Skip to content
Syed Asad Shah
Hire me
All articles
AI & Industry13 min read

Will Vibe Coding Take Your Developer Job? What the Data Says

Government labor statistics, payroll records, randomized trials, and job postings show how AI affects software engineering jobs in late 2026.

Syed Asad Shah

13 min read

In February 2025, Andrej Karpathy described a new style of programming called vibe coding. You give in to the vibes, embrace exponentials, and forget the code exists (Karpathy on X, archived). Nine months later, Collins Dictionary named vibe coding its 2025 Word of the Year. Soon, developers everywhere asked the same question. If you describe an app in plain English and Cursor, Copilot, Claude Code, or an agent builds the app, what is your job for?

You deserve an answer based on data instead of LinkedIn opinions. Here is what government labor statistics, payroll records, job indexes, developer surveys, randomized trials, and layoff trackers show in late 2026.

AI is not wiping out software developers as a group. AI is changing who gets hired, and junior developers feel the hit first.

What vibe coding means

In this post, vibe coding means AI-assisted development where you state your goal in natural language and iterate on generated code. This covers tools from autocomplete to autonomous agents, including Cursor, GitHub Copilot, Claude Code, ChatGPT, Replit Agent, and Devin.

Karpathy described the extreme version: accepting diffs without reading them and pasting errors back until the code runs. Most professional teams work with more care. Engineers use AI to draft code while humans own architecture, review, and tests. The key question is what happens to jobs when this workflow becomes normal.

By late 2026, AI tool usage is standard across the industry:

  • 84% of respondents in the 2025 Stack Overflow Developer Survey use or plan to use AI tools, up from 76% in 2024. 51% of professional developers use them daily.
  • Google’s 2025 DORA report found 90% of nearly 5,000 tech professionals use AI at work.
  • GitHub’s Octoverse 2025 reports 180M+ developers on the platform, with about 80% of new users trying Copilot in their first week.
  • At Google, Sundar Pichai said AI generated over 30% of new code in Q1 2025, up from 25% in late 2024 (Alphabet Q1 2025 remarks). Treat this company claim as directional, since Google published no methodology.

Almost every developer uses these tools now. Whether tool adoption costs people their jobs is a separate question.

The numbers at a glance

Indicator Figure Source
US software developers employed ~1.72 million BLS OOH, 2025
Projected developer growth (2025 to 2035) +10% (vs 3% all jobs) BLS OOH, Aug 2026
Previous 10-year projection (2024 to 2034) +15.8% (pre-revision) BLS TED, Jul 2026
Computer programmers (pure coding role) -7% projected (110,800 jobs) BLS OOH, 2025 to 2035
Median US software developer pay $135,980 per year BLS OEWS, May 2025
Indeed dev postings (Feb 2020 = 100) 61.1 low (May 2025) to 80.8 (Oct 2026) Indeed / FRED
Share of 2025 to 2026 hiring rebound in senior roles 71% Indeed Hiring Lab, Jul 2026
Dev employment aged 22 to 25 vs late-2022 peak ~-20% (to Sep 2025) Stanford DEL
Tech layoffs by year (Layoffs.fyi) 266k (’23), 153k (’24), 123k (’25), 135k (’26 YTD) CNBC-TV18, Oct 2026
Developers who distrust AI accuracy 46% (vs 33% who trust) Stack Overflow 2025
AI-generated code with OWASP Top 10 flaws 45% of tested tasks Veracode 2025

How much faster does AI make developers?

Claims about AI replacing entire teams depend on how much faster developers work with AI. Looking at the controlled studies in order tells a clear story.

  • 2023 lab trial (55.8% faster): In a randomized experiment, developers using GitHub Copilot built a JavaScript HTTP server in 71 minutes versus 161 minutes for the control group (Peng et al., arXiv). The study tested one small greenfield task, and only 35 of 95 recruited developers finished the task. Do not read this as proof of 2x output across production codebases.
  • 2024 enterprise field study (26.1% more tasks): A study of 4,867 developers at Microsoft, Accenture, and a Fortune 100 company found Copilot users completed 26.08% more tasks per week (Cui, Demirer et al., Management Science). Junior and short-tenure developers saw the biggest gains. This study gives the best real-world estimate for 2023 to 2024 tools.
  • Mid-2025 open-source trial (19% slower): METR ran a randomized trial with 16 experienced open-source maintainers working on their own large repos averaging over 1M lines of code. Using Cursor and Claude 3.5 or 3.7 Sonnet, developers took 19% longer. Before starting, these developers expected a 24% speedup. After finishing, they still believed AI made them 20% faster (METR, July 2025). Prompting, waiting, and fixing near-miss code ate up all the time saved on typing.
  • Early-2026 follow-up (4% to 18% faster, with selection bias): METR’s late-2025 update estimated an 18% speedup for returning developers and 4% for new recruits (METR, Feb 2026). METR calls this weak evidence because many developers now refuse to work without AI, even for pay. Between 30% and 50% of participants held back tasks they did not want to do without AI. Still, their refusal shows something important: developers do not fight to keep a tool if the tool slows them down.

