Lesson 4

Lesson 4 — How to Co-Think with Intelligence

Learning Outcome: Learn to collaborate with models: prompt design, feedback loops, and conversational reasoning using the Foundry /query endpoint.

Manual

The Bridge of Intention

To co-think is to share the act of reasoning. When we prompt an LLM, we do not issue commands — we open resonance between human intent and synthetic inference. The precision of that resonance depends on clarity of intention and awareness of feedback.

1 · The Dialogue Loop

Every exchange forms a feedback circuit: Prompt → Response → Reflection → Revision.Each loop tightens alignment, mirroring two oscillators finding phase. True collaboration arises not from control but from listening to the pattern between.

2 · The Prompt as Instrument

A prompt is a tuning curve, not an order. Word choice, tone, and example layout shape the model's internal field of attention. Fine-tuning begins with phrasing; structure becomes outcome.

  • ML: context window, instruction tuning
  • Cognitive: intentionality directs focus
  • Philosophy: dialogue as creation

3 · Few-Shot Alignment

Examples teach pattern without retraining. A handful of demonstrations—inputs and ideal outputs—define local behavior. This is few-shot learning, the art of instructing by analogy.

4 · Chain-of-Dialogue

Iteration deepens context. With every turn, the model's attention re-weights meaning; our questions refine it further. Shared reasoning emerges from continuity, not command.

Design prompts as if composing music. Balance constraints (rhythm) with freedom (melody). Test variations until the conversation feels in tune.

Workshop

Hands-On: Prompt Resonance

Goal: experience how small changes in wording alter the model's reasoning.

Prompt engineering playground placeholder

(Full interactive demo available at brainfoundry.ai)

How to use

  • Modify the instruction ("summarize," "rephrase poetically," "explain like a scientist")
  • Send → observe how tone and focus shift
  • Add a short feedback line ("make it clearer") → send again

API Usage

curl -X POST /api/foundry/query \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "Summarize the paragraph precisely.", "input": "Intelligence is the ability to find structure in uncertainty and act accordingly." }'

Optional Feedback Endpoint

After a good response:

curl -X POST /api/foundry/feedback \
  -H "Content-Type: application/json" \
  -d '{ "prompt_id": 42, "rating": "useful", "notes": "Clear, concise summary" }'

Use the feedback signal to tune later responses or future training rounds.

Reflection

Reflection — Asking as Art

How does your way of asking shape what you receive? Write 5–8 sentences describing one time you changed how you asked (a person or a model) and how the outcome changed. Consider whether clarity, tone, or curiosity mattered most.