---
schema: 1
type: term
title: _vibe-coded_
description: >-
  **_vibe-coded_** describes work — especially software or written output — that
  was produced by prompting an AI model and accepting what feels right, judged
  by hands-on intuition rather than by formal testing or benchmarks. In this
  channel's material it is closely tied to the practice of the **vibe check**,
  where new models are evaluated through direct, informal hands-on testing.
canonical: 'https://liquidgold.cc/channel/everyinc/explore/term/vibe-coded-0866c31198d1'
markdown: 'https://liquidgold.cc/channel/everyinc/explore/term/vibe-coded-0866c31198d1.md'
llms: null
parent: 'https://liquidgold.cc/channel/everyinc/llms.txt'
channel: everyinc
updated: '2026-10-23T22:12:35.346Z'
language: null
children: []
---


# \_vibe-coded\_

Canonical: https://liquidgold.cc/channel/everyinc/explore/term/vibe-coded-0866c31198d1

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

## Definition

**_vibe-coded_** describes work — especially software or written output — that was produced by prompting an AI model and accepting what feels right, judged by hands-on intuition rather than by formal testing or benchmarks. In this channel's material it is closely tied to the practice of the **vibe check**, where new models are evaluated through direct, informal hands-on testing.

In the captured material, **_vibe-coded_** sits inside the channel's broader practice of **vibe checks** — hands-on, informal testing of AI models rather than reading spec sheets or leaderboards[^vibe-check-topic]. The Every team's framework is that a model's real quality for your work comes from direct exploration: running your own benchmarks and informal tests, tuning reasoning effort per task, and weighing a model's ceiling against its consistency.Calling something *vibe-coded* signals that the result was produced by working with a model on instinct — accepting output that feels right rather than verifying every step against a formal standard. The channel treats this style of judgment as a legitimate evaluation method, but one that should be backed up with evidence: the team's vibe-check sessions are designed so that even where the team disagreed or hedged, the disagreement itself is preserved rather than flattened[^vibe-check-topic].- In day-zero testing coverage, such as the exploration of **Opus 4.8**, where a new model's coding, writing, and knowledge-work behavior is judged by hands-on feel first[^opus-video].- In model-comparison discussions spanning Claude Opus variants, GPT-5.x, Codex, Fable, and Astra, where *vibe-coded* judgments are the starting point for deeper analysis[^vibe-check-topic].- *Vibe-coded* describes work or judgments produced by intuitive, hands-on interaction with AI models rather than formal verification.- In this channel it connects directly to the **vibe check**: the Every team's core method for evaluating new model releases.- The team treats vibe-based judgments as a starting point, always pairing them with benchmarks and concrete evidence.