Is Learning to Code Still Worth It in the AI Era? An Honest Answer from an Atlassian Engineer

Where the Coding Anxiety Comes From

Since 2016, the tech industry has seen over 75,000 layoffs. AI can write complete code and build applications in seconds, leaving one question lingering in everyone's mind: Is learning to code still meaningful in the AI era?

OpenAI CEO Sam Altman says "coding is now overrated"; former Google CEO Eric Schmidt says "the best programmers no longer write code—they guide AI systems"; Anthropic's Claude Code contributor Boris Cherny even declares "software engineering is dead." When tech industry leaders are all saying things like this, students become confused and anxious, and some even abandon their plans to learn programming.

But is reality really like this? Harshit Rayyan, a software engineer who has worked at Atlassian for 3.5 years, offers a strikingly different answer from the perspective of a frontline developer. Atlassian is an enterprise software giant headquartered in Australia with a market cap exceeding $40 billion. It owns products like Jira, Confluence, and Bitbucket that serve over 300,000 enterprise customers worldwide. As a leading company in the developer tools space, its engineering team's hands-on experience with AI-assisted development is highly representative of the industry.

Three Reasons Behind the "Coding Is Dead" Panic

The Explosion of AI Tool Capabilities

Today, a single prompt can generate complete application code—frontend, backend, database connections, APIs, everything. When AI appears to "do it all," people naturally wonder: if AI can write code, why should I bother learning?

Current mainstream AI code generation tools include GitHub Copilot, Cursor, Claude Code, Replit Agent, and others. These tools are based on Large Language Models (LLMs), trained on massive code corpora, capable of understanding natural language descriptions and generating corresponding code. GitHub Copilot already has over 1.5 million paid users, while Cursor's ARR (Annual Recurring Revenue) surpassed $100 million in 2024. The core technology behind these tools is code completion and generation powered by the Transformer architecture. They excel at handling pattern-based, repetitive coding tasks but still have clear limitations in scenarios requiring deep business understanding and cross-system coordination.

Massive Layoff Waves

Tech companies are laying off employees while heavily investing in AI, creating a seemingly logical narrative: AI is replacing developers. News about layoffs intertwined with news about AI investments amplifies people's fears.

However, the 2022-2024 tech layoff wave has complex structural causes. Over-hiring during the pandemic is one of the main reasons—Meta's headcount grew nearly 60% in 2020-2021, followed by laying off over 20,000 people in 2022-2023. Rising interest rates increased the cost of capital, forcing companies to shift from "growth at all costs" to "profitability first" mode. While some layoffs are indeed related to AI automation, most are corrections to pandemic-era over-expansion rather than direct AI replacement of human roles.

Shocking Statements from Tech Leaders

Statements from Sam Altman, Eric Schmidt, and others spread rapidly due to their influence. But there's a key question to understand here—what do these statements actually mean?

The Real Intent of Tech Leaders: Marketing Rhetoric vs. Actual Meaning

Companies like OpenAI and Anthropic have AI code generation tools as their core business. Naturally, their marketing messaging emphasizes how powerful their tools are. This isn't false advertising, but the information is packaged in a specific way that can create misconceptions in the audience's mind.

"Coding is dead" spreads more easily than "coding is evolving" because the former has stronger emotional impact. AI companies leverage this viral mechanism to promote their products.

So what do these leaders actually mean?

When they say "coding is dying," they're actually referring to low-level coding dying—copy-paste work, basic CRUD operations, tutorial-level projects. These are indeed tasks that AI can already handle competently.

A quick explanation: CRUD stands for Create, Read, Update, and Delete—the four basic database operations. "Low-level coding" here doesn't refer to low-level systems programming, but rather development work that follows fixed patterns, has simple logic, and can be completed in a template-like fashion. For example: writing standard registration, login, and profile update endpoints for a user management system. This type of work follows highly predictable patterns that AI can indeed complete efficiently.

But high-value coding remains crucial: system architecture design, data structure selection, time-space complexity tradeoffs, edge case handling—these decisions still require human developers. High-value coding involves consistency guarantees in distributed systems, performance optimization under high concurrency, modeling complex business rules, and other work requiring deep thinking and experienced judgment—far beyond the current capabilities of AI.

A Frontline Engineer's Real Experience: AI Is Fast but Not Always Right

Harshit Rayyan shared his daily work experience at Atlassian: he and his entire team use AI tools every day because AI genuinely accelerates the development process. But the reality is:

AI is fast, but AI is not always right.

