2026-09-18

The Cornetto ice cream cone framework for prompting

Prompting still matters, but I’m always skeptical of overly technical and sophisticated frameworks and guides.

So here’s a sorta simple one, with two mnemonics to help you remember.

The Cornette ice cream cone framework for prompting: Cream — the task, the what; Context — the why; Constraints — how to, how not to; Control — the checking loop. A black-and-white cone diagram with a capitalized heading and thin horizontal label lines.

I think it’s mostly obvious. The important part is that the control part of the prompt is a different system, meaning it doesn’t know about the cream, context, and constraints; its only goal is to check the resulting output. This helps with all sorts of hallucinations. I usually use subagents/different AIs with a fresh context window for this, but it can also be static tests in programming or, if you want, a human in the loop as well.

The name and image are inspired by the Cornetto ice cream, with the broken-off horn! Cornetto Classico packaging showing an ice cream cone with its chocolate-filled tip broken off and displayed separately.

Or you can also call it the “three C” or “four C” (depending on whether you count the cream or not) framework.

If you think you will forget about this anyway, put something like this in the system prompt:

# ask for the three c's for medium to big tasks
Medium to big tasks are tasks that are not one off, aren't a simple question or something like filling out a form or parsing a pdf, it's building something new or synthesizing multiple things. Research doesn't fall under this.

### context
the why

### constraints
the how, and the how not

### control (aka, the loop)
the controlling, importantly, the control doesn't know about the context and constraints, usually subagents or more static control like red-green tests.