What is happening to jobs? Separating AI hype from reality
Claims that AI is already wiping out white‑collar jobs are met with skepticism here, with many pointing instead to pandemic over‑hiring, broader economic cycles, and companies using “AI” as a convenient pretext for layoffs. Commenters describe agentic coding tools as having only recently become genuinely useful, but say hard evidence of large, sustained productivity gains is still thin, highly uneven across experience levels, and often offset by new tech debt, higher infrastructure costs, and organizational inertia. Overall, people expect AI to change the nature of software and knowledge work—especially for juniors and routine tasks—but see mass unemployment and rapid, across‑the‑board displacement as far from proven.
State of AI and Coding Agents
- Many see a step-change around late 2025 / early 2026: coding “agents” start to feel like real workers rather than chat toys, especially with write–run–fix loops and better scaffolding.
- Others argue similar “this time it works” claims have been made since 2023; they see shifting goalposts and crypto‑style hype.
- Capabilities are described as “spiky”: excellent on some coding tasks (simple web apps, small features, debugging), poor or brittle on others (complex architecture, parallelism, some APIs, Docker/local setups).
Productivity Effects and Evidence
- Reported gains are mostly anecdotal; several note that rigorous studies lag the tech and often rely on self‑reports or weak metrics (SLOC, PR counts).
- One cited study found devs felt 20% faster but were actually ~20% slower.
- Some say agents now handle entire small apps or JIRA tickets; others experience regressions when tooling or resource limits change.
- There’s disagreement whether LLMs mainly boost juniors/average devs (regressing to the mean) or disproportionately empower already‑productive experts.
Jobs, Layoffs, and Labor Market
- Many think current tech layoffs are primarily pandemic over‑hiring and macroeconomics, with AI used as a convenient justification.
- Data in the article reportedly shows AI‑exposed sectors aren’t losing more jobs; commenters note unemployment moves together across sectors.
- Some recruiters report renewed hiring linked to “agent‑first” hype, while non‑startup environments still talk layoffs and AI pressure.
- New grads and juniors seem to face the worst market; several see sharply reduced junior hiring.
Junior vs Senior Engineers & Skills
- Seniors describe agents as strong multipliers if tightly guided; fully autonomous coding is called a “disaster”.
- Many managers say they now prefer fewer seniors plus AI and are reluctant to hire juniors, who may become “prompt‑and‑paste” operators and fail to build deep skills.
- Others contend effective AI use itself is a high‑skill activity that still favors strong engineers.
Code Quality, Tech Debt, and “Vibecoding”
- Multiple reports of “uncanny”, hard‑to‑maintain LLM code and growing tech debt when humans step out of the loop.
- Some companies are replacing niche SaaS with quickly “vibe‑coded” internal tools; long‑term reliability, security, and maintainability are flagged as unclear.
- Experienced devs see more incidents and lower average quality in some AI‑heavy codebases.
Organizational and Economic Dynamics
- Organizational inertia, internal AI bans, security concerns, and GPU/API costs significantly limit real deployment.
- Employees often hide true productivity gains, fearing they’ll be used to justify layoffs; AI adoption can be performative.
- Job ads with impossible “years of agentic AI” requirements are viewed as HR ritual, not reality.