Ola-Obaado

September 29, 2026

The Rules Changed While We Were Working

The interviewer that kept disconnecting

This year I was interviewed for a frontend role by an AI. Or I tried to be. Four times in a row, the AI interviewer froze, looped the same prompt, and kept the timer running while I sat there, prepared and unheard.

I wrote to support. They sent me a checklist: update your browser, check your connection. Reasonable advice. But it missed the point I kept thinking about afterwards.

A machine was now standing between me and a job. Not replacing me. Gatekeeping me. And a few weeks later, another platform invited me to a long-term project helping to teach AI how to do the kind of work I do. In one summer, AI had become my interviewer, my potential student, and my daily coding partner. That is what the future of work actually looks like from the inside. It is not one dramatic replacement. It is a hundred small shifts in who holds which role.

The rules changed while we were working

I think about work as a game. Not a trivial one. A game with rules, moves and payoffs, most of which nobody explains to you. You learn them by playing and losing. AI did not end that game. It rewrote the rulebook mid-match and did not announce it.

The data shows the rewrite clearly, if you read past the headlines.

  • There is no economy-wide wipeout. Stanford's Digital Economy Lab, using payroll data on millions of US workers through June 2026, found no evidence of widespread job displacement.
  • But the entry door is narrowing. The same study found employment of 22 to 25 year olds in AI-exposed jobs now sits 19% below where it would be had it tracked less-exposed peers. Experienced workers show no such gap. The drop comes mostly from companies hiring fewer young people, not firing more.
  • Demand is coming back, at the top. Indeed's Hiring Lab reports US software development postings up about 15% since agentic coding tools arrived in early 2025, while overall postings fell 7%. Yet 71% of that increase was senior roles, and 37% came from jobs with AI in the title.
  • The skill set is being rebuilt. The World Economic Forum expects nearly 40% of core job skills to change by 2030, with 170 million roles created and 92 million displaced.

Read those together and a pattern appears. The game still pays. It just pays for different moves, and it has quietly raised the price of entry.

What AI takes off the table, and what it can't

AI takes the parts of work that were always a translation job. Turning a clear spec into boilerplate. Wiring a form to an endpoint. Writing the fourth CRUD route that looks like the first three. Drafting documentation nobody wanted to write. If you can describe the task completely, a model can usually do it faster than you.

What it cannot take is everything that sits before and after that description:

  • Deciding what the spec should be. On a medical records system, a clinician's request is rarely the real requirement. The real requirement is buried in how the ward actually runs.
  • Carrying context over time. Why this workflow has an odd extra step. Which stakeholder will reject a change. What broke last quarter and why.
  • Owning the outcome. When a payment fails or a patient record is wrong, a person answers for it. Accountability does not transfer to a tool.
  • Knowing when the output is wrong. Generated code looks confident either way. Spotting the subtle bug, the security hole or the pattern that will not scale needs someone who understands the system underneath.

The work that remains is less about typing and more about judgment. The typing got cheap. The judgment got expensive.

The new job description: see the system

If I had to rewrite the engineer's job description for this era, it would be three lines.

  1. See patterns before you see code. Most bugs, delays and bad features are the same few patterns repeating. A state machine nobody drew. A data shape that does not match the real world. A process optimised for the wrong person. AI helps you build faster. Seeing the pattern tells you what to build.
  2. Think in systems, not tickets. A ticket is a local move. A system is the whole board. The engineer who asks "what else does this touch?" is now worth more than the one who closes tickets fastest, because closing tickets is exactly what the tools are getting good at.
  3. Direct, then verify. Working with AI is closer to leading a very fast, very literal junior teammate. You set intent, give context, review the result, and catch what it confidently got wrong.

None of this is new. Good senior engineers always worked this way. What changed is that it used to be the top of the ladder. Now it is closer to the entry requirement.

Why I still learn backend the slow way

I have spent years deep in the frontend. This year I decided to go properly into backend: TypeScript on Node, a real database, authentication, email delivery, deployment. A fair question is why. An AI can scaffold a NestJS service with Prisma and JWT auth in minutes.

It can. I have watched it do it. And then I have watched it get the parts that matter subtly wrong: a token check in the wrong layer, a migration that would lock a table in production, an email setup that passes every DNS check and still lands in spam because of a mail filter nobody mentioned.

