🔍 Read the full analysis: Is Learning To Program Still Relevant? What The Evidence Says on ThorstenMeyerAI.com
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
TL;DR
AI is generating a growing share of software code, but the available evidence does not show that programming skills are obsolete or that AI consistently makes developers faster. Entry-level hiring is tight, while experiments suggest learners can miss important programming skills when they rely on AI without careful practice.
AI is writing a substantial share of code at some companies, but current evidence does not show that learning programming has become pointless. Data cited in a November 2026 review by ThorstenMeyerAI.com show mixed productivity results, fewer entry-level software postings and a risk that learners using AI may retain less of what they study.
Anthropic reported that, in May 2026, Claude authored more than 80% of the code merged into the company’s codebase, compared with a share it described as “in the low single digits” before its coding tool Claude Code launched in February 2025. The company said its typical engineer merged eight times as much code per day in the second quarter of 2026 as in 2024. It cautioned that lines of code measure quantity rather than quality and that the eightfold figure was “almost certainly an overstatement” of productivity gains.
Whether AI makes developers faster is less settled. In a randomized METR trial involving 16 experienced open-source developers and 246 tasks in projects they already knew, developers took 19% longer when allowed to use AI tools available from February to June 2025. They had expected AI to reduce their time by 24% and, afterwards, believed it had reduced it by 20%. METR said a later study produced only “very weak evidence” about the size of any subsequent speed-up.
The labor market shows both pressure and continuing demand. Indeed data cited in the review put entry-level roles at 4.5% of US software-development postings in the first quarter of 2026, while senior positions made up 69.3%. Stanford researchers found employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it would otherwise have been. Meanwhile, Indeed’s US software-development postings were about 23% higher year over year in November 2026, after reaching a low in May 2025.
Skills Matter Beyond Code Generation
For students and career changers, the evidence points to a tougher first step into the profession, not the disappearance of programming as a skill. If employers expect fewer junior hires to produce code by hand, new workers may need to show they can understand existing systems, specify requirements and verify AI-generated work. A tool that can generate code does not, by itself, establish that the result meets a user’s needs or is safe to deploy.
The learning evidence also matters. 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 involved debugging. That result does not establish that AI always impairs learning, but it suggests that getting a working answer quickly may not build the same understanding as attempting, testing and correcting code.
For employers, the question is not only how much code a tool produces. Review, testing and security decisions remain part of the work described by Anthropic chief executive Dario Amodei: programmers still need to specify what is being built, how it fits with other code and whether the design is secure. Those tasks depend on enough technical knowledge to judge the output.
Products Worth Considering

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

Samsung 27-Inch Odyssey G5 QHD Gaming Monitor (G50F)

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black
As an affiliate, we earn on qualifying purchases.
Different Tasks, Different Results
The studies do not measure one identical kind of programming. A 2022 experiment with 95 professional developers found that participants using GitHub Copilot completed a small task—writing an HTTP server in JavaScript—55.8% faster on average than participants without it. Developers with less programming experience benefited most. That result differs from METR’s trial, which involved real tasks in large codebases familiar to experienced contributors. The findings can coexist: AI may help with a discrete, well-defined task while adding overhead in a complex project.
Company disclosures also require care. Anthropic is both a developer of AI coding tools and a heavy user of them, so its internal figures may not describe a typical employer. 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 cover different organizations, periods and measures; they should not be treated as directly comparable.
Amodei made his prediction about AI writing code at a Council on Foreign Relations event on March 10, 2025, saying it could write 90% of code within three to six months and essentially all of it within twelve months. The later Anthropic disclosure records substantial use inside his company, but it does not establish that the prediction applied across the software industry or that AI independently handles all programming work.
““More than 80% of the code we merge into Anthropic’s codebase was authored by Claude.””
— Anthropic
Productivity and Hiring Remain Unsettled
How much AI improves software productivity in everyday work remains uncertain. METR’s 2025 randomized trial found slower completion, while a later study produced estimates pointing toward time savings but with confidence intervals that included no effect. METR said developers increasingly resisted tasks without AI, complicating the later comparison. Tools have also changed, so results from early 2025 may not represent tools used in late 2026.
Labor figures do not prove that AI caused the drop in employment among younger workers or the small share of entry-level postings. The figures describe conditions and patterns, not a single cause. It is also unclear whether the recent rise in software-development postings will translate into more junior opportunities, or how broadly Anthropic’s internal code-production figures apply to other workplaces.
The learning experiment was limited to a particular task and a group made up mostly of junior developers. It raises a concern about learning through AI, but does not show that every learner, subject or way of using these tools produces the same result.
Watch Hiring and Learning Evidence
The next useful evidence will come from longer-running studies of real software teams that measure quality, security, maintenance and delivery time—not just the volume of generated code. METR’s account of its later trial points to a measurement problem: researchers need methods that can compare AI-assisted work when developers may not want to work without AI.
Job-posting data will show whether the recent increase in US software-development listings continues and whether employers begin advertising more junior positions. For learners, the key test is whether AI-assisted study produces durable skills: can people explain, debug, test and safely modify code after the assistant is removed? Until those results are clearer, the evidence supports learning programming while treating AI as a tool whose output must be understood and checked.
Key Questions
Is learning to program still worthwhile?
Current evidence does not show that programming skills are obsolete. It suggests that code generation is changing, while specifying requirements, reviewing output, testing, debugging and security judgment remain relevant.
Does AI make software developers faster?
The evidence is mixed. In METR’s early-2025 randomized trial, experienced developers took 19% longer with AI on tasks in familiar codebases. A later study offered only weak evidence about the size of any speed-up.
Are entry-level programming jobs disappearing?
Indeed data cited in the November 2026 review show that entry-level roles accounted for 4.5% of US software-development postings in the first quarter of 2026. That is evidence of a narrow entry-level share, not proof that such jobs have disappeared or that AI alone caused the pattern.
Can using AI make it harder to learn coding?
It may, depending on how it is used. In one Anthropic experiment, mostly junior developers who used AI while learning a library scored 50% on a follow-up quiz, versus 67% for those who coded by hand. The study does not establish that every form of AI-assisted learning has the same effect.
Source: ThorstenMeyerAI.com
