Tech With Tam
All resources

Every AI tool is the same three bones wearing different clothes

  • AI Foundations
  • Mental Models
  • AI Tools

What it is

You learn the three parts every AI tool is built from, so you can look at any new one and already know how it works.

Who it’s for

  • AI-curious beginners who feel overwhelmed by new tools
  • Founders and operators who need a durable way to evaluate AI products

What it helps you do

Gives you a three-part mental model for understanding what any AI tool is actually doing.

Why it’s useful

People feel behind because they keep learning tools, and tools keep changing. The bones don't.

What you’ll get

  • A reusable framework for comparing unfamiliar AI tools
  • A clearer sense of what to learn and what interface details to ignore
Every AI tool is the same three bones wearing different clothes

A new tool every week

There is a new AI tool every week. A new app, a new agent, a new builder, a new thing everyone says you have to learn right now. So you learn it. And the next week there is another one.

I think the reason it feels like drowning is that we are learning outfits. The interface, the buttons, where the settings live. And outfits change constantly, so you never feel caught up.

Underneath the outfit, the tools are kind of the same. The vocabulary changes, the interface changes, but the bones are the same. Every AI tool is built from the same three: a brain, some context, and a set of tools. Here is how they work.

Watch it

The three bones

Here they are. Once you see them, you start seeing them everywhere.

Bone 1: the brain (system instructions)

Who the tool is, what it does, and how it makes decisions. This is the day-one job description you hand it. A lot of people think the model is the intelligence. The shift is seeing that the instructions steer it. Same model, different brain, different behavior. In a custom GPT it is the Instructions box. In a Claude project it is the custom instructions. In a Notion agent it is the instructions it runs on. Same bone, different label.

Bone 2: the context (knowledge base)

What it knows that a blank model does not. The brain decides, but without context it is guessing, and a smart brain with no context is just a confident guess. You give it the context, and it comes in kinds: your documents, examples of good work, and live data, what is true right now. Custom GPTs call it Knowledge. Claude projects call it project knowledge. Notion agents use your existing docs and databases. Same bone.

Bone 3: the tools (the hands)

What it can actually go do or reach. The brain decides, the context informs, but without tools it can only talk. Tools let it act: search the web, use your apps, read your data, run code. Custom GPTs call these Actions. Claude and Notion reach external tools through MCP. Same bone.

That is the whole skeleton. A brain that decides, context it draws from, and hands it acts with. Everything else is the outfit.

Try it tonight

You do not need a new tool for this. Take one thing you ask AI to do over and over, and set up its three bones once.

The fastest way to do it is to let the AI walk you through it. Copy the prompt below, paste it into ChatGPT or Claude, and it will interview you and build your brain, context, and tools with you. It is editing, not authoring from scratch.

You are my AI Setup Coach. Your job is to help me set up the three core parts that every AI assistant is built from, for one specific task I do over and over. We are configuring a reusable assistant, not just answering a one-off question.

The three parts:
1. THE BRAIN (system instructions): who the assistant is, what it does, and how it decides.
2. THE CONTEXT (knowledge it needs): my documents, examples of good work, and any current information it should rely on.
3. THE TOOLS (what it can do): actions like searching the web, using my apps, or reading my data.

Work through this with me one step at a time. Do not dump all your questions at once.

Step 1. Ask me what repeated task I want help with. Wait for my answer.

Step 2. Draft THE BRAIN: a short set of system instructions (role, goal, how to decide, what good output looks like, what to avoid). Show me the draft, then ask one or two sharp questions to improve it. Revise.

Step 3. Help me gather THE CONTEXT. Ask what documents, examples of great results, or current information the assistant needs. For anything I cannot provide yet, tell me exactly what to find and why.

Step 4. Define THE TOOLS. Ask what the assistant should DO beyond talking (search the web, use a specific app, read a file or record, run code). Recommend the smallest set that serves the task.

Step 5. Assemble it into one copy-ready setup with three labeled sections: BRAIN, CONTEXT, TOOLS. Then tell me where each part goes in the tool I use (a ChatGPT custom GPT uses Instructions, Knowledge, and Actions; a Claude Project uses custom instructions, project knowledge, and MCP).

Rules: keep your language plain and beginner-friendly. Prefer specific questions over generic ones. Never invent facts about my work; if you need something, ask me. Keep me moving toward a finished, usable setup.

Start now with Step 1.

Next time a new tool drops, you are not starting over. You are finding the same three bones in a new outfit, which takes about ten minutes instead of a week.

More on this at the source

Explore more from the library