Should Students, Developers And Managers Keep Learning To Code?
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

AI is producing a growing share of code at some companies, while evidence on productivity and effects on early-career jobs remains mixed. The available research supports learning how to specify, read, test, debug and assess code, even when AI writes the first draft.

AI is writing a growing share of code at some companies, but available evidence does not show that students, developers or managers can stop learning coding skills. Anthropic said Claude authored more than 80% of code merged into its own codebase in May 2026; research and job data point to continued demand for technical judgment, while leaving the scale of AI’s productivity and employment effects uncertain.

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.

Anthropic’s figure is a company-reported measure, not an industry-wide estimate. The company said Claude authored more than 80% of its merged code in May 2026, compared with what it described as the low single digits before Claude Code launched in February 2025. Anthropic said engineers directed and reviewed the AI-generated work. It also reported that its typical engineer merged eight times as much code per day in the second quarter of 2026 as in 2024, while cautioning that lines of code measure quantity rather than quality and that the increase almost certainly overstates the true productivity gain.

Independent evidence gives a less settled picture of speed. In a randomized METR trial conducted in early 2025, 16 experienced developers took 19% longer to complete tasks with AI tools, despite estimating that the tools had made them 20% faster. METR later said it believed speed-ups were likely larger by early 2026, but described its newer data as only very weak evidence about their size. The source also reports an Anthropic learning experiment in which mostly junior developers who used 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 involved debugging.

Labor-market measures are mixed as well. Indeed data cited in the source show that 4.5% of US software development job postings were entry-level in the first quarter of 2026, while 69.3% were senior. Stanford researchers found employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it would otherwise have been expected to be, according to the source. At the same time, US software development postings on Indeed were about 23% higher year over year in November 2026, after reaching a low in May 2025. These indicators describe different measures and do not by themselves establish that AI caused the changes.

At a glance
reportWhen: Evidence and figures reported through N…
The developmentA review of 2026 job data, developer surveys and coding experiments finds that AI is changing coding work, but does not establish that learning to code is no longer useful.
Bar chart titled Developers felt faster with AI. They were slower. METR trial in early 2025 with 16 experienced open-source developers and 246 real tasks: before the study they expected AI to cut their time per task by 24%; afterwards they believed it had cut it by 20%; measured, tasks took 19% more time with AI allowed. A highlighted note: plus 19% time; METR thinks tools sped people up more by early 2026 but cannot yet say by how much
Expectation, belief and measurement in METR's developer trial. Sources: Becker et al. (METR), arXiv:2507.09089, p. 1; METR, February 24, 2026. Chart: Thorsten Meyer AI.

Coding Skills Move Beyond Typing

For students and career changers, the findings argue for learning how software works, not just how to produce lines of code. AI can draft an implementation, but someone still has to describe the task, judge whether the output fits existing systems, test behavior, find defects and assess security. Those skills also help users spot when an AI answer is plausible but wrong.

For working developers, AI may shift time from writing code toward reviewing and maintaining it. A developer survey cited in the source found 66% named “almost right, but not quite” as their biggest AI frustration. Managers face a related challenge: if fewer postings are entry-level, the traditional route for new workers to gain experience may narrow. The figures do not prove that AI caused this pattern, but they make training and supervised practice relevant questions for hiring plans.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

From Code Generation to Review

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 of it within 12 months. Those were predictions, not measured outcomes. Anthropic’s later figures describe extensive use inside its own company, with engineers directing and reviewing the generated code.

Other figures cited in the source suggest adoption beyond Anthropic, but they are not directly comparable. 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 about 4% of GitHub commits were authored by Claude Code. Differences in date, company and measurement mean these numbers do not establish a single industry-wide share.

“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

Productivity and Hiring Remain Unsettled

The available figures do not establish how much AI raises software productivity across the industry. Anthropic’s internal output measures come from a company that builds AI coding tools, and it warns that lines of code are an imperfect proxy. METR’s controlled trial found slower performance in its early 2025 setting, while its later evidence on speed-ups was weak. Results may differ as tools and tasks change.

It is also unclear how much AI explains shifts in entry-level postings or employment among young workers. The reported labor-market measures show patterns, not proof of cause. The source does not provide enough detail to assess how these figures vary by field, employer or region, or what share of affected workers moved into other roles.

Watch Hiring and Skill Requirements

The next useful evidence will come from comparable, repeated measures of task completion, code quality and hiring as AI tools evolve. Students can watch how employers describe entry-level roles and the skills they require. Developers and managers can track whether AI changes review workloads, training practices and the number of junior opportunities.

For now, the evidence supports learning to code alongside learning to work with AI: specify requirements, read generated code, test it, debug failures and evaluate design choices. How much of the day-to-day work shifts to AI, and whether that creates or removes jobs overall, remains an open question.

Key Questions

Does the evidence show that students should stop learning to code?

No. The reported evidence shows AI generating code in some settings, but it does not establish that coding knowledge has become unnecessary. Skills in specifying tasks, reviewing code, testing and debugging remain relevant to evaluating AI-generated work.

Is AI making software developers more productive?

The evidence is mixed. A METR trial in early 2025 found experienced developers took 19% longer with AI on the tested tasks. Anthropic reported much higher code output internally but cautioned that its lines-of-code measure likely overstates productivity gains.

What happened to entry-level software job postings?

According to Indeed data cited in the source, 4.5% of US software development postings were entry-level in the first quarter of 2026, compared with 69.3% that were senior. The figures do not show on their own why the distribution changed.

Can using AI to learn programming affect learning?

In an Anthropic experiment described in the source, mostly junior developers who used AI while learning a new library scored 50% on a follow-up quiz, compared with 67% for those who coded by hand. The largest reported gap was in debugging; the result comes from that experiment and should not be treated as a universal outcome.

Source: Thorsten Meyer AI

You May Also Like

TypeScript Generics: Why They’re Less Scary Than You Think

A deeper look at TypeScript generics reveals how their simplicity can transform your coding experience—discover why they’re easier than you think.

Exapunks (2018)

A 2018 game, Exapunks, is cited in recent cybersecurity discussions, highlighting its impact on understanding hacking culture and digital threats.

Rust vs. Go: Which Language Gives You the Bigger Salary Bump in 2025?

Just as the tech world evolves, discovering which language—Rust or Go—will boost your salary in 2025 depends on industry trends and your career goals.

Why Developers Need Better Debugging Habits

Great debugging habits can transform your coding process, but the key to mastering them lies in understanding what truly makes a difference.