Is Coding Still Worth Learning? Evidence For Students, Developers And Managers
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

AI is writing a large share of code at some firms, but the evidence does not show that coding skills are obsolete. Hiring data, productivity studies and a learning experiment point to a changing job market where specifying, reviewing, testing and debugging code remain important.

AI is writing much of the code at some companies, but 2026 data and studies do not establish that learning to code is no longer worthwhile. Anthropic says Claude authored more than 80% of code merged into its codebase in May; meanwhile, job-posting figures show a small share of entry-level software roles, while research on AI’s effect on developer speed remains mixed.

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

Anthropic’s own figures show how quickly AI has entered its software work. The company says Claude authored more than 80% of code it merged in May 2026, compared with a share in the low single digits before Claude Code launched in February 2025. It also says the typical engineer merged eight times as much code per day in the second quarter of 2026 as in 2024. Anthropic cautions that lines of code measure quantity rather than quality and that the eightfold figure is almost certainly an overstatement of productivity gains.

Evidence about speed is less settled. In a randomized METR trial conducted in early 2025, experienced developers took 19% longer with AI tools, although they believed the tools had made them 20% faster. METR later said newer evidence suggested the speed-up may have grown by early 2026, but was only very weak evidence of its size. Anthropic also reports an experiment in which mostly junior developers learning a new library scored 50% on a follow-up quiz after using AI, compared with 67% among developers who coded by hand.

The US entry-level market appears tight, though software job postings overall have shown signs of recovery. Indeed data cited in the source put entry-level roles at 4.5% of software development postings in the first quarter of 2026, against 69.3% for senior roles. The source also reports that postings were up about 23% year over year in November, after reaching a low in May 2025. These figures describe postings, not hires or the number of people employed.

At a glance
analysisWhen: Evidence and figures reported through N…
The developmentA collection of 2026 company data, hiring figures and research suggests AI is changing coding work while leaving its productivity benefits and effects on early-career jobs uncertain.
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 Beyond AI Drafts

The findings point to a change in what coding work involves. If an AI system produces a draft, people still need to describe the desired result, fit the code into an existing system, check whether it works and judge whether it is safe to ship. Those tasks depend on understanding code, even when a person types fewer lines.

For students and career switchers, the hiring figures suggest that getting an initial role may be difficult, and the learning experiment raises a separate concern: using AI to complete exercises can reduce what learners retain. For working developers, being able to read, test, debug and review generated code may matter alongside writing it. Managers face a related question: whether teams have enough skill to verify AI output and take responsibility for decisions about its use.

The evidence does not settle whether these skills will translate into stable job growth or higher productivity across the industry. It does show why raw output counts alone cannot answer whether AI has made programming knowledge unnecessary.

Products Worth Considering

DJI Mini 4K Fly More Combo Drone with 4K Camera and 3 Batteries

DJI Mini 4K Fly More Combo Drone with 4K Camera and 3 Batteries

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

As an affiliate, we earn on qualifying purchases.

From Code Generation to Review

In March 2025, Anthropic chief executive Dario Amodei predicted at the Council on Foreign Relations that AI would write 90% of code within three to six months and essentially all of it within a year. The figures now available from Anthropic show substantial adoption at his own company, though they do not establish that the prediction came true across the software industry.

Other figures in the source are lower and refer to different measures and dates. 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. Company code shares, GitHub commits and code merged per engineer measure different things, so they should not be treated as direct comparisons.

Anthropic’s report describes engineers directing and reviewing Claude’s work. That distinction helps explain why code generation does not by itself measure the time saved, the quality of the result or the amount of human work remaining.

“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

Limits of the Available Evidence

It remains unclear how much AI improves software productivity across employers and types of work. Anthropic’s figures come from a company that builds AI coding tools and uses them heavily; its measures count code volume, and the company itself warns that this can overstate gains. METR’s early controlled trial measured a specific group of developers and tasks, while its later assessment described the new evidence as weak.

The job figures do not establish that AI caused the low share of entry-level postings or the reported employment decline among 22- to 25-year-olds in AI-exposed jobs. The source says Stanford researchers found employment in those jobs was 19% below where it would have been, but does not provide enough detail here to establish the causes or the full comparison method. Posting counts also cannot show how many positions were filled or how many applicants competed for them.

It is also uncertain how well the learning experiment applies to longer courses, different AI tools or experienced programmers. The available material does not establish which coding skills employers will prioritize, or how quickly hiring patterns may change.

What New Hiring Data Shows

Further hiring data will help show whether the reported recovery in software development postings continues and whether entry-level opportunities grow alongside it. The figures cited here cover postings through November 2026, but do not establish what employers will do next.

Researchers and companies will also need to publish clearer, comparable evidence on time saved, code quality, debugging and the effects of AI-assisted learning. Until then, students, developers and managers can use the existing findings as evidence of a changing workflow, not a settled verdict on the future value of coding skills.

Key Questions

Does Anthropic’s 80% figure mean AI writes 80% of all code?

No. Anthropic says Claude authored more than 80% of code merged into Anthropic’s own codebase in May 2026. The figure does not measure all software companies or all code written.

Has AI been proven to make developers faster?

No. In METR’s early 2025 randomized trial, experienced developers took 19% longer with AI, though they thought they were faster. METR said later evidence may indicate a speed-up by early 2026, but was very weak evidence of its size.

Are entry-level software jobs disappearing?

The source reports that entry-level roles made up 4.5% of US software development postings in the first quarter of 2026. That shows a limited share of postings, but does not by itself establish why the share was low or that such jobs are disappearing.

What coding skills does the evidence point to?

The reported findings highlight the ability to specify what software should do, read and review code, test and debug it, and assess design choices, including security. The evidence does not establish a definitive list of skills employers will require.

Source: Thorsten Meyer AI

You May Also Like

Claude Code’s Auto Mode Will Be On By Default, Anthropic Confirms – Yahoo Tech

Anthropic has announced that Claude Code’s auto mode will be enabled by default, though rollout details and user controls remain unspecified.

OpenAI’s AI Findings Challenge A Hypothesis—But What’s Established?

OpenAI says an AI system found an algorithm contradicting ETH. Lean checked a weaker bound, while independent expert review is still pending.

How Claude Frontier Academy Plans To Train 10,000 Engineers With $100M

A headline links Claude Frontier Academy to $100 million and 10,000 engineers, but program details and the figures’ meaning remain unconfirmed.

Genf Real Estate Trends In Focus: Signal Monitoring Insights

Analysis of recent real estate signals in Geneva shows Solvalor 61’s purchase of a residential property, highlighting market trends and rapid development monitoring.