Speedups are real on greenfield work and boilerplate, modest on routine enterprise tasks, and sometimes negative on large codebases you already know well. DORA’s 2025 report adds another warning: AI adoption links to higher delivery throughput, yet also higher delivery instability. Teams ship more code, and more code breaks.

Hiring and layoffs from 2022 to 2026

The crash started before AI matured

Software job postings on Indeed peaked in 2022 and then dropped hard. The index hit a low of 61.1 in May 2025 (with February 2020 set at 100), sitting 39% below pre-pandemic levels (Indeed Hiring Lab, Aug 2026).

Indeed notes the drop in AI-exposed jobs started before ChatGPT launched (Indeed, Jul 2026). Pandemic overhiring, rising interest rates, and the 2023 layoff wave (265,660 tech cuts on Layoffs.fyi) caused the 2022 to 2024 slump. During those years, AI models were too weak to drive the hiring slowdown.

Postings rebounded alongside coding agents

Since Claude Code launched in late February 2025, the same month Karpathy coined vibe coding, US software development postings on Indeed rose almost 15%. Over the same period, total US job postings fell 7% (Indeed Hiring Lab).

From 2022 to early 2025, AI-exposed roles lost the most job postings. Over the past year, AI-exposed roles rebounded the fastest. By October 23, 2026, the Indeed software index reached 80.8, sitting 19% below February 2020 (FRED).

Correlation does not prove causation. Still, the market does not show AI wiping out software work. Instead, cheaper software development creates demand for more software.

Layoffs are a noisy signal

Layoffs.fyi counted 134,917 tech layoffs in 2026 through October 21. This figure passes the 2025 total of 122,606, while staying near half the 2023 peak (CNBC-TV18). Keep three facts in mind when reading layoff news:

  • Trackers disagree on totals. TrueUp counts 196,349 tech layoffs in 2026.
  • Tech layoff numbers include all company departments, such as recruiting, sales, and support, rather than only engineers.
  • Executives use AI as a cover story for cost-cutting and data-center spending. Barclays cited needing fewer developers as AI raised output, yet the bank’s CEO called AI “the icing on the cake, not the cake.”

Juniors vs. seniors: the main fault line

Entry-level hiring shows the sharpest drop in the entire dataset.

Stanford’s Digital Economy Lab analyzed ADP payroll records for millions of US workers. Employment for software developers aged 22 to 25 fell nearly 20% from its late-2022 peak to September 2025, while employment for older developers kept growing (Brynjolfsson, Chandar & Chen). Their mid-2026 update shows young workers in AI-exposed jobs sit 19% below peers in less-exposed jobs. This gap widened after 2025 and comes mostly from slower hiring rather than firing (Stanford DEL).

The Stanford authors treat these numbers as early warning signs rather than proof of direct AI causation, since high interest rates and hiring freezes also hurt new graduates. Other sources confirm the same trend:

  • 71% of the 2025 to 2026 rebound in software job postings came from senior roles, and 37% came from jobs with AI in the title (Indeed).
  • The NY Fed reported a 5.7% unemployment rate for recent college graduates in June 2026, compared to 4.1% for all workers (NY Fed). Major-level data placed computer engineering unemployment at 7.8% and computer science at 7.0%, ranking among the highest of all majors (Forbes, Feb 2026).

Why do juniors take the hit? AI excels at boilerplate, CRUD endpoints, unit tests, and small, well-defined bugs. Companies used to assign these tasks to new hires for on-the-job training.

Stanford found employment drops where AI automates work and grows where AI augments work. Anthropic’s usage data aligns with this split: 79% of Claude Code conversations involved automation rather than augmentation, versus 49% on Claude.ai (Anthropic Economic Index, 2025).

Juniors face a tough paradox. Field studies show junior developers gain the largest speed improvements from AI tools, yet those same tools handle the entry-level backlog companies once hired juniors to clear.

Which tasks shrink, and which grow

Tasks shrinking or getting cheap

  • Boilerplate, scaffolding, glue code, and CRUD endpoints.
  • First-draft unit tests and documentation.
  • Turning a strict spec into code. BLS projects computer programmers to shrink 7% by 2035, citing AI automation of repetitive coding tasks (BLS).
  • Simple prototypes and internal tools built directly by product managers or designers.

Tasks growing in value

  • Reviewing and verifying AI code: 66% of developers spend more time fixing near-miss AI code (Stack Overflow 2025).
  • Security: Veracode found AI introduced OWASP Top 10 flaws in 45% of coding tasks, with newer models showing little security progress (Veracode). BLS projects information security analysts to grow 28.5% from 2024 to 2034.
  • System design, architecture, and clear specifications: coding agents depend on the boundaries you set.
  • AI, ML, and data engineering: BLS projects data scientists to grow 33.5% from 2024 to 2034 (BLS TED). About 47% of Indeed software postings now mention AI (Indeed, Apr 2026).
  • Platform, reliability, and observability: when AI raises delivery instability, teams need stronger monitoring and testing to catch failures.

