Tuesday evening.

I was on the couch, half-watching something I can’t remember, when the WhatsApp notification came in.

A friend I hadn’t spoken to in a while. He works in HR. The message was short. He was considering switching to data analytics and wanted to know whether I still thought it was a good idea in 2026.

My first reaction was happiness.

Well, not because I had a clever answer ready.

Because he was moving. Most people I know who are unhappy in their work aren’t doing anything about it. They are sitting with it, reading LinkedIn posts about AI taking jobs, and calcifying. He was asking a question. That alone put him ahead of a lot of people.

I wrote back a few bullet points. Standard stuff.

Don’t quit your HR job.

Run both tracks in parallel, at least for the first year.

Give yourself margin.

The burnout path is the one where you bet the house on day one and then have to keep the bet alive through months of tutorials and rejected applications.

Build a portfolio, but not the usual kind.

Don’t stop at the dashboards and the SQL.

Show how you used AI tools to get there. That’s the skill that matters in 2026. Plenty of people can follow a pandas tutorial. Fewer can show they have figured out a working loop with an AI assistant for real analytical work.

Contribute to open source if you can. It counts as experience, and it forces you to read other people’s code, which is where the learning actually happens.

Sent.

And I sat with it for a while, because there was a bigger thing underneath those bullet points that wouldn’t fit into a WhatsApp reply.

For about a year now, I have been the AI champion at my company. The role where you go around helping teams figure out how to use AI tools in their actual work. Going in, I thought the role would make me more anxious about the future of this job. Spending every day watching people try tools that can write code would chip away at your sense of what you do, wouldn’t it?

The role clarified something I would never put into words. Software engineering, as a job, isn’t about writing syntax or knowing which framework is hot this quarter. It’s about solving problems in the shortest time at the highest quality the situation allows. Tools change, but the underlying work doesn’t.

Most engineers conflate the job with the tools. I used to. You spend years getting good at a specific stack, and you start to feel like the stack is the skill. When a new tool arrives that makes the stack cheaper to operate, it feels like the skill itself is being devalued.

It isn’t. The skill sits underneath.

My actual day now looks like it did three years ago in the shape of things. Features, bugs, and the usual arguments with customers about scope. What’s changed is the share of time I spend on each part. The syntax-wrestling part is shorter. The thinking part is longer. I spend more of the day on design decisions, on trade-offs, on whether we should even be building the thing at all.

Ask me whether that’s a better job or a worse job, and I would say better. The parts of the work I liked most were always the design parts. The parts I tolerated were the mechanical parts. AI collapsed the mechanical parts into a shorter form. The design parts are still mine.

When my friend asks whether it’s worth switching into tech in 2026, what I am answering is a different question. Is it worth becoming a person who solves problems for a living, using whatever tools are current, knowing those tools will change every two years for the rest of your career?

Not because tech is still an easy path in 2026. It isn’t.

The job market is harder than it was five years ago, the bar for entry is higher, and the noise of people telling you the industry is over is louder than ever. But those are conditions, not conclusions. If you accept that the underlying skill is problem-solving and not any particular tool, the conditions become things you navigate, not things that end the conversation.

When I told him to run both tracks in parallel and give himself a margin, I framed it as practical advice. Don’t burn your runway. Don’t bet the house.

It is that. It’s also something else.

The filter on this career isn’t whether you can learn Python, but whether you can keep showing up to an unfinished skill for long enough to see it turn into something. Months of tutorials that you don’t understand. Weeks of projects that don’t work. A gap between what you can almost do and what you can ship, which is always bigger than you think.

If you can’t give yourself the conditions to stay in that gap, you won’t make it through. Not because you are not smart. Because the gap eats people who don’t have the margin to tolerate it.

That’s why I told him to keep his HR job. Not as a safety net in a financial sense, although it is that too. As a patience budget. It lets him stay mediocre at the new skill for six or twelve months without that mediocrity costing him his rent.

A lot of career advice online treats conviction as the key variable. Go all in. Bet on yourself. I have come to think the key variable is closer to stamina. And stamina needs a source. For most people, the source is the thing they are still getting paid for while they learn.

If he can’t do that, if keeping HR while building the other thing feels like too much, then the honest answer is it probably won’t work anyway. Once the new skill becomes a job with deadlines, clients, and Monday mornings, the same tolerance is what gets tested. Better to find out now, cheaply, than later, expensively.

I learned programming mostly by reading books and following courses that weren’t quite for me. I remember the specific feeling of being stuck on something for hours, Googling the error, finding a Stack Overflow answer from 2013 that almost matched my situation, and not knowing how to bridge the gap.

Starting today, I will use an AI model as my patient tutor. Talk to it. Explain what I think I understand, and let it catch me where I am wrong. Ask it to generate five versions of the same concept until one of them clicks. This kind of learning didn’t exist when I started, and it’s better than what I had in a way that matters.

I wish I had it. And now he does. That’s a non-trivial advantage, and most people entering tech right now aren’t making the most of it.

Not industry analysis or tool tutorials, but as a human engineer in Singapore, working through this period out loud, trying to stay a full human being while the job keeps shifting shape underneath. If that sounds like something you would want in your inbox every week, you are in the right place.

Back to Tuesday evening.

I sent the bullet points. He read them. He sent back, “Thanks, bro, appreciate it.”

And I sat there thinking about everything I didn’t tell him, which is this article. The WhatsApp reply was the version for him, in that moment, that he could act on. This is the version for me, because writing it out is how I figure out what I actually think.

I hope he runs both tracks, builds the portfolio, and finds the patience to stay in the gap.

And I hope a year from now, he’s the one getting the WhatsApp message from someone else asking whether it’s still worth switching into this thing.

Cheers.