How to Stay Relevant When AI Writes Your Code

Welcome back.

"If I'm not writing the code, what am I doing?"

I've heard this question, or some version of it, from developers at every level over the past few months. Senior engineers. CTOs. People who've been in the industry for over a decade.

It's the question of 2026. And today, we're going to answer it.

Let’s dive in.

AI News Roundup

↗ OpenAI's $60 Billion Bet
Nvidia, Microsoft, and Amazon are in talks to invest up to $60 billion in OpenAI. This would be one of the funding rounds in tech history.

↗ Meta Acquires Manus for $2.7 Billion
Meta acquired autonomous agent startup Manus for $2.7 billion, signaling they're betting big on AI agents that execute complex tasks with minimal oversight. Agentic AI is the next frontier, and Meta just bought a head start.

↗ Musk Pivots Tesla to AI and Robotics
Elon Musk announced that Tesla is officially shifting from cars to AI and robotics. Optimus humanoid robots and full self-driving. The pivot isn't new, just official. Can Tesla execute this at scale?

YOU’RE INVITED: AI Hack & Learn

Amazon just announced 16,000 layoffs. Meta cut 1,500 last week. Half a million tech workers have lost their jobs since ChatGPT launched.

And the companies laying off? They're investing more in AI, not less.

So, I'm hosting a free 3-hour workshop on Friday (tomorrow) for anyone who wants to actually understand these tools, or finally build that project you've been putting off.

Whether you've never touched generative AI or you've been tinkering but haven't shipped anything, come spend three hours getting clear on what matters.

I'll walk you through the basics. Answer your questions. And give you the space to actually build something instead of just reading about it.

The Question Nobody Wants to Ask

"If I'm not writing the code, what am I doing?"

I've heard this question in nearly every conversation with developers over the past few months. It's whispered in Slack DMs. It comes up in 1-on-1s. It's the subtext of every thread about AI tools.

For twenty years, developer identity was built on translating logic into syntax. That mastery, earned through years of debugging at 2 am, defined who you were.

GitHub's research team studied this and found developers move through four stages: Skeptic → Explorer → Collaborator → Strategist.

The identity crisis peaks during that Explorer-to-Collaborator transition. That's when developers shift from using AI for boilerplate to genuine pair-programming.

It's the moment when you realize: you're not coding anymore. You're directing.

And for a lot of people, that feels wrong. Like they're not doing "real work" anymore.

The Supervision Paradox

Validating AI code requires the very expertise that not coding might erode.

Researchers call it the "supervision paradox."

AI handles 70% of the work blazingly fast—the boilerplate, the standard patterns, the CRUD operations. This is where all those productivity stats come from.

But the remaining 30% — edge cases, security hardening, production readiness — requires deep expertise that only comes from hands-on experience.

And here's where it gets uncomfortable.

Studies show AI-generated code with inadequate review produces:

• 2.5x more security vulnerabilities

• 322% more privilege escalation paths

• 153% more design flaws

Someone has to catch those issues. And if you haven't been building your intuition through hands-on work, can you?

Meanwhile, junior developer hiring dropped nearly 20% between 2022 and 2025. The entry-level pipeline that traditionally built that expertise? It's narrowing right when we need it most.

What the Code Analysis Shows

GitClear analyzed 211 million lines of code. The findings are troubling:

• Copy-pasted code up 17%

• Duplicated code blocks up 8x

• Code churn (revisions within two weeks) up 26%

Developers themselves rank code quality as their top concern, even above job security.

Addy Osmani from Google wrote something that stuck with me: "Vibe coding is not the same as AI-assisted engineering."

There's a crucial difference between using AI as a productivity tool and blindly accepting whatever it generates.

When coding becomes conversation, something shifts. The friction that forced you to truly understand the problem? It disappears. The struggle that built intuition? Gets outsourced. The tacit knowledge accumulated through countless small debugging victories? It never forms in the first place.

That's the loss nobody's quantifying yet.

What Actually Matters Now

The developers thriving in this new world aren't better coders. They're better orchestrators.

Here's where they're spending their time:

Understanding the work: AI fluency, yes. But also product thinking. What should we build? Why? What success criteria matter?

Directing the work: Specification writing. Architecture design. Delegating to multiple AI agents while maintaining coherence across all of them.

Verifying the work: Many advanced AI users report spending more time verifying than generating. The job shifts from writing to validating.

Here's a data point that tells the whole story: TypeScript became GitHub's #1 language by contributors in 2025.

Not because developers suddenly love types (despite what they might tell you)…but because structure helps AI-assisted work!

The tools that constrain AI creativity are becoming more valuable than the ones that enable yours.

Think about that for a second.

Three Paths Forward

Annie Vella, a Distinguished Engineer, wrote about the software engineering identity crisis and identified three strategies:

1. Resist. Double down on domains requiring deep expertise like systems programming, performance optimization, and security. Places where AI's 70% isn't good enough.

2. Adapt. Fully embrace the orchestrator role. Become the person who knows how to get reliable results from AI, every time.

3. Balance. The "engineer/manager pendulum" approach. Swing between direct problem-solving and strategic guidance. Never fully abandon hands-on work.

The smartest developers I'm seeing aren't picking one path. They're deliberately practicing all three.

Side projects to maintain craft. AI orchestration for productivity. Strategic oversight for career growth.

They're not choosing between AI and hands-on coding. They're cycling between both.

The Honest Answer

Is this evolution or loss?

The optimistic take: Developers are reclaiming their original engineering identity. Before we became code monkeys, engineers did requirements, design, architecture, and operations. AI is returning us to that broader role.

The pessimistic take: We're trading deep craft for shallow fluency. The generation that learned to code by building will be replaced by one that learned to code by asking.

The honest take: Both are true.

And where you land depends on the choices you make right now.

Within five months of launch, developers merged over 1 million pull requests using GitHub's Copilot coding agent.

The shift isn't coming. It's here.

Your Turn

I want to know what you're experiencing:

If you're using AI coding tools:

• What surprised you most?

• Where is it genuinely helping?

• Where are you catching issues that it missed?

If you're not using AI tools yet:

• What's holding you back?

• What would need to change?

Hit reply. I read every single one, and these conversations shape what I explore next.

Until next week 🖖,
Haroon