Should You Still Learn To Code? What The Evidence Says For Students, Developers And Managers
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Should You Still Learn To Code? What The Evidence Says For Students, Developers And Managers on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

AI now writes much of the code merged at Anthropic, but that figure comes from a company that builds and heavily uses AI coding tools. Research on productivity is mixed, entry-level hiring is tight, and one learning experiment found lower quiz scores among developers who used AI. The evidence supports learning to code, while putting more emphasis on specifying, reviewing, testing and debugging software.

AI is writing a growing share of software code, but the evidence available through November 2026 does not show that students or developers should stop learning to code. Company figures, hiring data and research experiments instead point to a changing skill mix: people still need to define software requirements and review, test and debug what AI produces.

Anthropic says more than 80% of the code merged into its codebase in May 2026 was authored by Claude, compared with a share it described as in the low single digits before its coding tool launched in February 2025. The company also reported that its typical engineer merged eight times as much code per day in the second quarter of 2026 as in 2024. Anthropic cautioned that lines of code measure quantity rather than quality, making the eightfold figure “almost certainly an overstatement” of the productivity gain.

Whether AI makes developers faster is less settled. In a 2025 randomized trial, the nonprofit METR assigned 16 experienced open-source developers 246 tasks in projects they knew well. Developers using AI tools available between February and June 2025 took 19% longer to complete the tasks, despite believing AI had made them faster. METR’s later work suggested developers may have gained more from newer tools, but the group said that evidence was weak and its estimates did not rule out no effect.

The labor market picture is also mixed. Indeed data cited in the source show 4.5% of US software development postings were entry-level in the first quarter of 2026, while 69.3% were senior. Yet US software development postings were about 23% higher year over year in November, after reaching a low in May 2025. The source also cites Stanford researchers’ estimate that employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it otherwise would have been; that finding does not, on its own, establish AI as the sole cause.

At a glance
analysisWhen: Evidence and figures reported through N…
The developmentA review of 2026 company data, hiring figures and research experiments finds AI is changing coding work, but does not establish that learning to code is no longer worthwhile.

Coding Skills Beyond Code Generation

For students and career switchers, the figures suggest a narrower entry-level hiring path, not proof that programming knowledge has become unnecessary. When employers expect smaller teams or AI-assisted output, candidates may need to show they can do more than produce a working first draft: they must explain what a system should do, detect errors and understand how a change affects an existing codebase.

For working developers, AI can alter the balance between typing and directing work. Anthropic says its engineers use Claude to write code while engineers direct and review it. That is evidence about the company’s workflow, not a universal productivity benchmark. The METR trial also warns against assuming that faster code production means faster completion of real work.

Learning method matters, too. In an Anthropic experiment, mostly junior developers using AI to learn a new library scored 50% on a follow-up quiz, compared with 67% among those who coded by hand; the largest gap was in debugging. The result comes from one experiment, but it gives learners and educators a reason to check whether AI assistance is replacing the practice needed to build independent understanding.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

From AI Predictions to Measured Work

On March 10, 2025, Anthropic chief executive Dario Amodei told the Council on Foreign Relations that AI would be writing 90% of code within three to six months and essentially all code within a year. Anthropic’s later internal figure shows substantial AI use at the company, though it does not validate that broader forecast across the software industry.

Other public estimates differ in scope and date. Google chief executive Sundar Pichai said in October 2024 that AI generated more than a quarter of new code at Google, which engineers then reviewed and accepted. METR cited an estimate in February 2026 that Claude Code authored about 4% of GitHub commits. These figures should not be treated as directly comparable: they cover different organizations, tools, periods and measures.

Productivity studies also test different kinds of work. A 2022 experiment with 95 professional developers found that those using GitHub Copilot completed a small JavaScript HTTP-server task 55.8% faster on average. METR’s later trial involved real tasks in large codebases familiar to experienced developers. The contrast suggests results can depend on task type and setting; neither experiment settles how AI affects every developer’s work.

“The programmer still needs to specify what is being built, how it fits with other code and whether this is a secure design or an insecure design.”

— Dario Amodei, Anthropic chief executive, speaking at the Council on Foreign Relations on March 10, 2025

What the Studies Cannot Settle

There is no single, industry-wide measure of how much code AI writes or how much time it saves. Anthropic’s figures describe its own workforce and tools; the company has a strong reason to use its products, and its reported code volume does not directly measure quality or total project time.

METR’s trial was small and focused on experienced developers working in familiar open-source projects. Its later study faced a different problem: some developers said they were holding back tasks they did not want to do without AI. METR reported raw estimates of tasks taking 18% less time for developers from the first study and 4% less for newly recruited developers, but confidence intervals included no effect. It said the results offered only weak evidence about the size of any change.

The employment figures also cannot establish how much of the change came from AI rather than other economic or industry conditions. The source does not provide enough detail to determine whether the rise in postings continued after November 2026, or whether the learning experiment’s quiz results predict longer-term skill or job performance.

What Students and Employers Should Track

Students can keep learning programming while using AI in ways that preserve active practice: attempt problems independently, read generated code, test its behavior, debug failures and explain design choices. The available evidence supports those as relevant skills; it does not establish one ideal curriculum or prove that every learner should use the same balance of manual and AI-assisted work.

Developers and managers can track project outcomes rather than relying on lines of code or self-reported speed alone. Useful measures include whether software works, how much review and rework it requires, and how long tasks take from specification to delivery. Future studies with larger samples and varied teams could clarify how results differ by experience, task and tool.

For now, the evidence supports a qualified answer: yes, learning to code remains worthwhile, but typing syntax is only part of the work. The next useful evidence will come from longer-term research on learning and hiring, alongside transparent measures of software quality and productivity across organizations.

Key Questions

Should students still learn to code?

Yes, the evidence does not show that coding education is obsolete. It suggests students should also learn to specify requirements, read and test code, debug problems and judge design decisions.

Does AI make software developers more productive?

The answer depends on the task and evidence used. METR’s 2025 trial found experienced developers took 19% longer with the AI tools tested, while Anthropic reported a large rise in code merged per engineer but cautioned that this was not a direct measure of productivity.

Is it harder for new developers to get hired?

Indeed data cited in the source show entry-level roles made up 4.5% of US software development postings in the first quarter of 2026, compared with 69.3% for senior roles. Those figures describe postings, not the full hiring market, and do not show that AI alone caused the imbalance.

Can using AI make it harder to learn programming?

One Anthropic experiment found mostly junior developers who used AI while learning a new library scored 50% on a follow-up quiz, versus 67% for those who coded by hand. That single result does not settle the question for all learners, but it suggests students should check that assistance is not replacing practice.

Source: ThorstenMeyerAI.com

You May Also Like

Google Surges In Global Coverage

Google’s mentions in global coverage have surged, with GDELT reporting 130 mentions in a recent window, marking a 5.5-fold increase. The development signals broader media attention.

Android Show 2026: all the news and announcements

Google’s Android Show 2026 revealed new features for Android 17, Gemini, Android Auto, and upcoming devices like Googlebook laptops, ahead of I/O 2026.

Using Nemotron-Personas-Finance-Japan To Explore Demographics And Finance

Elith and NVIDIA say their 9-million-persona dataset supports financial AI development using synthetic Japanese demographic and finance profiles.

N-able Pulls N-central Offline After 30+ “Zero-Days” Go Public. What To Ask Your IT Provider Today

N-able took hosted N-central offline and advised on-premises customers to do the same after a researcher published more than 30 alleged flaws.