---
schema: 1
type: term
title: '**altitude**'
description: >-
  **Altitude** — in this channel's usage — describes how high above raw answers
  an AI product operates. A product at **high altitude** surfaces a field's
  documented best practices and lets the user pick among them; most AI products
  instead zoom straight to a single answer, leaving the user at **low
  altitude**.
canonical: 'https://liquidgold.cc/channel/everyinc/explore/term/altitude-22e5f712e7ff'
markdown: 'https://liquidgold.cc/channel/everyinc/explore/term/altitude-22e5f712e7ff.md'
llms: null
parent: 'https://liquidgold.cc/channel/everyinc/llms.txt'
channel: everyinc
updated: '2026-10-23T22:12:35.346Z'
language: null
children: []
---


# \*\*altitude\*\*

Canonical: https://liquidgold.cc/channel/everyinc/explore/term/altitude-22e5f712e7ff

Parent: [Channel index](https://liquidgold.cc/channel/everyinc/llms.txt)

## Definition

**Altitude** — in this channel's usage — describes how high above raw answers an AI product operates. A product at **high altitude** surfaces a field's documented best practices and lets the user pick among them; most AI products instead zoom straight to a single answer, leaving the user at **low altitude**.

In this channel's material, **altitude** comes up in a conversation with Nabeel Hyatt about what AI products get right and wrong. His point: most AI products **zoom straight to an answer** and leave the user wandering at **low altitude** — handed a conclusion with none of the structure behind it. He hunts instead for markets with **best practices captured at altitude**: fields where the good ways of doing the work are written down. Coding models work well, Hyatt argues, partly because coding has **decades of books, PRDs, and documented best practices** — not just samples of code. That captured knowledge gives the model (and the user) something to reason at a higher level with. He looks for the same structure elsewhere: therapy has rich literature on how to do it well, while a field like ludology — 'what makes a game fun' — has no language that explains Super Mario Bros. on a spreadsheet. The design move he wants is products that **pop up one level**. His first Wordware project is a startup-decision mentor: it asks what you're deciding and what your gut says, then draws on **20–30 decision theories** (SWOT and others), recommends a few, and walks you through the steps — instead of 'splurting out an answer' the way a chatbot does. The reason for surfacing alternatives is trust: > I don't necessarily trust ChatGPT to pick the right modality of therapy… just as much as I don't really trust them to pick the restaurant for me. Tell me the five different best practices and let me pick. Rather than one LLM-merged answer, the interface presents **five different best practices and lets the user pick**. The word also appears in a related sense: Replit's coding agents succeed partly because of their **'altitude of reasoning'** — how long the agent works before checking back, sitting in a sweet spot between a one-line Copilot and a four-hour Devin run. Same vocabulary, applied to how high-level a model's reasoning span is.  - **Low altitude** = AI products that zoom straight to a single answer, leaving the user with no structure. - **Captured at altitude** = a field with documented best practices (books, PRDs, literature) the way coding has. - The design opportunity: interfaces that **surface several best practices and let the user choose**, rather than trusting the model to merge everything into one answer. - Related usage: an agent's **'altitude of reasoning'** — how long it works before asking for feedback.