From prompt engineering to context engineering to harness engineering — the three eras that turned a chatbot into a workforce.
I don't study agents from the outside. I run a production agent fleet every day — coder, deployer, CRM operator, delivery, people ops, expenses — each one wired into live business systems and judged by real outcomes, not demos.
This deck is the map I use to build: three engineering eras, and why the third one is where the leverage lives now.
Each era ended when its core artifact hit a ceiling — and the industry engineered around it.
May 2020 — GPT-3 shows that a 175-billion-parameter model learns from examples placed inside the prompt itself. No retraining, no gradients. The prompt stops being a search box and becomes a programmable interface.
An LLM is a stateless function: output = f(weights, context). You can't fine-tune your way out of every gap — but you fully control what goes in the window. Context engineering is designing that input on purpose: the right facts, the right tools, the right history, at the right moment.
A harness is everything that isn't the model: the loop, the tools, the permissions, the file system, the memory, the skills. The model proposes; the harness disposes.
Three products, three answers to “what should a harness be?”
| ERA | UNIT OF WORK | KEY ARTIFACT | FAILURE MODE | LEVERAGE |
|---|---|---|---|---|
| 01 · PROMPT | one completion | the prompt itself | silent drift on rewording | wording craft |
| 02 · CONTEXT | a grounded task | RAG + memory + tools in the window | stale or wrong context | retrieval & curation |
| 03 · HARNESS | an outcome, end to end | loop + tools + permissions + skills | runaway or unsupervised actions | supervision & evals |
Read the table as a stack, not a history: 2026 harnesses still contain 2022 prompts and 2023 retrieval. Nothing was replaced — each era became a layer.
While the market debates prompts, I operate a twelve-lane agent fleet wired into a live enterprise — CRM replacement, proposal engine, delivery pipeline, people ops — running on schedules, not good intentions.
If your organization wants that layer built, that's the conversation.