Agentic engineering works on decade-old engineering practice, not more skills
alex_verem · x · 2026-11-17
- @mardehaym argues most companies fall into the trap of piling up skills and subagents in agentic engineering—writing a skill shows results the same afternoon, but a test suite running on every PR is what matters.
- His team has run managed delivery for a performance-marketing platform since 2023: 22 production releases in Q1 2026 with zero rollbacks; the client's CTO ships code via his own AI agent, gated by the team's engineering process.
- Core claim: where agents writing most of the code actually works, it comes down to engineering practices that were already good ten years ago.
More from coding & agent
- Self-learning agent loop takes Honor of Kings win rate from 0% to 60% in three rounds, no retraining — kimmonismus · 2026-11-17
- Dev builds a 'Flags' Claude mod so the AI keeps a session list of issues and reminders — dSebastien · 2026-11-17
- 10 prompts that treat AI coding models as senior engineers, not autocomplete — hey_abusiddik · 2026-11-17
- Termination Lag: coding agents find correct patches early but keep working, sometimes undoing them — PennState · 2026-11-17
- moveOS launches as an operating system for agentic engineering in autos — ycombinator · 2026-11-17
- Injecting Spring HTTP interfaces into MCP tools: a cleaner pattern with MCP Inspector testing — therealdanvega · 2026-11-17