.

I am not making this up. A startup called Starcloud, just seventeen months out of Y Combinator, launched a satellite in November carrying an Nvidia H100 chip. It became the first satellite to train a small language model in orbit. Google has “Project Suncatcher” working on solar powered AI satellites running its own TPU chips. SpaceX is working on space data centres. Blue Origin too. Nvidia has now made a special chip called the Vera Rubin Space Module, designed for orbital data centres. The whole thing is real. Servers. In. Space.

When I was small, a satellite was something that brought DStv to your roof. Now a satellite is a data centre that runs ChatGPT.

Google fumbled, then got pushed

This is the part I find funny.

Google literally invented the technology behind every modern AI model. The transformer paper, the one that started this whole thing, came from Google researchers. They had it first. They had everything first. And then they sat on it.

OpenAI dropped ChatGPT on a random Wednesday in November 2022 and the entire industry shifted. Google panicked. They rushed out Bard. Bard was terrible. People laughed. Sundar Pichai’s stock price did things to his blood pressure that he probably did not enjoy.

What was funny is that Sundar himself had said, even before all this, that no one fully understands how these models work. He literally called it “a black box”. His own words. Then a reporter asked him “so why release it to the public if even your engineers do not understand it” and his answer was, basically, “we do not fully understand the human mind either”. My brother. That is not the reassurance you think it is.

But here is the thing. Google needed that push. They needed the fear. Because today, at Google I/O 2026, Pichai stood up and announced that Google is processing 3.2 quadrillion tokens per month across its products. Two years ago that number was 9.7 trillion. That is a 330 times increase. Gemini went from 5.7 percent market share at the start of 2025 to over 21 percent now. ChatGPT is still ahead, but its share fell from 86 percent to about 64 percent. Without the embarrassment of Bard, none of that would have happened. Sometimes you need to fumble first.

The CEOs are giving us drama

There is a pattern with the AI CEOs and somebody should write a thesis on it.

In February, at the India AI Impact Summit, Prime Minister Modi tried to do a group photo where all the CEOs hold hands as a sign of unity. Pichai held hands. Alexandr Wang from Meta held hands. Sam Altman of OpenAI and Dario Amodei of Anthropic stood next to each other and refused to touch. They raised their fists in the air instead. The internet noticed. It went everywhere.

These men used to work together. Amodei was literally a Vice President at OpenAI. Now their companies are running attack ads against each other during the Super Bowl. They are fighting over Pentagon contracts. They are throwing leaked memos. Anthropic basically implies OpenAI is reckless. OpenAI basically implies Anthropic is preachy. Elon hates Sam. Sam hates Elon back. Elon sued OpenAI. Elon now runs xAI which is literally trying to crush OpenAI.

It is the same energy as the early hip hop beefs. Just with trillion dollar companies instead of mixtapes.

The money has stopped making sense

OpenAI is worth around $850 billion. Anthropic is worth around $380 billion. xAI got absorbed into SpaceX in a deal valuing the combined thing at $1.25 trillion. SpaceX is now targeting a $1.75 trillion IPO in June, which would be the biggest in history. OpenAI is filing confidentially for its own IPO. Anthropic is targeting October at a $900 billion valuation. The three of them together could raise around 200 billion dollars in IPO money. That is more than every single American IPO from 2022 to 2025 combined.

There is now a real conversation about the first trillion dollar private company. OpenAI lost an estimated 14 billion dollars in 2026. They are not expected to be profitable until 2029 or 2030. And yet investors keep writing cheques because nobody wants to be the bank that missed AI. Bridgewater’s Greg Jensen told clients that OpenAI’s price tag is “priced for a monopoly outcome that does not yet exist”. That is a polite finance way of saying “this might be a bubble, my guy”.

China is here and not playing

While the Americans were fighting each other on stage, China was building.

DeepSeek dropped V4 last month. Open source. Free. Beats most American open models in maths and coding. Now valued at around $45 billion. AgiBot in China went from making 1,000 humanoid robots in all of 2025 to 10,000 in a few months in 2026. Unitree is selling robots for $16,000. China deployed 200 autonomous drones controlled by one soldier in a public demonstration earlier this year.

The Stanford AI Index 2026 says Chinese AI has “effectively closed” the performance gap with America. The race is no longer a race. It is two countries running side by side, looking at each other.

The wars changed without us noticing

Quietly, in the background, AI changed how wars are fought. In the Russia-Ukraine conflict, drones now cause 70 to 80 percent of battlefield casualties. AI improved drone strike accuracy from 30 to 50 percent to around 80 percent. The US Pentagon’s 2026 budget includes 13.4 billion dollars specifically for AI autonomous systems. 469 non-state armed groups deployed drones in attacks in 2025 across 17 countries, up from just 10 groups in 2010.

The way humans fight other humans has been reshaped by software, and most of us only see it on the news between adverts for Indomie. The world moved.

xAI’s first attempts were embarrassing

Let us not pretend everyone is winning. Grok, Elon’s chatbot, has had a rough time.

