AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Guide To Coding Skills For Students, Developers And Managers In The AI Era on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys 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 authors a large share of code at some companies, but evidence that it consistently makes developers faster remains mixed. Hiring data point to fewer entry-level software postings, while research suggests students who rely on AI while learning may retain less. The practical implication is to build coding fundamentals alongside skills in specification, testing, debugging and code review.

AI is writing a growing share of software code, but the evidence does not show that learning to program has become unnecessary. Data and studies reviewed by ThorstenMeyerAI.com point to a shifting skill mix: students and developers still need coding fundamentals, while specification, testing, debugging and review are increasingly important as AI tools generate more code.

Anthropic reported that, in May 2026, Claude authored more than 80% of the code merged into the company’s codebase, compared with a share in the low single digits before Claude Code launched in February 2025. The company also said 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 and that the eightfold figure is “almost certainly an overstatement” of productivity gains.

Evidence on whether AI makes developers faster is mixed. In a randomized METR trial conducted in early 2025, 16 experienced open-source developers completed real tasks in familiar projects. They took 19% longer when allowed to use the AI tools tested, despite expecting the tools to reduce their time. METR said in February 2026 that newer results suggested possible speed gains, but offered “only very weak evidence” about their size.

Hiring figures show a difficult market for newcomers, though not an end to software demand. Indeed data cited in the report show that 4.5% of US software development postings were entry-level in the first quarter of 2026, while 69.3% were senior. The same source reported that US software development postings were about 23% higher year over year in November 2026, after reaching a low in May 2025.

At a glance
reportWhen: Published November 18, 2026; evidence s…
The developmentA review of 2026 company data, hiring figures and studies finds AI is changing software development, but does not establish that learning to code is no longer useful.

The Skills Behind AI-Written Code

The findings matter to students, career switchers and employers deciding what to teach, learn or hire for. If AI handles more typing, people still need to define the task, recognize when generated code is wrong, and judge how it fits into a larger system. Those responsibilities are central to making software reliable, not merely producing code quickly.

Evidence on learning also cautions against treating AI assistance as a substitute for practice. In an Anthropic experiment, mostly junior developers who used AI while learning 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. That result does not prove AI use always harms learning, but it suggests that delegating too much of the work can leave learners less prepared to understand and fix code.

For managers, the available data argue against equating more code with more productivity. Anthropic itself says its code-volume measure overstates the true gain. Hiring and training decisions may need to account for whether candidates can reason about system design, test outputs and maintain software, not just generate code with a tool.

Products Worth Considering

Dell XPS 13 13.4-Inch Touchscreen Laptop, Intel Core 5 320, 16GB RAM, 512GB SSD

Dell XPS 13 13.4-Inch Touchscreen Laptop, Intel Core 5 320, 16GB RAM, 512GB SSD

As an affiliate, we earn on qualifying purchases.

From Code Generation to Review

The debate follows high-profile predictions about automation. On March 10, 2025, Anthropic chief executive Dario Amodei told the Council on Foreign Relations that AI would write 90% of code within three to six months and essentially all of it within 12 months. Anthropic’s later internal figures show extensive use at its own company, though they do not establish that the same share applies across the software industry.

Other published estimates differ in date and scope. 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 measure different organizations and activities, so they should not be treated as a single industry-wide rate.

Productivity studies also depend on the work being tested. A 2022 experiment involving 95 professional developers found participants using GitHub Copilot completed a small JavaScript HTTP-server task 55.8% faster. METR’s later trial instead tested experienced developers on real tasks in codebases they already knew. The contrasting results underline that a short, self-contained coding exercise may not predict performance on complex maintenance 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

How Far the Findings Reach

Several figures come from companies describing their own tools and workflows. Anthropic is both a developer of AI coding tools and an unusually intensive user, so its internal code share may not represent typical workplaces. Its code-volume figures also do not show whether the code is more useful, secure or maintainable.

Research on speed remains unsettled. METR’s early-2026 study was affected by developers’ reluctance to take tasks without AI, and its reported estimates had confidence intervals that included no effect. It is not yet clear how much current tools improve results across different experience levels, programming languages and types of work. The hiring statistics also describe postings and employment patterns, not the full range of software jobs or the cause of changes in entry-level opportunities.

What Training and Hiring Should Track

The next useful evidence will come from larger, repeated studies that measure completed work over time, including quality, security, maintenance costs and how much developers learn. Employers and researchers will also need to separate the impact of AI tools from changes in hiring demand and broader economic conditions.

For now, the report’s evidence supports a measured approach: learn to write and read code, use AI as an assistant rather than a replacement for understanding, and practice testing and debugging generated work. Managers can assess whether teams can explain and maintain what they ship. How these skills translate into long-term hiring demand remains uncertain, and the available figures do not settle whether entry-level opportunities will recover.

Key Questions

Is learning to code still worthwhile?

The evidence reviewed does not establish that coding skills are obsolete. It points to continuing value in understanding code, defining requirements, testing outputs and debugging problems, even when AI generates code.

Does AI reliably make software developers faster?

No single result establishes that. METR’s early-2025 trial found experienced developers took 19% longer with the AI tools tested, while a 2022 Copilot experiment found faster completion on a small coding task. The results differ by task and study design.

What happened to entry-level software job postings?

Indeed data cited in the report show that entry-level roles accounted for 4.5% of US software development postings in the first quarter of 2026. The figures describe postings during that period; they do not establish why the share was low or predict future hiring.

Can using AI interfere with learning programming?

An Anthropic experiment found mostly junior developers who used AI to learn a new library scored 50% on a follow-up quiz, compared with 67% for those who coded by hand. This is one experiment, not proof that every form of AI assistance reduces learning.

What skills should developers and managers prioritize?

The cited evidence points to coding fundamentals alongside clear task specification, code reading, testing, debugging, security judgment and review. Managers should also look beyond code volume when evaluating productivity and software quality.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Are Three AI Models Enough To Capture The Complexity Of The Real World?

Exploring whether a limited number of AI models can accurately interpret complex real-world events and the risks of interpretive homogeneity.

SenseTime SenseNova U1.5: Building Better AI With 8B-MoT And Open Access

SenseTime unveils SenseNova U1.5, an 8B-parameter unified vision-language model with open training code, marking a key move in open multimodal AI development.

How AI Is Enhancing The Top Drawing Tablets Of 2026

Artificial intelligence enhances top drawing tablets of 2026, improving precision, usability, and creative workflows for artists and designers.

Runway Opens Praxis-1 Amid OpenAI’s Robotics Team Shake-Up

OpenAI has moved its robotics teams into other divisions, while Runway released an open-weight robot model built for faster control.