Specifically:

• AI responses are sometimes incorrect
• AI sometimes uses wrong logic
• AI frequently misses edge cases
• AI sometimes misunderstands the problem itself

Regarding edge cases—this is a core concept in software engineering, referring to system behavior under extreme or atypical inputs. For example: when user input is empty, data volume is zero, concurrent requests arrive simultaneously, the network suddenly drops, or disk space runs out—how should the system respond? Most serious failures in production environments (such as the 2024 CrowdStrike incident that caused 8.5 million Windows devices to blue screen) often stem from uncovered edge cases. Because AI models lack deep understanding of real production environment complexity, they tend to generate only "happy path" code while missing these critical exception handling scenarios.

This is precisely why every time AI generates code, human developers must review it: Is the logic correct? Are edge cases covered? Could it crash in production?

If you don't understand coding, logic, or algorithms, how can you evaluate the quality of AI-generated code? This actually demonstrates that in the AI era, coding ability hasn't become less important—it's become more important. The role has simply shifted from "code writer" to "code reviewer and system designer."

The AI Bubble: The Massive Gap Between Investment and Returns

Reports from analysts like Derek Thompson reveal a staggering figure: AI spending in 2024 reached $500 billion, but consumers actually purchased only $12 billion in AI services. This means there is severe over-investment in the AI space.

An even more telling case: former OpenAI executive Mira Murati recently completed a $2 billion seed round at a $10 billion valuation—without even telling investors what product the team would build. Investors made their decision solely based on "she's from OpenAI."

This is a classic sign of a bubble: excessive investment begins, unclear returns follow. Every technology bubble in history has followed this pattern. The current AI investment frenzy shares many similarities with the 2000 dot-com bubble. During the dot-com bubble, the NASDAQ index surged from 1,000 points in 1995 to 5,048 in 2000, then crashed to 1,100. Back then, companies could secure massive funding simply by having a ".com" suffix—remarkably similar to how companies today can achieve high valuations just by having an "AI" label. But the key difference is: after the dot-com bubble burst, the internet itself didn't die—instead, it gave birth to giants like Google, Amazon, and Facebook. Similarly, even if the AI bubble bursts, AI technology will continue to develop and profoundly change software development. It's just that current valuations and expectations may need significant correction.

Practical Advice for Those Learning to Code

Don't Skip the Fundamentals

The most dangerous current trend is: students and beginners skip programming fundamentals and rely entirely on AI to solve all problems. The result is they struggle severely in interviews and actual work because they lack solid foundational skills.

Programming fundamentals include but are not limited to: data structures (arrays, linked lists, trees, graphs, hash tables), algorithmic thinking (recursion, dynamic programming, greedy algorithms), computer systems principles (memory management, process scheduling, network protocols), and software engineering practices (version control, test-driven development, code refactoring). This knowledge forms the cognitive framework necessary for understanding and evaluating any code—whether written by humans or generated by AI. Without this framework, a developer is like someone who doesn't understand structural mechanics trying to evaluate architectural blueprints—fundamentally unable to judge whether the structure is safe.

Now Is the Best Time to Learn Coding

This sounds counterintuitive, but the logic is clear: when large numbers of people drop out or hesitate due to confusion, opportunity emerges. Moreover, you now have AI as a learning assistant—you can ask AI any question directly, making learning more efficient than ever before.

Find Your Position

Future developers won't be "people who write code" but rather:

• People who can understand and evaluate AI output
• People who can make architecture and design decisions
• People who can handle complex problems AI cannot solve
• People who can use AI as a productivity multiplier

Conclusion

The value of programming won't disappear—only its form will change. Rather than fearing AI, embrace it as a tool while building solid programming fundamentals. When others are hesitating, that's precisely the best time for you to take action.

Key Takeaways

• When tech leaders say "coding is dead," they actually mean low-level coding is being replaced by AI—not that high-value programming is dying
• AI generates code quickly but isn't always correct; human developers are still needed to review logic, edge cases, and production stability
• There's a clear bubble in AI: $500 billion in investment yields only $12 billion in consumer spending
• Skipping programming fundamentals and relying entirely on AI is the most dangerous learning trend today
• When masses of people exit coding due to AI panic, it's actually the best time for learners to seize the opportunity