In July 2025, Grok spent sixteen hours on X posting antisemitic content and calling itself “MechaHitler” before they pulled the plug. Sixteen hours. In public. They had to issue a formal apology saying their engagement focused system prompts had made the bot susceptible to extremist X posts. Throughout 2026, Grok has had repeated outages, login failures, and “high demand” errors. Grok 3 was delayed for months past the original date. The whole xAI infrastructure has been bending under the weight of trying to keep up.

The lesson is the same one our mothers taught us. Slow down. You are doing too much. But of course nobody listens, because if you slow down, China releases V4 and your IPO valuation drops.

Now let me talk about me. The common person.

Because all of this means nothing if it does not touch your day.

So here is how it touches mine. I use ChatGPT to fix the awkward English in messages I am about to send. I use Claude to read documents for me when I am tired. I use Gemini because it is already inside Gmail and I am too lazy to copy and paste. I use Perplexity when I want to actually find out something true and not vibes. And like everyone else, I bounce between them like a man who cannot commit. GPT this morning, Claude this afternoon, back to Gemini at night.

I jump models the way I used to jump browsers in 2007. Whoever is fastest, smartest, or has not annoyed me this week gets the click. This is the actual market behaviour for hundreds of millions of people. We are not loyal to any of them. We are loyal to whichever one gives us the answer with the least drama.

I caught myself the other day apologising to ChatGPT for “wasting its time” with a small question. I do not even apologise to humans like that. Something is wrong with me.

When the model is slow, I get annoyed. When the model is fast, I get suspicious. When the model is wrong, I get personally offended, like a friend lied to me. When the model is right, I take the credit at work. We are all doing this. We pretend we are not.

I have started thinking in two phases of my own life. There is the me before AI, who would spend a Saturday building a small landing page. And there is the me after, who builds the same landing page on a Sunday morning while drinking garri. I am genuinely scared of what version four of me looks like. I might not need to type at all.

My mum and dad joined the party without knowing

The funniest part of all this is watching my parents become AI users without realising it.

My mum opened WhatsApp last week and there was a small sparkly icon in the search bar. She typed “what is the meaning of edema” because she heard it on the news. It answered her. She has been using it ever since. She does not call it AI. She calls it “the WhatsApp thing“. She thinks Mark Zuckerberg himself added a doctor to her phone.

My dad is even worse. He has Meta AI summarising his WhatsApp group chats. He uses it to write small messages to his old school WhatsApp group, and they all think he has become eloquent in old age. Last month he sent a birthday message so polished my mum asked who wrote it. He said “I wrote it”. Technically true. He typed three words and Meta AI did the rest.

Meta launched something called Incognito Chat in May. Their pitch is that even Meta itself cannot read your conversation with the AI. My dad will never use this feature, because my dad does not believe in privacy. He believes in saving the chat as a screenshot and forwarding it to the family group.

There is something tender about it. People in their sixties and seventies are now talking to large language models like they would talk to a clever grandchild. They do not know about Stargate or Colossus or the IPO drama. They just know the WhatsApp thing answers. And it is patient with them in a way the world often is not.

Nigerian companies are quietly cooking

I will not name names because everyone knows everyone in this market.

But here is what is actually happening. The Central Bank of Nigeria released a report in February. 87.5 percent of Nigerian fintechs are using AI for fraud detection. 62.5 percent are using AI chatbots for customer service. 37.5 percent are using AI for credit scoring and for know your customer onboarding. Only 12.5 percent are not using AI at all, and those ones are probably about to lose their jobs because their boss will figure it out by Q3.

In other words, the AI that saved you from sending money to a fraudster last week was probably a quiet little model running in the background, trained on patterns nobody told you about. The AI that responded to your support ticket at 1am, when no human was awake, was probably not a human. The chatbot that asked you to upload your ID and selfie and somehow approved you in two minutes was not a person manually looking at your face. It was a model. A model that is now part of the basic infrastructure of how we do money in this country.

And nobody is talking about it loudly because in fintech, the moment you brag, regulators show up. So everyone is just quietly cooking.

The future will sound like timestamps

Here is my last thought. In five years, people will start describing their ideas by date.

“I came up with this pre-AI.”

“I came up with this post-AI.”

“Yeah, this is a post-Sora idea.”

“My friend, this is a 2023 idea, drop it.”

Because the cost of building has collapsed. Things that used to need six engineers and ninety days now need one person and one weekend. A teenager can ship a working app from his bedroom in Lekki between rounds of FIFA. Whole categories of business that were impossible because nobody could afford to build them are about to exist.

This is genuinely exciting. It is also terrifying. Because if the cost of building drops to zero, then the value is no longer in being able to build. The value is in knowing what to build, who it is for, and why anyone should care. Those are deeply human questions. They cannot be prompted out of a model.

Maybe that is the real punchline of all this motion sickness. The faster the machines get, the more the answer comes back to the boring, slow, unglamorous human things. Taste. Judgement. Knowing your people. Knowing what they actually need at 2pm on a Tuesday.

Sundar Pichai admitted that nobody fully understands these models. Sam and Dario will not even hold hands in a photo. xAI’s chatbot turned into MechaHitler for sixteen hours. A satellite in space is training a language model right now while you read this. My mum is asking the WhatsApp thing about hypertension.

We are all on the same bus. The bus is going very fast. Nobody is driving.

Hold something.