Context Engineering vs. Prompt Engineering: Discourse Map

1. The received view

The dominant framing across mid-2025 to early-2026 commentary: context engineering is a superset / natural successor of prompt engineering. Prompt engineering is what you do when you type something into a chatbot; context engineering is what you do when you build an LLM-powered system. Anthropic’s official line crystallized this: “context engineering is the natural progression of prompt engineering.” Most thought leaders treat the new term as a course-correction against the way “prompt engineering” got colonized in 2023 by SEO blog posts about “act as an expert” tricks and jailbreak hacks.

The term was popularized in a roughly two-week span in June 2025: Tobi Lütke floated it (June 19), Karpathy endorsed and elaborated (June 25), Lance Martin (LangChain) and Simon Willison wrote it up (June 23/27), and Anthropic formalized it (Sept 29, 2025).

2. Direct quotes from named thought leaders

Tobi Lütke (CEO, Shopify), June 19, 2025:

“I really like the term ‘context engineering’ over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.” (https://x.com/tobi/status/1935533422589399127)

Follow-on: “DSPy is my context engineering tool of choice.” (https://x.com/tobi/status/1937967281599898005)

Andrej Karpathy, June 25, 2025:

“+1 for ‘context engineering’ over ‘prompt engineering’. People associate prompts with short task descriptions you’d give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window with just the right information for the next step.” (https://x.com/karpathy/status/1937902205765607626)

Karpathy lists what fills the window: “task descriptions, few-shot examples, RAG outputs, multimodal data, tools, state, history.”

Simon Willison, June 27, 2025:

“I think ‘context engineering’ is going to stick - unlike ‘prompt engineering’ it has an inferred definition that’s much closer to the intended meaning, which is to carefully and skillfully construct the right context to get great results from LLMs.” (https://simonwillison.net/2025/Jun/27/context-engineering/)

He adds that prompt engineering had suffered because “many people’s inferred definition is that it’s a laughably pretentious term for typing things into a chatbot.”

Anthropic engineering blog, September 29, 2025:

“At Anthropic, we view context engineering as the natural progression of prompt engineering.” “Building with language models is becoming less about finding the right words and phrases for your prompts, and more about answering the broader question of ‘what configuration of context is most likely to generate our model’s desired behavior?’” “Good context engineering means finding the smallest possible set of high-signal tokens that maximize the likelihood of some desired outcome.” (https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)

Anthropic’s X promo: “Most developers have heard of prompt engineering. But to get the most out of AI agents, you need context engineering.” (https://x.com/AnthropicAI/status/1973098580060631341)

Lance Martin (LangChain), June 23, 2025:

“Context engineering is the art and science of filling the context window with just the right information at each step of an agent’s trajectory.”

The canonical OS analogy: “LLMs are like a new kind of operating system. The LLM is like the CPU and its context window is like the RAM, serving as the model’s working memory.”

Organizes the discipline into four buckets: Write, Select, Compress, Isolate. (https://rlancemartin.github.io/2025/06/23/context_engineering/)

Walden Yan (Cognition / Devin):

“Prompt engineering was coined as a term for the effort needing to write your task in the ideal format for an LLM chatbot. Context engineering is the next level of this. It is about doing this automatically in a dynamic system.” “At the core of reliability is context engineering.” (https://cognition.ai/blog/dont-build-multi-agents)

Drew Breunig: Catalogued specific failure modes that only show up at the systems level: context poisoning, context distraction, context confusion, context clash. (https://www.dbreunig.com/2025/07/24/why-the-term-context-engineering-matters.html)

Hamel Husain has not blessed the term itself but argues substantively the same. His “LLM bullshit knife” tweet cuts through buzzwords:

“RAG → Provide relevant context; Agentic → Function calls that work; CoT → Prompt model to think/plan; FewShot → Add examples; PromptEng → Someone w/good written comm skills.” (https://x.com/HamelHusain/status/1798757828100047063)

3. Successor, superset, subset, or parallel?

Three views coexist:

4. Concrete side-by-side example

Most widely-circulated illustration is the travel-booking agent:

Second illustration from Lance Martin: Anthropic’s multi-agent researcher has the LeadResearcher write its plan to external memory before its 200K context window fills up — pure context engineering, no prompt re-wording could solve it.

5. Historical arc: what changed in the tech

6. The skeptic view

Pushback clusters on three lines:

Hamel Husain’s “bullshit knife” tweet implicitly aligns: every term is just a label for an existing concrete practice; the labels keep changing, the practice doesn’t.

Source URLs

Pedagogical note for the workshop opening

Cleanest anchor for a mixed-experience audience:

  1. Karpathy + Lütke pair — one-tweet definition + CEO co-sign
  2. Anthropic “natural progression” — institutional weight
  3. Lance Martin’s OS / RAM analogy — intuition
  4. Travel-agent side-by-side example — concrete difference
  5. Brief skeptic acknowledgment — inoculates the experienced attendees against feeling pandered to