🔍 Read the full analysis: A Practical Look At Coding Skills For Students, Developers And Managers on ThorstenMeyerAI.com
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
TL;DR
A November 2026 review of company figures, hiring data, experiments and surveys finds that AI can author a large share of code, but its effect on developer productivity remains uncertain. The evidence points to continued value in learning to specify, read, test, debug and evaluate code, even as entry-level hiring appears constrained.
A November 18, 2026 review by ThorstenMeyerAI.com finds that AI is writing a substantial share of code at some companies, while evidence that it consistently makes developers faster remains mixed. For students, working developers and hiring managers, the report’s central finding is that coding skills still matter, but more of the work may involve specifying, reviewing, testing and debugging code generated with AI.
Anthropic reported that, in May 2026, Claude authored more than 80% of the code merged into the company’s codebase. The company said the share had been in the low single digits before Claude Code launched in February 2025. It also reported that 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 the productivity gain.
Independent experimental evidence is less straightforward. In a 2025 randomized trial, the nonprofit METR studied 16 experienced open-source developers completing 246 tasks in projects they knew. Developers took 19% longer with AI tools, despite expecting a speed-up and later believing the tools had made them faster. METR said in February 2026 that a newer study offered only very weak evidence about how much faster developers might now be: its estimates were uncertain and included no effect.
The labor market presents a separate concern for people entering the field. The report cites Indeed data showing that 4.5% of U.S. software development postings were entry-level in the first quarter of 2026, while 69.3% were senior. It also says U.S. software development postings were up about 23% year over year in November 2026, after reaching a low in May 2025. Those figures describe different points in time and do not by themselves establish why hiring changed.
Coding Skills Beyond Code Generation
The figures point to a changing division of work, not a simple end to programming. At Anthropic, the company says engineers direct and review code authored by Claude. That distinction matters for learners and employers: generating code is only one part of delivering software, and someone still has to define the task, fit a change into an existing system and check that the result works safely.
For students and career switchers, the narrow share of entry-level listings suggests a difficult route into the field, although the data cited here do not show that AI alone caused the change. A skill set built around reading, testing and debugging may help learners understand and verify generated code rather than simply accept it. For managers, code volume is an incomplete measure of output: Anthropic itself warns that more lines do not necessarily mean better or more productive work.
The report also flags a learning trade-off. In an Anthropic experiment, mostly junior developers using AI while learning a new library scored 50% on a follow-up quiz, compared with 67% among those who coded by hand. The result raises a question about how AI should be used in training; it does not establish that all AI-assisted learning reduces skill or that the result applies to every developer.
Products Worth Considering

FLEXISPOT 3-in-1 Convertible Chair and Table for Four

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

Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER), QHD
As an affiliate, we earn on qualifying purchases.
From Coding Predictions to Measured Results
The review revisits a prediction made by Anthropic chief executive Dario Amodei at the Council on Foreign Relations on March 10, 2025. Amodei said AI would be writing 90% of code within three to six months and essentially all of it within twelve months. The review compares that forecast with Anthropic’s later internal figures, as well as research and hiring data. Anthropic’s adoption figures are important but come from the company that develops Claude, so they should not be treated as a measure of every workplace.
Other cited estimates use different measures and periods. 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 about 4% of GitHub commits were authored by Claude Code. These figures are not directly comparable: they refer to different companies, time periods and definitions of AI-written 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. The contrast suggests that results from a short, bounded exercise may not predict performance on complex work, and that tool capabilities and study conditions have changed over time.
“More than 80% of the code we merge into Anthropic’s codebase was authored by Claude.”
— Anthropic
Limits of the Productivity Evidence
It remains unclear how much AI changes software productivity across the industry. Anthropic’s internal measures come from a company with unusually extensive use of its own coding tools, and its code-volume figures do not directly measure quality or time saved. METR’s 2025 trial found slower completion with AI, while its later study was affected by developers who were reluctant to do tasks without AI and produced estimates with confidence intervals that included no effect.
The hiring numbers also need careful interpretation. The reported shares of entry-level and senior postings show the composition of listings, not the total number of jobs available or the reason employers set those requirements. The report cites Stanford researchers’ finding that employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it would have been, but that comparison does not establish that AI caused the gap. The available evidence does not settle how future demand, wages or training pathways will change.
What Learners and Employers Can Track
The next useful evidence will be longer-term studies that measure completed work, quality, security and time across varied teams—not just code volume or short tasks. METR’s newer results remain uncertain, so its assessment may change as studies include more developers and account for the growing reluctance some participants reported about working without AI.
For learners and managers, the immediate question is how to use AI without skipping the understanding needed to check its output. The report’s evidence supports paying attention to clear task specification, code review, testing and debugging, while recognizing that it does not prescribe a single curriculum or hiring standard. Hiring trends and workplace results will need to be monitored as more comparable data emerge.
Key Questions
Does the report say students should stop learning to code?
No. Its conclusion is that coding remains worth learning, but learners should build skills beyond typing code, including reading, testing and debugging AI-generated work.
Does AI make developers faster?
The evidence cited is mixed. In METR’s randomized 2025 trial, experienced developers took 19% longer with AI tools, while a 2022 Copilot experiment found faster completion on a small programming task. METR said its later evidence on possible speed-ups was weak and uncertain.
What did Anthropic report about Claude-written code?
Anthropic said Claude authored more than 80% of code merged into its codebase in May 2026. The figure applies to Anthropic, and the company cautioned that its measure of code volume likely overstates the productivity gain.
What do the entry-level job figures show?
The review cites Indeed data showing that entry-level listings made up 4.5% of U.S. software development postings in the first quarter of 2026, while senior roles made up 69.3%. These figures do not establish why the distribution changed or how many jobs were available overall.
Source: ThorstenMeyerAI.com
