Ask HN: With all the AI hype, how are software engineers feeling?

Hacker News Discussion ↗

Overall Morale and Emotional Climate

  • Morale ranges from energized and “loving coding again” to exhausted, angry, or deeply demotivated.
  • Some feel like they’ve gained a “superpower”; others feel like “cavemen” wasting time trying to make tools useful.
  • A number report psychological damage: loss of motivation to learn, blog, or publish OSS because it feels like “why bother, AI will do it / train on it.”

Perceived Productivity Impact

  • Claims split sharply:
    • Enthusiasts report 30–95% of their code or workflow now aided or written by AI, enabling solo devs to ship work that once needed teams.
    • Many others say AI does 0–10% of their work, or even slows them down due to bad suggestions and extra verification.
  • One linked study (of experienced OSS devs) is cited as showing ~19% productivity decline when using LLMs, reinforcing skepticism for seniors.

Hiring, Job Security, and Offshoring

  • Most say hiring hasn’t stopped; some orgs are still expanding engineering teams, especially seniors.
  • Others report slowed hiring and salary pressure, but attribute more to offshoring and cost-cutting than AI directly (though AI is sometimes used to justify offshoring).
  • A few are actively planning to leave tech, assuming AI will erode remote/outsourced opportunities first.

Management Expectations and Pressure

  • Common complaint: leadership believes (or pretends) AI can do “30–50% of the work,” cutting headcount while workload stays the same.
  • Some managers wield AI as a cost‑cutting pretext, or mandate AI usage and measure people on “shipping with AI.”
  • Devs are frustrated by PMs waving oversimplified LLM outputs as proof tasks are “trivial” and should be done “by end of day.”

Where AI Helps vs. Where It Fails

  • Helpful niches repeatedly mentioned:
    • Boilerplate code, simple CRUD, tests, small scripts, config scaffolding.
    • Documentation stubs, meeting transcription, rewriting tickets/notes, summarizing large codebases or papers.
    • Debugging when you can paste large logs + code, or exploratory work in unfamiliar libraries.
  • Weaknesses:
    • Complex, domain-heavy, legacy, safety-critical, or hardware/embedded code.
    • Existing large codebases with messy history and undocumented domain rules.
    • Hallucinated APIs, deprecated patterns, brittle tests, and “slop” PRs that OSS maintainers often reject.
    • Code agents that frequently go off the rails, require micromanagement, and still can’t complete end‑to‑end tasks.

Junior vs. Senior Engineers

  • Many observe juniors or “struggling” devs benefit more: help with syntax, patterns, and basic scaffolding.
  • Several seniors say their own net productivity gain is tiny or negative; they spend more time reviewing, correcting, and ensuring quality.
  • Some argue that if you see 50%+ gains, it may reflect prior low productivity or reliance on shallow work; others strongly disagree and point to solo‑consultant success stories.

Impact on Knowledge Ecosystem and the Web

  • Multiple comments worry that LLMs are killing Stack Overflow and niche info sites by diverting traffic while being trained on their content.
  • Maintainers and site owners report:
    • Floods of low‑quality AI PRs in OSS.
    • Traffic drops due to AI summaries, forcing them to divert months to “damage control.”
  • Concern that future LLMs will have worse training data as today’s Q&A and documentation ecosystems degrade.

Societal / Ethical Concerns and Personal Futures

  • Some see AI as primarily a “ruling class” profit play, with workers and independent publishers paying the price.
  • Worries about younger devs depending on AI, never learning fundamentals, and long‑term software quality collapsing.
  • A few in precarious situations (e.g., in war zones) feel AI specifically undermines one of the few portable, secure careers they had.

Views on the Hype Cycle and the Future

  • Many compare the current hype to crypto, Agile/Scrum, or “year of Linux on the desktop”: real utility, but wildly oversold.
  • Some are “biding time” for the bubble to burst; others think we’re at an early stage of a real shift where devs become more like architects/AI‑orchestrators.
  • Consensus points: writing code was never the main bottleneck; communication, requirements, domain knowledge, and organizational drag still dominate, and AI hasn’t fixed that.