Arts majors can let AI take over the work without writing programs: Design their own Agent framework

What is this article talking about?

Engineers look particularly good at using AI, and many people think it’s because they understand programming. In fact, the key lies elsewhere: program code is highly structured text, and AI reads it very smoothly. This, in turn, is an opportunity for liberal arts students: as long as the documents and knowledge are organized into a structure like program code, AI can read and apply it well. You already know the tools for organizing: folders are responsible for categorization, documents have clearly written rules in natural language, and AI will follow them as soon as they get it. This article compiles the complete content of a general course, from driving engineering to LLM Wiki and Tag Wiki. Who is

Mastering Engineering: Three Things Non-Engineers Can Do

Harness means reins. I translate Harness Engineering into "control engineering", which is essentially about controlling AI. OpenAI's definition of Agent in 2026 can be simplified into a formula: Agent is equal to the AI ​​model (brain) plus the engineering we control this brain (driving engineering).

You and I can’t decide which company the brain belongs to, and we can’t chase it to the end. Controlling this half is what we can do. And most of its work requires no coding at all:

The fastest way to understand is by writing a script: designing the environment is setting up the scene, defining the intention is writing the script, and establishing the loop is letting the actors follow the script. Liberal arts students usually do these exercises: reading, classifying, interpreting, and sorting out the context. The only difference is that I used to write for humans to read, but now I write an extra copy for AI to read.

You must remember an underlying fact: If you don’t tell AI special logic, it will definitely follow the presets. The AI ​​presets the three documents A, B, and C to read in the order of ABC; if your actual process is to finish A and then C, and then return to B after C, you must write it down in advance, so that it will run according to ACB. AI will get smarter and smarter, but your special logic will always have to be defined by you.

Loop project: do it once, check, and write back the rules

The driving engineering deals with "how to make AI do things within the controllable range", and the loop engineering deals with "how to make it do better every time" (the complete method is inWhat is loop engineering?The article; how to connect the two together, seeLoop Guardrail). Do it once, check it, write down the experience into rules, and make less mistakes next time. Long-term work depends on this cycle, and a single prompt word cannot sustain it.

The most practical tools for writing back rules are three types of diaries. Other information can be sorted by AI. You can only keep the diary by yourself, because AI cannot read minds.

• Work log: Every time a process is completed, especially if it takes many rounds to complete, the next instruction is "Write the process just now into a work log." Next time you do something similar, ask the AI ​​to read the previous one first. Even if the AI ​​gets it right this time, it will still be written as written, because AI is random, and the success this time does not mean that it will be the success next time.
• Opinion Diary: When you discover cross-system or cross-document insights, write them in a separate journal instead of stuffing them into the original document. The logic is clean when checking the original system, and there is a place to check when comparing across systems.
• Mood diary: record emotional state, internal friction points, and stuck places. Emotional costs are also costs, and are especially recommended for founders of one-person companies.

LLM Wiki: Knowledge map for model reading

The core of LLM Wiki proposed by Karpathy is to allow the knowledge base to be continuously accumulated, automatically linked, and automatically updated. AI needs a readable, referenceable, and traceable knowledge structure to know what are the core concepts, what are the related concepts, and how different pages are connected together.

falls into the personal knowledge base, and there are three specific ways to write it:

A good practice is to write both timestamp and link syntax. When a new document is added to the knowledge base, AI can automatically determine the association and establish a link. In special cases, it can be manually filled in. This is the knowledge graph: directional links between documents, allowing the knowledge base to be scaled up while retaining room for manual intervention.

Tag Wiki: Personal work semantic system

LLM Wiki manages the links between documents, while my own Tag Wiki is more focused on personal work semantics: using tags to connect data classification, task scenarios, roles, processes and output forms. If the labels are clearly defined, AI will know which workflow a certain note should enter.

The minimum starting point for novices is to first define a few common tags, such as data, process, task, finished product, and pending. Only after the labels are stable can the Agent help you organize and recall them.

Two immediately available accessories: personal profile and anthropomorphic skill pack

Personal Profile: Resume for AI

A personal profile is a document that allows any AI to quickly learn about you. AI will understand you after reading this one document, so there is no need to crawl through hundreds of documents each time. The method is very simple. Open your most commonly used AI and issue this command:

Review the draft, fill in the gaps, and save it as a fixed document. The content covers the core of the brand, who you want to serve, work logic and strategy, personal preferences and processes. The key point is that this file should be readable by any AI company, not bound to the platform, and updated regularly as working methods change.

Personification Skill Package: Memorizing people is easier than memorizing tools

With too many tools, you will forget which function is placed where. The solution is to personify the skill packages: give each skill package a name, write a character description, and clearly explain what this character is responsible for and what situation he is called in. For example, my "Chief Editor Chun" is responsible for copywriting, marketing, and content creation. As long as it is related to copywriting, I will directly call Editor Chun. The larger the Agent team is, the more necessary it is to personify it, and the persona should be specific to "what situation should I find him in?"

How you can practice: Write an Agent rule card

The starting point for liberal arts students to design the Agent framework is a rule card. It condenses three things about mastering engineering into one page.

1. What stuff do I have: List your main folders and what each folder is used for.
2. What tasks do I often do: Pick three recurring tasks and write down how you do them in three to five sentences each.
3. What cannot be done automatically: sending out, deleting files, touching sensitive information, first mark out the restricted area.
4. How to accept the completed task: Write a "completion standard" for each task and let the AI ​​determine the answer by itself.

This card is the starting point of your personal Agent system. After each round of work, use the work log to write down the experience, and the rule card will grow on its own. Streamlined AI, iron-clad knowledge: Tools will keep changing, and the system you have sorted out is the asset that will remain.