TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Foundations · AGE 15—17 · FACTUAL · EVOLVING

What a chatbot is made of

A chatbot is a stack around a model, and the stack sets its manners.

The idea

A chatbot is not one thing but a stack of parts: a trained model at the core, a sampler that picks each token, a hidden system prompt that sets its manners, the interface you type into, and, in many products, more layers — memory that carries things between sessions, retrieval that looks documents up, tools it can call, safety filters. Not every chatbot has every layer. When a chatbot changes behaviour, one or more layers changed. If you know the stack, you can work out which.

Why it matters

"The AI got worse" gives you nothing to fix. "The wrapper probably changed its system prompt" is something you can test and work around.

See it in the studio

The same model behind two apps: one refuses site photos, one accepts them. The model is identical. The stack around it differs. Choose tools by the whole stack, not by the famous name at the core.

Watch for this

Blaming the model for its manners. Politeness, refusals, and personality mostly live in the hidden system prompt: a text somebody wrote, sitting above every conversation you have.

Try it

Ask the same question in two apps built on the same model. Log the differences. Everything that differs is the stack, not the model. You have just mapped the wrapper.

Prove it

Name the layers of the stack and give one behaviour that belongs to each.

How it works

The system prompt is a set of instructions added invisibly in front of your conversation. Memory and retrieval are ways of putting more into what the model is given; tools let the product act outside the conversation. The sampler is the draw you met at THE DRAW. None of these is the model. That is why "which model?" and "which app?" are two different questions with two different answers.

What this idea builds on

What this idea opens up

Sources

Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.

Age grows from 11 at the centre to 22 at the edge, and six sectors show the learning strands. Tab into the map and the arrow keys step from idea to idea, following the links where there is one. Enter opens the idea under the cursor, and E reads out its links and the reason recorded on each. Press slash for Search, question mark for the full key list, and Escape to leave. Open Ideas for the complete readable list, including what each idea builds on and what it opens up.

LOGIKA · RBDS AI LAB INDIA
ON-RAMP · AGE 11 · FIRST ENCOUNTERS, NOT GATES — IDEAS · — DEPENDENCIES
DONE
OPENS NEXT
SOLID — STANDS ON