AI made coding easier, not software engineering easy, dev argues
zishanverse · reddit · 2026-08-16
A developer's long Reddit post on AI's real impact on software engineering:
- AI genuinely makes "standard parts" — landing pages, CRUD apps, APIs, dashboards, auth flows — trivially easy to generate; for some projects clients can indeed DIY with AI.
- But "I can generate the code" ≠ "I understand the system": architecture, where business logic lives, handling 10,000 users, database design, mid-request API failures, auth and data protection, wrong LLM answers, prompt injection, production monitoring, six-month maintainability — that's the hard core of software.
- AI can generate an answer to almost all of these, but generating an answer isn't knowing whether it's correct. The dangerous slide is from "AI helps devs write faster" to "AI means we no longer need to understand development" — two completely different statements.
- The author's take: as code gets cheap, the valuable skill shifts from writing code to judging whether code is correct, so good developers become more valuable, not less. The scary future isn't AI writing bad code — it's humans deploying systems they don't understand because "the AI said it was fine."
The post ends by asking the community: has AI actually reduced the need for developers, or just moved the hard part somewhere else?
More from coding & agent
- Scheduled Chaos Testing Framework for Production AI Agents — blaizedsouza · 2026-08-16
- Hermes turns any website into an API for agents with one browser run — Saboo_Shubham_ · 2026-08-16
- Graphify (107k Stars) Turns Codebases and PDFs into Queryable Knowledge Graphs — tom_doerr · 2026-08-16
- User reports Opus 5陷入 fix-regress loop in Claude Code — dseven4evr · 2026-08-16
- Claude Code gets 'find-skills' to auto-discover and install required tools — dr_cintas · 2026-08-16
- Blender MCP connection lost error with incomplete JSON response — jcam12312 · 2026-08-16