🔍 Read the full analysis: What Learning To Code Offers In An AI-Driven Workplace 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 producing a growing share of software code, but the available evidence does not show that learning to code has become unnecessary. It points instead to a shift in the work: developers need to direct AI tools and assess, test and debug their output, while the effects on productivity and entry-level jobs remain unsettled.
AI is writing a growing share of software code, but evidence gathered through November 2026 does not show that learning to code is now a waste of time. Instead, company figures, controlled experiments and employment data point to a changed skill mix: developers increasingly need to specify tasks, review generated code and test or debug it, even as productivity gains and the impact on entry-level jobs remain uncertain.
Anthropic reported that, in May 2026, Claude authored more than 80% of the code the company merged into its codebase, compared with a share it described as being 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. It cautioned that lines of code measure quantity rather than quality and that the eightfold figure is almost certainly an overstatement of the productivity gain.
Whether AI makes developers faster is less settled. In a 2025 randomized trial, nonprofit research group METR found that 16 experienced open-source developers took 19% longer to complete 246 tasks when using the AI tools available during the study. Participants had expected AI to make them faster and afterwards believed it had done so. METR said later data suggest developers may have gained more speed by early 2026, but described those results as only very weak evidence about the size of any effect.
Employment indicators send mixed signals. Indeed data cited in the source report show that 4.5% of US software-development job postings were entry-level in the first quarter of 2026, while 69.3% were senior. The report also cites Stanford researchers’ estimate that employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it otherwise would have been. Yet US software-development postings on Indeed were up about 23% year over year in November 2026, after reaching a low in May 2025. These figures describe different measures and do not establish that AI alone caused changes in hiring.
Coding Skills Beyond Writing Code
For students, career changers and employers, the practical question is not just whether AI can generate code. It is whether a person can tell the tool what a system should do, judge whether its proposed changes fit the rest of the software, and identify faults or security risks. Those tasks require enough technical understanding to evaluate an answer rather than accept it on appearance.
Learning methods may matter, too. 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 finding does not prove that AI always impairs learning, but it raises a concrete risk: getting working code quickly may not build the understanding needed to maintain it or find errors.
The evidence supports a distinction between automating code production and replacing software judgment. If organizations rely on AI for more routine implementation, workers may need stronger review and testing skills. At the same time, fewer entry-level postings could make it harder for newcomers to gain the experience needed to develop those skills. The available figures show a concern, not a settled explanation or forecast.
Products Worth Considering

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

Bose QuietComfort Wireless Bluetooth Noise-Cancelling Headphones, Cypress Green

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
From AI Predictions to Workplace Data
On March 10, 2025, Anthropic chief executive Dario Amodei told an audience at the Council on Foreign Relations that AI would be writing 90% of code within three to six months and essentially all of it within a year. In the same answer, he said programmers would still need to specify what was being built, how it fit with other code and whether its design was secure. Anthropic’s later account of its own codebase gives one company-specific measure of how quickly AI use expanded, but it does not establish that the same share applies across the software industry.
Other published figures use different dates and measures. 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 are not directly comparable: they cover different organizations, periods and definitions of AI-authored code.
Experiment results also depend on the work being tested. A 2022 study of 95 professional developers found that participants using GitHub Copilot completed a small JavaScript HTTP-server task 55.8% faster on average. METR’s later trial involved real tasks in large codebases familiar to experienced developers. That contrast helps explain why a tool may speed up a bounded coding exercise without producing the same result in complex maintenance work.
“The programmer still needs to specify”
— Dario Amodei, Anthropic chief executive, speaking at the Council on Foreign Relations on March 10, 2025
Productivity and Hiring Remain Unsettled
The available evidence does not settle how much AI improves developer productivity across different kinds of work. METR’s 2025 trial found slower completion, while a 2022 task-based experiment found a substantial speed-up. Its later study produced estimates pointing to tasks taking 18% less time for developers from the first study and 4% less for newly recruited developers, but the confidence intervals included no effect. METR said those results were unreliable as a measure of the true impact.
Anthropic’s figures also come from a company that builds and heavily uses the tool being measured. Its code-volume results do not establish code quality or prove that other employers will see similar output. On jobs, the cited decline in the relative share of entry-level postings and the estimate for younger workers do not, by themselves, show that AI caused those changes. How hiring, wages and training will develop remains unclear, as does whether employers will create new routes for junior workers to gain experience.
Track Hiring, Learning and Output
The next useful evidence will come from studies that track task completion, code quality and maintenance across a wider range of developers and software projects, rather than relying on code volume or a single type of exercise. METR’s later results suggest tools and working habits are changing quickly, but its stated uncertainty means those estimates should not be treated as a firm productivity forecast.
Readers weighing training or career choices can watch whether software employers expand entry-level hiring and structured training, and whether studies find that AI-supported learning improves or weakens later debugging and review skills. For now, the evidence supports learning how software works alongside learning to use AI coding tools; it does not establish that either traditional coding skills or AI assistance alone will guarantee a job.
Key Questions
Does the evidence show that learning to code is no longer useful?
No. The source material indicates that AI can generate substantial amounts of code, but developers still need to specify tasks, review output, test it and find defects. The evidence does not establish that coding knowledge has become unnecessary.
Does AI consistently make software developers faster?
No consistent productivity effect is established. METR’s 2025 randomized trial found experienced developers took 19% longer with the tools tested, while a separate 2022 experiment found faster completion on a small coding task. METR’s later estimates were uncertain.
What did Anthropic report about AI-written code?
Anthropic said Claude authored more than 80% of the code it merged into its own codebase in May 2026. The company also warned that its reported eightfold increase in code per engineer may overstate the productivity gain and does not measure code quality.
Are entry-level software jobs disappearing?
The cited Indeed data show entry-level roles made up 4.5% of US software-development postings in the first quarter of 2026, while senior roles made up 69.3%. Those figures do not prove that AI caused the pattern or show that entry-level work is disappearing; postings were also up about 23% year over year in November 2026.
Source: ThorstenMeyerAI.com
