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

🔍 Read the full analysis: Coding And AI: How Students, Developers And Managers Can Weigh The Evidence on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • 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

Evidence compiled by ThorstenMeyerAI.com points to AI writing a large share of code at some companies, while controlled research has not established a consistent speed gain. Entry-level hiring is weak in the cited data, but software job postings have rebounded; studies also warn that relying on AI can hinder some learning.

A November 18 review by ThorstenMeyerAI.com finds that AI now writes much of the code at some companies, but available research does not show that developers consistently finish work faster. The report also points to few entry-level software postings in recent U.S. data and a learning experiment in which junior developers using AI scored lower on a follow-up quiz.

The clearest company figure comes from Anthropic, which says Claude authored more than 80% of the code it merged in May 2026, compared with “the low single digits” before Claude Code launched in February 2025. Anthropic also reported that 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 and that the eightfold figure is “almost certainly an overstatement” of the productivity gain.

Research on work speed is less settled. In a randomized METR study involving 16 experienced open-source developers and 246 real tasks, participants using AI tools available from February to June 2025 took 19% longer to complete tasks. They had expected a 24% reduction in time and afterward believed AI had made them 20% faster. METR’s later study produced estimates suggesting shorter task times, but its confidence intervals included no effect; the group described that evidence as weak.

The labor-market figures in the report show pressure at the entry level alongside a recovery in overall postings. Indeed data cited by the report put entry-level roles at 4.5% of U.S. software development postings in the first quarter of 2026, while senior roles accounted for 69.3%. The same source reported that U.S. software development postings were about 23% higher year over year in November 2026, after reaching a low in May 2025. Stanford researchers, as summarized in the report, estimated employment among 22- to 25-year-olds in AI-exposed jobs was 19% below where it otherwise would have been.

At a glance
analysisWhen: Published November 18, 2026; evidence i…
The developmentA November 18, 2026, report reviews company disclosures, job data and research on how AI is changing coding work, hiring and learning.

Coding Work Is Shifting Upstream

The evidence matters to students, developers and employers because it separates AI’s ability to produce code from proof that it improves the whole software development process. More generated code does not, by itself, establish better software, lower costs or faster delivery. Anthropic’s own caveat about code volume underlines that distinction.

For students and career switchers, a thin share of entry-level listings could make it harder to get an initial role, although the cited figures do not establish that AI caused the change. For working developers, the results suggest that review, testing, debugging and clear task specification remain relevant when AI generates code. Managers deciding on hiring or training should treat company output figures and developer surveys as evidence with limits, not as a direct measure of business-wide productivity.

Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

As an affiliate, we earn on qualifying purchases.

From Code Generation to Review

In March 2025, Anthropic chief executive Dario Amodei predicted that AI would write 90% of code within three to six months and essentially all code within a year. In the same remarks, he said programmers would still need to specify what was being built, how it fit with other code and whether the design was secure. The report uses that distinction to frame the evidence: code production is changing, while human judgment remains part of the work.

Other company figures differ in scope and date. Google chief executive Sundar Pichai said in October 2024 that AI generated more than a quarter of new Google code, 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 measures: they cover different organizations, tools, periods and definitions of code authorship.

Experiments also test different kinds of work. A 2022 test of 95 developers writing a small JavaScript HTTP server found that participants with Copilot completed the task 55.8% faster on average. METR’s later trial involved tasks in large, familiar codebases, where developers had to work within established project rules. The contrast cautions against treating one result as a universal measure of AI’s effect.

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

— Anthropic

Productivity and Hiring Remain Unsettled

The cited evidence does not establish a single, general productivity effect across software work. METR’s first trial found longer completion times, while its later study was inconclusive; tools and developer habits have also changed. Anthropic’s internal figures come from a company building and heavily using the technology, and its output measure does not capture code quality or maintenance costs.

The report also describes an Anthropic learning experiment in which mostly junior developers using AI to learn a new library scored 50% on a follow-up quiz, compared with 67% for those who coded by hand, with the largest gap in debugging. The result does not show that all AI use weakens learning, or how the effect changes with different teaching methods. Nor do the job-posting and employment figures alone show whether AI caused changes in entry-level hiring. The report provides no single, comparable measure covering software quality, long-term employment or productivity across the industry.

Watch Hiring and Replication

The next useful evidence will come from studies that test current tools on varied, real-world tasks and measure more than completion time, including defects, review effort and maintenance. METR’s reported difficulty recruiting developers willing to work without AI also shows why later experiments need to account for changing expectations and tool dependence.

Readers can track whether the gap between entry-level and senior postings narrows as overall software vacancies change, and whether employers report altered training or hiring practices. Until those data are available, the report’s findings support a measured takeaway: learning to code still has value, but coding education and workplace training may need more emphasis on reading, testing, debugging, specifying requirements and judging security. The size of AI’s longer-term effect on jobs and output remains unresolved.

Key Questions

Does the report show that AI makes developers faster?

No consistent speed gain is established. METR’s 2025 randomized trial found that experienced developers took 19% longer with AI on the tasks tested. Its later estimates suggested possible time savings but were uncertain and included no effect.

How much code does AI write at Anthropic?

Anthropic said Claude authored more than 80% of the code it merged in May 2026. That is a company-specific figure, not an industry-wide estimate, and Anthropic cautioned that code volume overstates productivity gains.

Are entry-level software jobs disappearing?

The cited Indeed data show that entry-level roles made up 4.5% of U.S. software development postings in the first quarter of 2026. The figures do not prove AI caused the limited share, and overall postings were about 23% higher year over year in November 2026.

Should students still learn to code?

The report’s evidence supports continuing to learn programming while developing skills in reading, testing and debugging code, specifying requirements and evaluating security. One experiment found lower quiz scores among junior developers who used AI to learn a library, but it does not establish that every use of AI harms learning.

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

The $60 Billion Bargain: Why Cursor Could Be a Steal for SpaceX

SpaceX’s recent $60 billion all-stock acquisition of AI coding startup Cursor is a strategic move, offering growth, market control, and potential profitability.

Jack Clark Says It Out Loud — Reading the Co-Founder’s 60%/2028 Estimate on Automated AI R&D

Anthropic’s co-founder Jack Clark states there is a 60%+ probability that AI systems capable of autonomously building their successors could emerge by 2028.

Breakeven Analysis: Find the Point

Navigating breakeven analysis reveals how to determine the point where profits begin, unlocking essential insights for your business success.

Apple greift nach China-Speicher. Europa hat nicht einmal diese Option.

Apple plant, Speicherchips vom chinesischen Hersteller CXMT zu beziehen, während Europa keine vergleichbare Option hat. Das zeigt die Abhängigkeit Europas in der Halbleiterindustrie.