November 12, 2025
Angels in your IDE, demons in comments
Software Development in the Time of New Angels
AI ‘angels’ crash the $150/hour party—devs split between hype, panic, and eye-rolls
TLDR: An engineer says agent-like AI tools can now write code, shaking the $150/hour developer model. Comments clap back with clickbait complaints, defense of past mishaps, and a reality check that humans still must exceed AI—use it as a helper, not a replacement—making this shift urgent and unavoidable.
The author claims the $150/hour world of coding tilted five months ago when “agentic AI”—bots that plan and write code—arrived via Claude Code. Cue the comments: aetherspawn called out stealth hype (“Put (AI) in the title!”), while datadrivenangel defended engineers in the infamous Knight Capital meltdown, saying the story unfairly blames devs. hdivider brought practical rain: yes, the AI thrill is real, but who’s handling maintainability, scalability, and “overly complicated” solutions? The crowd split fast: some say AI angels are lifting productivity and slashing grunt work; others fear shadow demons lurking in the codebase and long-term tech debt.
Even the fans agree this isn’t the death of developers—it’s the end of coasting. datadrivenangel highlighted the author’s stance: AI won’t replace devs wholesale. Then williamstein dropped a reality check: hiring today means expecting more than what Claude can do. Translation: AI is table stakes, not a miracle. The jokes flew—“fancy autocomplete with wings,” “heaven vs legacy hell”—while aetherspawn’s clickbait jab kept the vibe spicy. The consensus-ish: agentic AI boosts speed, but humans own judgment, architecture, and accountability. Your IDE got an angel; your pull requests still need a grown‑up. And that $150/hour? The comments say it’s officially on trial. For now.
Key Points
- •$150/hour is presented as a typical cost for senior software developers in the United States, shaping how software development is organized.
- •The article claims a major shift occurred five months ago with the initial release of agentic AI for software development.
- •Earlier experiments with GPT-3.5 and completion-based automation reportedly tripled the author’s coding velocity.
- •The author adopted Anthropic’s Claude Code to implement a Java parser, AST, and tooling in Scala for a side project.
- •The side project aimed to render software projects as knowledge graphs using language processing, static analysis, and semantic web technologies, later named Project Sagrada.