Five ways your daily work changes

  1. Your core output shifts from lines of code to intent. You write specs, constraints, and tests, then review diffs.
  2. Prototyping costs almost nothing. Non-developers build working demos, raising expectations for production engineers.
  3. Code review becomes the main bottleneck. With 46% of developers distrusting AI output versus 33% trusting AI output, verification takes up much of your day.
  4. Teams get flatter and more senior. Companies hire fewer juniors and expect each senior engineer to supervise more AI output.
  5. AI fluency becomes a baseline skill. Employers expect you to steer coding agents as part of normal development.

Historical parallels in software

  • Compilers (1957): IBM pitched FORTRAN as an automatic coding system, cutting coding and debugging work by over 80%. One example turned 47 statements into 1,000 machine instructions (Backus et al., 1957, IBM). Programmer jobs did not vanish. Cheaper software led companies to build far more software, growing the profession many times over.
  • IDEs, Stack Overflow, and open source: Each wave removed hours of typing and searching. Developer demand kept rising because lower costs created new software use cases.
  • Offshoring (2000s): Offshoring did not end US software engineering, yet offshoring hollowed out routine spec-to-code jobs and pushed onshore engineers toward design, architecture, and system integration.
  • No-code and low-code (2010s): Many predicted no-code tools would replace developers. Instead, business users built thousands of prototypes, and those apps later needed real engineers for security, integration, and scale. Vibe-coded apps follow the same pattern.

What makes this wave different: past tools automated one layer at a time, such as assembly, search, or hosting. Today’s agents attempt entire tickets from start to finish, and model capabilities improve fast. History offers context, not a guarantee.

Will AI take developer jobs?

  • For the profession overall, the data shows augmentation so far. BLS projects software developer jobs to grow 10% by 2035, more than triple the 3% average across all jobs, though down from the prior 15.8% forecast. Job postings are climbing off their 2025 floor, and employment for experienced developers keeps rising. Goldman Sachs estimates AI adoption will displace 6% to 7% of the US workforce over a decade, with most workers moving into new roles (Goldman Sachs).
  • For entry-level and pure coding roles, displacement pressure is happening now. A 20% drop in young developer employment and a 7% projected decline for computer programmers are clear warning signs.

Expected timeline

  • Now to 2027: Fewer junior job openings, a senior-heavy and AI-focused hiring rebound, and tougher technical interviews.
  • 2027 to 2030: As agents handle multi-file tasks with fewer errors, team sizes per product will shrink. Total developer jobs will still grow if cheaper software expands total demand. The World Economic Forum employer survey still ranks software developers among the fastest-growing roles through 2030 (WEF Future of Jobs 2025).
  • The 2032 pipeline problem: The biggest risk is the broken training ladder. If companies hire no juniors today, the industry faces a shortage of senior engineers in 2032.

Key takeaways

  • AI tool adoption is near universal at 84% to 90%, while developer trust stays low, with 46% distrusting AI accuracy.
  • Real-world speedups average around 26% more completed tasks, rather than 10x, and drop below zero on complex legacy code.
  • Macro conditions caused the 2022 to 2024 hiring crash, while the 2025 to 2026 rebound favors senior and AI-skilled engineers.
  • Junior developers take the hardest hit, with employment for ages 22 to 25 down nearly 20% from 2022 levels.
  • Raw coding shrinks as a share of the job, while specs, code review, security, and system design grow in value.

FAQ

Should you still learn to code in 2026?

Yes. Aim to judge and debug code, rather than only type syntax. Strong fundamentals in debugging, security, databases, and system design let you turn raw AI output into reliable software.

What should you do as a junior developer?

Show employers you ship complete projects using AI tools. Write thorough tests, practice reviewing and fixing AI-generated diffs, and focus on areas where AI augments human judgment, such as security, infrastructure, data engineering, and complex domain logic.

Are companies replacing engineers with AI?

Some companies say they are, including Barclays. Many AI layoff announcements also hide routine cost cuts or spending shifts toward data centers. Gartner predicts half of companies cutting staff for AI will rehire for similar roles by 2027 (CNBC-TV18). Treat this as a forecast rather than a proven result.

Is vibe coding safe for production?

Not without human review. About 45% of AI-generated code samples failed basic security checks in Veracode’s tests.

Sources

Government and labor data

Hiring and layoffs

Research

Surveys and industry reports

Data current as of early November 2026. Employment projections are estimates, and AI labor research remains early and mostly correlational.

Written by

Syed Asad Shah

Computer Systems Engineer, Full Stack Developer, and AI Engineer. I build custom software — web, mobile, and desktop apps — and AI systems you can actually ship.

Have a project like this in mind?

Tell me what you’re building and where you’re stuck. I’ll tell you how I’d approach it.

Hire me