The post did not define the term. By our arithmetic from the dated posts we cite, the jump from context engineering to harness engineering took 237 days, and from harness engineering to Chase’s goal engineering 273 days.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on why products built on the same model differ: the harness around it. It teaches the general idea only and does not discuss goal engineering.
The key insight: Harrison Chase’s 11 November post struck through “context engineering” and “harness engineering” and left “goal engineering”. By our arithmetic from the dated posts we cite, context to harness took 237 days and harness to goal 273 days.
The post, published at 01:37 UTC on 11 November, has three lines. The first two are rendered in strikethrough characters; the third is plain. It had about 84,000 views and 802 likes when we read it on 12 November.
We could not read Chase’s own replies in the thread. Two replies we did read played on the format: one suggested “Intent engineering”, another “next up goal orchestrator engineering”.
Dated posts that used each term, from Tobi Lütke’s 19 June 2025 post on context engineering to Harrison Chase’s 11 November 2026 post on goal engineering. The dates are of the posts we cite, not first-ever uses; the day gaps are our arithmetic.
LangChain co-founder and chief executive Harrison Chase posted this at 01:37 UTC on 11 November 2026: ‘context engineering’ and ‘harness engineering’, both struck through, and ‘goal engineering’ left standing. The post gives no definition; the piece reads it alongside the ‘Goal Tracking’ section of Chase’s 2 November LangChain blog post on new context engineering challenges.
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.
Shopify chief executive Tobi Lütke posted this on 19 June 2025, saying he preferred ‘context engineering’ over prompt engineering. It is the earliest dated post in the piece’s timeline, not necessarily the first use of the term.
+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. Science because doing this right involves task descriptions and explanations, few shot examples, RAG, related (possibly multimodal) data, tools, state and history, compacting… Too little or of the wrong form and the LLM doesn’t have the right context for optimal performance. Too much or too irrelevant and the LLM costs might go up and performance might come down. Doing this well is highly non-trivial. And art because of the guiding intuition around LLM psychology of people spirits.
On top of context engineering itself, an LLM app has to: – break up problems just right into control flows – pack the context windows just right – dispatch calls to LLMs of the right kind and capability – handle generation-verification UIUX flows – a lot more – guardrails, security, evals, parallelism, prefetching, …
So context engineering is just one small piece of an emerging thick layer of non-trivial software that coordinates individual LLM calls (and a lot more) into full LLM apps. The term “ChatGPT wrapper” is tired and really, really wrong.
Andrej Karpathy posted this on 25 June 2025, backing ‘context engineering’ over ‘prompt engineering’ and describing what goes into the context window. The piece quotes his one-line description.
Context Engineering, June 2025
Shopify chief executive Tobi Lütke wrote on X on 19 June 2025 that he preferred the term 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”, he wrote.
Four days later, Chase’s LangChain post gave a definition: “Context engineering is building dynamic systems to provide the right information and tools in the right format such that the LLM can plausibly accomplish the task”. We explain the idea in our context engineering explainer.
On 25 June 2025, Andrej Karpathy added his support on X, describing context engineering as “the delicate art and science of filling the context window with just the right information for the next step”.
Harness Engineering, Late 2025 to Early 2026
On 26 November 2025, Anthropic’s engineering post on long-running agents called its Claude Agent SDK “a powerful, general-purpose agent harness adept at coding”. To stop an agent declaring a project done too early, it had an initializer agent write a file of over 200 features, all marked as failing at the start.
On 11 February 2026, OpenAI published “Harness engineering: leveraging Codex in an agent-first world”, by Ryan Lopopolo of its technical staff. It described a team whose job was to “design environments, specify intent, and build feedback loops that allow Codex agents to do reliable work”. Its summary line: “Humans steer. Agents execute”.
The nearest thing to a definition is Chase’s 2 November LangChain post, nine days before the X post. Under a heading called Goal Tracking, he wrote: “Agents are moving from doing tasks to working towards goals”.
He listed the open problems: “It’s challenging to get agents to think about broad goals and not just small tasks”, and “It’s challenging to keep them consistent over long time horizons”.
Under Alignment he added a warning: “Goal design is hard by itself. It will be very hard for users to design goals that agents can not cheat”. His example was an instruction to get rid of all bugs that ends with the agent deleting the codebase.
A separate LangChain post on proactive agents describes how this runs in practice: “The cron then runs on some cadence (once an hour) and prompts the agent to explore new actions to progress on its goal”.
Why It Matters for LangChain
As we read the three terms, each moves the engineer’s work one step further from the single prompt: what the model sees, then the system around it, then what the agent is working towards. That reading is ours, not Chase’s.
The round is aimed at the same problem. The Journal reported that LangChain plans to use the money to expand LangSmith, its platform for monitoring, evaluating and deploying AI agents. Chase told the paper “The agent’s behavior is the hard part to control”. The valuation is 2.4 times the $1.25 billion of October 2025, by our arithmetic in our earlier piece.
The Structural Read
As we read it, the term names a problem more than a product: Chase’s own 2 November post lists goal tracking and goaldesign as open challenges, and says LangChain and the industry do not necessarily have general solutions for them.
Harrison Chase, LangChain blog, 2 November 2026
“Agents are moving from doing tasks to working towards goals”
Three Implications
FOR AGENT BUILDERS Chase’s post names the hard parts: getting agents to try new experiments, think about broad goals and stay consistent over long time horizons.
FOR BUYERS Chase wrote that it will be very hard for users to design goals that agents cannot cheat, with the example of an agent deleting a codebase to get rid of bugs.
FOR LANGCHAIN The Wall Street Journal reported that LangChain plans to use its $300 million to expand LangSmith, its platform for monitoring, evaluating and deploying AI agents.
The framework puts the fight one layer above the model: “It is about controlling the operating layer through which autonomous work is delegated, orchestrated, and governed.”
As we read it, goals are what gets delegated through that layer. The framework predates Chase’s post and does not discuss goal engineering.
What Is Not Established
We found no definition of goal engineering by Chase in the posts we read, and the LangChain blog posts we read on 12 November do not use the phrase. Those posts name no LangChain product for it. The dates above are of the posts we cite, not the first-ever uses of each term.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Goal engineering is, for now, Chase’s three-line framing, not a defined practice. His 2 November post sets out what it would have to solve: agents that pursue broad goals over long periods without gaming them. It arrived 273 days after OpenAI’s harness engineering post, by our arithmetic, and a day before LangChain’s $3 billion round was reported.
A note on sourcing. We read Harrison Chase’s 11 November post on X and LangChain’s own blog posts of 23 June 2025 and 2 November 2026 and on proactive agents; Tobi Lütke’s and Andrej Karpathy’s June 2025 posts on X; Anthropic’s 26 November 2025 post; and OpenAI’s 11 February 2026 post. The round figures are as reported by The Wall Street Journal and covered in our earlier piece.
Engagement figures are as read on 12 November. The day counts are our arithmetic from the posts’ own dates. We did not contact LangChain.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
