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

🔍 Read the full analysis: Is Learning To Code Still Worth It In The Age Of AI? on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

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

TL;DR

AI is writing a large share of code at some companies, but evidence does not show that software development has become a task-free profession. Job postings, productivity studies and a learning experiment suggest that coding remains useful, while the value is shifting toward specifying, testing, debugging and reviewing software.

AI is writing much of the code at some technology companies, but available 2026 evidence does not show that learning to program has become pointless. Anthropic says Claude authored more than 80% of code merged into its own codebase in May, while job data, developer experiments and a study of AI-assisted learning point to a more complicated shift: less emphasis on typing code, and more on directing, testing and judging it.

Anthropic reported that the share of code merged into its codebase that was authored by Claude rose from the low single digits before February 2025 to more than 80% in May 2026. It also said its typical engineer merged eight times as much code per day in the second quarter of 2026 as in 2024. The company cautioned that lines of code measure quantity rather than quality, making that eightfold figure “almost certainly an overstatement” of productivity gains. Anthropic builds AI coding tools, so its experience may not represent other employers.

Controlled evidence on speed is mixed. In a METR randomized trial, 16 experienced developers completed 246 tasks in familiar open-source projects. With AI tools available, they took 19% longer, despite believing afterward that AI had made them 20% faster. METR said in February 2026 that a later study produced only weak evidence about the size of any improvement: its estimates included no effect, and researchers said developers were increasingly reluctant to take tasks without AI.

Hiring data point to pressure on new entrants, not an end to demand. Indeed data show that 4.5% of US software-development postings were entry-level in the first quarter of 2026, compared with 69.3% classified as senior. Stanford researchers found employment among 22- to 25-year-olds in AI-exposed jobs 19% below where it would otherwise have been. At the same time, Indeed’s 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: Evidence and figures reported through N…
The developmentA review of 2026 job-market data, controlled studies and company disclosures finds that AI adoption is reshaping coding work, while evidence on productivity and entry-level prospects remains mixed.

The Skills Behind AI-Written Code

The evidence matters to people choosing what to study, developers deciding how to adapt and employers setting training plans. If AI handles more routine implementation, the job may put greater weight on defining requirements, understanding a codebase, checking outputs and spotting security or reliability problems. Those skills require enough programming knowledge to recognize when an answer is wrong, incomplete or unsafe.

The entry-level figures also suggest that the route into software work may be changing. When a smaller share of postings is explicitly entry-level, new developers may face more competition for the roles that provide early experience. But those figures do not establish that AI caused the change, and the recovery in total postings complicates any claim that demand for software developers has disappeared.

Learning methods matter as much as tool use. 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 suggests a potential trade-off: AI can help produce an answer while reducing the practice needed to understand how the answer works. It is one experiment, not proof that all AI-assisted learning is less effective.

Products Worth Considering

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity

Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity

LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDR True Black 400

LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDR True Black 400

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 company’s later disclosure that Claude authored more than 80% of its merged code shows substantial adoption inside Anthropic, though it does not establish that his broader timeline or prediction applies across the industry.

Other published figures offer a different scale. 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. In February 2026, METR cited an estimate that Claude Code authored about 4% of GitHub commits. These measures cover different organizations and periods, so they should not be treated as a direct comparison.

Study design also helps explain why productivity findings diverge. A 2022 experiment with 95 professional developers found participants using GitHub Copilot completed a small task—writing a JavaScript HTTP server—55.8% faster on average. METR’s later trial involved real work inside large projects developers already knew. A short, self-contained coding task and a task constrained by an established codebase test different kinds of work.

“More than 80% of the code we merge into Anthropic’s codebase was authored by Claude.”

— Anthropic, in its report on coding activity

How Far the Job Shift Reaches

The scale of productivity gains across the industry remains unsettled. Anthropic’s figures come from a company that develops coding AI, and its own report warns that output volume can overstate gains. METR’s randomized trial measured slower completion with AI for its participants, while its later estimates were too uncertain to establish a clear effect. The results are tied to particular tools, tasks and time periods.

The hiring figures also do not prove that AI is responsible for fewer entry-level opportunities. They show the composition of postings and an employment difference among younger workers in AI-exposed occupations; they do not, by themselves, isolate AI from other economic or hiring factors. It is also unclear whether the rise in overall software-development postings will translate into more junior positions.

Finally, the learning experiment tested a particular library-learning exercise and group of mostly junior developers. It does not establish that every use of an AI tutor weakens learning, or that AI cannot support learning when learners actively explain, test and revise generated code.

What Learners and Employers Track

The next useful evidence will come from studies that measure completed work, quality, security and learning over time—not just lines of code or workers’ impressions of speed. More results across different experience levels and project types could clarify when AI helps, when it slows developers down and whether those effects have changed as tools have improved.

For students and career switchers, the current evidence supports learning programming fundamentals alongside practical AI use. Practice reading code, writing tests, debugging failures and explaining design choices, rather than relying on generated code as a substitute for understanding. Working developers can apply the same approach by reviewing AI-generated changes and checking them against project requirements.

Employers and job seekers will also be watching whether the recovery in software-development postings leads to more entry-level hiring. Until that becomes clearer, the defensible conclusion is narrower than either “coding is over” or “AI changes nothing”: code generation is increasingly automated in some settings, while the broader effects on productivity, training and hiring are still developing.

Key Questions

Is learning to code still worth it?

Yes, the evidence supports learning programming, but not treating typing code as the only skill. Understanding software makes it possible to specify tasks, test AI-generated work, debug problems and assess whether a design meets requirements.

Does AI make software developers faster?

It depends on the task and study. METR’s 2025 trial found experienced developers took 19% longer with AI on familiar project work, while a 2022 Copilot experiment found participants completed a small coding task 55.8% faster. METR said its later evidence on speed gains was weak.

Are entry-level software jobs disappearing?

Indeed data show entry-level roles made up 4.5% of US software-development postings in the first quarter of 2026, but that figure does not prove AI caused the decline. Total postings were about 23% higher year over year in November 2026, and it remains unclear how much of that demand reaches new graduates.

Can AI help people learn programming?

It can provide assistance, but one Anthropic experiment found mostly junior developers who used AI while learning a new library scored lower on a follow-up quiz than those who coded by hand. The result does not settle how other forms of AI-assisted instruction affect learning.

Source: ThorstenMeyerAI.com

You May Also Like

Why Intelligent Tutoring Systems Are Essential for Personalized Learning

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

How to Choose AI-Powered Note-Taking Apps

Learn how to set up and maximize AI-powered note-taking apps for efficient, smart note management. Step-by-step guide for all levels.

Revolutionize Your Learning With ChatGPT Work And Codex AI

OpenAI releases new guidance for ChatGPT Work and Codex, enabling multi-step research, course development, and software projects in educational settings.

AI Tutors for Every Student: Personalized Learning at Scale

Great innovations like AI tutors are transforming education, but understanding their full potential can reveal how they might reshape learning for every student.