跪拜 Guibai
← All articles
JavaScript

The End of Prompt Engineering: Why Context Frameworks Are the Real AI Workhorse

By 默_笙 ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

LLMs have gotten good enough that elaborate prompt-crafting is obsolete — the models auto-optimize short inputs. The bottleneck is no longer how you phrase the instruction but whether you supply the right context, constraints, and output format. Developers who treat AI integration as a context-engineering problem, not a prompt-writing problem, get deterministic results from non-deterministic models.

Summary

The three-stage evolution of AI engineering moves from Prompt Engineering (2022-23), through Context Engineering (2024-25), and into Harness Engineering (2025-26). Prompt Engineering treated LLM output like a gacha pull — a well-crafted prompt raised the odds of a good answer but never guaranteed one. Context Engineering flips the approach: instead of relying on pre-trained knowledge alone, it retrieves and injects relevant background, constraints, and format requirements before the model generates a response. RAG is the most common implementation, and tools like Cursor already apply this by feeding an entire codebase as context.

A practical example structures a milk-tea design task into three fields — background (shop owner, student customers, 15-20 yuan price point), constraints (summer drink, under 8 yuan cost), and output requirements (visually appealing, JSON format). This framework gets assembled into a system prompt and sent to DeepSeek with temperature 0.7 for creative balance. The code wraps the call in two layers of try-catch: one for API failures, one for JSON parsing failures, because AI output is never 100% predictable.

Modern LLMs already auto-optimize user prompts internally, which is why short prompts now outperform the long, detailed ones that GPT-3.5 required. But context engineering remains essential because no model knows your internal documents, your specific scenario, or your constraints unless you supply them explicitly.

Takeaways
Prompt Engineering (2022-23) treated LLM output like a gacha pull: a good prompt raised the odds but never guaranteed a correct answer.
Context Engineering (2024-25) retrieves and injects background, constraints, and format requirements before generation, making output reliable.
RAG is the most common high-level implementation of Context Engineering.
Cursor outperforms ChatGPT on code because it feeds the entire project codebase as context — tech stack, code style, and dependencies.
Harness Engineering (2025-26) adds rules, guardrails, safety layers, and loops for deterministic delivery akin to traditional software.
Modern LLMs (Claude 4.6, Gemini 3, GPT-5) auto-optimize user prompts internally, so short prompts now beat the long, detailed ones GPT-3.5 required.
A context framework has three elements: background (who you are, what you do), constraints (limits and boundaries), and output requirements (format and spec).
Temperature 0.7 balances creativity and coherence for design tasks; 0.0-0.3 suits code generation and data analysis.
Every AI call needs two layers of fault tolerance: one for API/network failures, one for malformed or unexpected output formats.
Conclusions

Prompt Engineering is effectively dead as a standalone skill — the models now do the prompt optimization themselves, so the value has shifted upstream to context assembly.

The three-era framing (Prompt → Context → Harness) maps cleanly onto the maturity curve of any engineering discipline: first you hack the tool, then you feed it better data, then you build guardrails around it.

Calling temperature a 'creativity dial' undersells it; the choice of 0.7 for a milk-tea design task versus 0.2 for code generation is really about whether you want the model to interpolate within known patterns or extrapolate beyond them.

The two-layer try-catch pattern is a microcosm of AI engineering: one layer handles infrastructure failure, the other handles semantic failure, and both are mandatory for anything that touches production.

Concepts & terms
Context Engineering
An AI engineering paradigm (2024-25) that structures background information, constraints, and output requirements into a framework and injects them into the model's context before generation, producing more reliable results than raw prompts alone.
RAG (Retrieval-Augmented Generation)
The most common implementation of Context Engineering: before answering, the system retrieves relevant documents or data from an external knowledge base and adds them to the prompt, grounding the LLM's response in specific, up-to-date information.
Harness Engineering
An emerging AI engineering paradigm (2025-26) that layers rules, guardrails, safety checks, feedback loops, Skills, and MCP on top of LLMs to achieve deterministic, production-grade delivery comparable to traditional software.
temperature
A parameter controlling randomness in LLM output. Low values (0.0-0.3) produce conservative, deterministic results suitable for code; mid values (0.5-0.7) balance creativity and coherence for design tasks; high values (0.8-1.0) maximize divergence for brainstorming.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