My old job feels distant now
AI coding tools and autonomous “agents” are rapidly reshaping software work, with many engineers saying the intellectually satisfying parts of programming are being automated away while the remaining job tilts toward coordination, stakeholder management, and review. Commenters are split between enthusiasm for huge productivity gains and new solo-building opportunities, and anxiety about collapsing salaries, diminished demand for traditional engineering roles, and a future where only a smaller elite of architect‑ or product‑type engineers remain in demand. Underneath is a broader economic argument over whether AI-driven code generation creates an unsustainable glut of low-quality software and precarious careers, or simply accelerates a long-term shift toward more abstract, business-driven forms of “knowledge work.”
Impact of AI on Dev Jobs and Wages
- Many expect downward pressure on salaries: first juniors, then seniors, with some imagining experienced devs working for very low wages rather than starve.
- Others argue that AI primarily deletes “code monkey” roles while increasing demand for people with product, business, and relationship skills.
- Some see AI adoption as non-optional: refuse the tools and you’ll struggle to get hired; accept them and you may still face a tougher market later.
- Comparisons are made to elevator operators vs technicians and to agriculture: fewer workers needed even as total output grows.
Developer Workflow and Productivity with Agents
- Several posters describe radical productivity gains: delegating complex CI fixes overnight, building full apps or platforms with minimal budget, never getting “stuck” anymore.
- Some like that AI removes drudgery (boilerplate, syntax, refactors), letting them focus on higher-level design or even personal hobbies during work hours.
- Others feel their job has become monotonous: feed tickets to an LLM, review, merge, repeat.
Quality, Reliability, and Oversight Concerns
- Multiple comments report LLM-generated code as messy, overgrown, inconsistent, or subtly wrong, requiring constant steering and review.
- Open-source maintainers complain about huge, low-signal AI-generated PRs and documentation, increasing review burden.
- There is worry that speed gains might quietly come from skipping tests or lowering standards.
- Some developers choose to avoid AI entirely, viewing it as overhyped and transient.
Economics and Sustainability of AI
- Debate over whether current AI economics are sustainable: some point to unprofitable cost structures; others stress explosive growth and falling “cost of intelligence” as justification.
- Open-weight models (including from China) are seen as reducing dependence on major labs.
- Corporate environments are described as slower to adopt AI than startups, both for capability and governance reasons.
Changing Nature of Software Work and Identity
- Several posters mourn the loss of “fun” problem-solving in favor of stakeholder wrangling and performance politics, now seemingly amplified by AI.
- Others argue software was already boring; AI just makes the grind “less unfun.”
- Broader concern: as code becomes commoditized, value shifts to defining problems, making tradeoffs, and influencing people—skills many engineers don’t enjoy.