What the internet is saying about Prompt Engineering

What the internet is saying about "prompt engineering"

Evidence caveat: None of the 17 items discusses "prompt engineering" directly. The closest signals are adjacent topics: agentic engineering, harness engineering, and AI coding tools. Everything below is inference from those, not a measured view of prompt engineering itself.

🟢 Supporters

• •No source makes a direct case for prompt engineering as a discipline.
• •Adjacent signal: Hacker News carries items on the engineering practices around LLMs. These include "100 Levels of Agentic Engineering" [1] (5 points, 1 comment) and "What we've learned 3 years into harness engineering" [13] (5 points, 0 comments). They suggest interest in structured, skill-based approaches to working with AI. Only titles are available, so I can't say what they conclude.
• •Adjacent signal: Dashlane's MV3 feedback says "I used AI to help write it" [6]. It drew 34 reactions, which shows AI-assisted authoring is accepted in serious technical contexts. It is not evidence about prompting technique.

🔴 Skeptics

• •No source criticizes prompt engineering by name.
• •Adjacent friction with AI tooling:
  • •The VS Code issue on "granular control over Copilot changes" [3] has 232 comments and 295 reactions. It is framed around "more agentic workflows" and how changes are "made, reviewed," and users want more control. This is a sign of pushback on agent behavior, not on prompting.
  • •A Codex Desktop issue [2] reports the weekly quota falling from 77% to 10% in 38 seconds (67 points). It has 47 comments and 44 reactions. That is a cost and reliability complaint about AI tooling, not about prompts.

⚪ Neutral Observers

• •The shift in vocabulary is visible in the titles. "Agentic engineering" [1] and "harness engineering" [13] appear, while "prompt engineering" does not. This is consistent with, but doesn't prove, discussion moving from prompt wording toward system-level engineering.
• •The remaining items are unrelated to the topic: CVEs [5][16], YouTube GIFs [7], Kia login [9], Leaflet [11], CNCF [12], Deno [15], pressure regulators [17], and others. They are likely matches on the word "engineering."

Cross-Platform Signals

• •GitHub has the most activity and the highest engagement. It is practical and complaint-driven, centered on Copilot controls [3] and Codex quota [2].
• •Hacker News has only a few low-engagement items (0–1 comments, 5–6 points), mostly about agentic and harness engineering [1][13].
• •There is no consensus or divergence on prompt engineering, because neither platform addresses it directly.

Community Verdict

The evidence does not support a verdict on prompt engineering. Attention in this window appears to sit with agentic workflows, harnesses, and the cost and control problems of AI coding tools [1][2][3][13]. Those threads show users wanting predictability and oversight of AI agents, but they say nothing about prompt craft as a skill. A proper read would need sources that actually mention prompt engineering.

---

Based on Hacker News + GitHub data from the last 30 days, researched with last30days by Matt Van Horn.