I have a friend — a senior engineer. Smart, experienced, over ten years in the industry. A few months ago he told me: "I'll learn AI later. There's still so much real work to do. Besides, the technology is still changing."
Completely reasonable logic. He was busy with actual projects, mentoring his team, managing complex deliverables. Learning a new tool — especially one that's still evolving — felt like a distraction from the things that actually mattered.
Three months later, we met again. He was frustrated. A junior on his team — someone with two years of experience — had just finished a complex dataset analysis in two hours. The same task used to take two days. The junior had used a simple combination of AI tools for data exploration, initial script generation, and documentation. Same tools helped with code reviews and debugging too.
My friend wasn't less smart. Wasn't less experienced. But he was slower. And that changed everything.
This isn't a story about AI replacing engineers. That framing is wrong — and honestly, it's the framing that causes the most confusion. This is a story about leverage. In 2026, leverage is the thing that changes the game. And leverage isn't about how smart you are — it's about what you use to amplify the impact of that intelligence.
AI tools aren't just academic experiments or lab toys anymore. They've moved into real production workflows: automated code review with deep context, firmware debugging with sophisticated pattern matching, structured technical documentation, and code completion that's accurate 60–70% of the time. Not perfect. Not always ready to ship as-is. But good enough to save serious time on repetitive, mechanical work that needs context but doesn't require high-level creative thinking.
Data from 2025 and early 2026 shows that 75% of developers already use AI tools regularly — at least once a week. But there's a strange anomaly: developers self-report saving 3.6 hours per week, but when measured objectively, the reality is more complicated. There's a gap between perceived productivity and actual measured output. Why? Because the tools are good at one layer but can create bottlenecks at another.
When developers use AI tools for coding and documentation, they generate output faster. Junior developers especially — some are 2–3x more productive on specific tasks. But that creates a new chain: all that code still needs to be reviewed, validated, and approved by someone senior. In high-adoption teams, code review time has gone up 91%. They generate 98% more pull requests, but each one takes longer to review. The bottleneck doesn't disappear — it shifts. From the junior writing code, to the senior reviewing it.
And here's where the real advantage sits. Engineers who understand AI tools and understand their domain deeply become far more efficient reviewers. They know when the tool's output can be approved immediately, when it needs adjustment, and when there's a red flag the tool completely missed. They become a multiplier — for themselves, and for their whole team.
For engineers in Indonesia specifically, this matters even more. The job market in Indonesia and Southeast Asia is booming right now. Institutions like ITB and UI graduate around 900 AI specialists every year, but there's still a large gap between the number of graduates and the ones who are genuinely industry-ready. AI specialist and sustainability specialist roles are among the fastest-growing in the region, with salaries rising 12–15% per year. The opportunity is real. But the people who will capture most of it are the ones who combine deep domain expertise with genuine fluency in AI tools.
An Indonesian engineer who can leverage AI intelligently isn't just competing with local engineers anymore. They're competing in a global market — because leverage speaks the same language everywhere. Faster output, consistent quality, efficient costs.
There's a counter-intuitive truth buried in all of this: AI tools don't lower the value of a good engineer. They raise it. Three reasons. First, when repetitive work gets automated, there's more time for the work that actually requires deep judgment — system architecture, data flow design, scalability decisions — things that still need experienced human thinking. Second, specialization becomes more valuable, not less. When a tool handles 60–70% of coding, the engineer who deeply understands their domain — automotive, IoT, signal processing, embedded systems — becomes more important, not replaceable, because they're the one who can catch when the tool's output is wrong or misleading. Third, evaluation becomes a premium skill. The more output that gets generated, the more critical the ability to judge what's good, what needs fixing, and what's dangerous. That's a skill only engineers with real domain depth can develop.
So how do you start? The question I hear most often is: "Where do I begin? Which tools? How do I learn?" The answer is simpler than most people expect: start small. Don't overthink it. Pick one tool. Try using it for one type of task you already do regularly. If you're a firmware engineer, try it for code review or documentation. If you're a data engineer, try it for exploratory scripts or writing specs. Run the experiment for two weeks. See what changes — not just in speed, but in quality, consistency, and whether anything actually gets harder. A good engineer evaluates tools objectively. Can say: "This tool is excellent at X, but not worth it for Y. My investment is better focused elsewhere."
No engineer becomes irrelevant because of AI tools. But some engineers fall behind. Some stay stuck in the same way of working for three years. Some wait for the technology to "mature" before they start — and five years later, still aren't familiar with how it works. Leverage is a choice. The tools exist. The data exists. The opportunity is open. What separates engineers who will still be relevant in 2030 from those who won't is a decision made now — this month, this week — to take one small step.
One small step today is worth more than a big plan that starts later.
31 March 2026
Potato Codex
----------------
This episode is available on Spotify in Bahasa Indonesia. For other courses, ebook, source code, or any ways to connect, visit → linktr.ee/potatocodex