You only catch those if you understand what is underneath. So I learn by building a real project end to end, and I use AI as a sparring partner, not a substitute. I ask it to explain, to challenge my design, to generate the boring parts. I do not let it make the decisions I have not yet understood.

The risk in this era is not that AI knows too much. It is that we let it think for us before we have learned to think for ourselves. Understanding is what lets you direct the tool instead of just trusting it.

The junior engineer problem

This is the part that worries me most, because I see it up close. I mentor junior engineers and review their projects. The easy tasks that used to train them, the small bug fixes and simple features, are exactly the tasks AI now handles. Companies see that and hire fewer juniors. The Stanford numbers above are that decision, repeated across thousands of firms.

It looks efficient. It is a slow-motion trap. Every senior engineer was once a junior who got paid to be slow while learning. Remove that stage and in five years there is no one to promote. You cannot hire experience that was never allowed to form.

So the question for teams is not whether juniors still have value. It is how to redesign their first years. A few things I have seen work:

  • Make them reviewers early. Have juniors critique AI output against the codebase. Explaining why something is wrong teaches more than writing it from scratch.
  • Give them whole slices, not fragments. One small feature end to end, database to UI, teaches the system in a way ten isolated tickets never did.
  • Teach the why out loud. Knowledge-sharing sessions on data structures or architecture matter more now, not less. The tools give answers. People still need the reasoning.

If you are a junior reading this: your edge is not speed. It is understanding, curiosity and the ability to explain what the machine produced. Build those on purpose.

Access is not evenly distributed: the view from Lagos

Most writing on the future of work assumes stable power, fast internet and a card that pays for every subscription. I work from Lagos, and that assumption does not hold.

Nigerians are not behind on AI. A May 2026 report found 88% of Nigerian respondents had used at least one AI tool in the past year, and 74.1% of digitally enabled firms in Lagos, Rivers and Abuja had built AI into operations. The appetite is here.

The friction is elsewhere. The same report warns that access still depends on connectivity, affordability, devices and geography. An August 2026 IMF study named unreliable electricity, thin digital infrastructure and skills shortages as major brakes on AI adoption in sub-Saharan Africa. And the global race for AI chips is pushing up the cost of entry-level phones, the one device many young builders here depend on.

That is my AI interviewer story again. The tool was built for a smooth connection. When the network hiccuped, the system did not adapt. It just kept the clock running. If AI becomes the front door to global work, then a flaky connection or a declined foreign card becomes a new kind of barrier, invisible to the people who designed it.

The opportunity is real too. Remote work plus AI means a well-trained engineer in Akure can compete for work anywhere. But that only happens if we treat access, and the training pipeline, as infrastructure. Not as an afterthought.

A script audit for the AI era

There is an idea I keep coming back to: you are often playing a game you did not choose, by rules you were never shown, toward goals someone else set for you. The way out is not to quit the game. It is to audit your own script. What am I running, who wrote it, and is it still serving me?

AI makes that audit urgent. Here are the questions I am asking myself, and that I would put to any engineer:

  1. Which parts of my week could a model do today? Be honest. Those parts are not your career. They are overhead you can now hand off.
  2. What do I understand that the tool does not? The domain, the users, the history, the system. That is your moat. Deepen it on purpose.
  3. Am I learning, or only shipping? If AI writes everything and you understand none of it, you are getting faster and weaker at the same time.
  4. Who am I pulling up behind me? The pipeline breaks if seniors stop teaching. Mentoring is no longer a nice extra. It is how the profession survives.
  5. Whose rules am I playing by? Hype says "learn AI or die." Doom says "it is all over." Both are scripts written by someone else. Look at the data and your own work, then decide.

The game still runs

The future of work is not humans versus AI. It is people who understand systems working with tools that are very good at execution. The typing gets cheaper every month. The judgment, the context and the willingness to own an outcome get more valuable.

That is good news for anyone willing to go deep. It is hard news for anyone hoping the tools will do the thinking. And it is an urgent message for every team that stopped hiring juniors: the next generation of seniors has to come from somewhere.

As for that AI interviewer that kept freezing on me, I do not hold it against the machine. It did exactly what it was built to do. The lesson was for the people who built it, and for the rest of us: the tools will keep getting better. Whether the work gets better depends on us.

Sources