Dejan Chat Protocol: Response Rules for LLMs and Claude Code

X postDEJANResource

Dejan introduced the 'DEJAN Chat Protocol', a set of response guidelines for AI answer engines aimed at reducing cognitive load when coding with Claude. The protocol enforces leading with a direct one-sentence answer and deferring elaboration entirely to explicit user decisions.

Key points

  1. Lead with the answer: Every turn begins with a single plain sentence; detailed expansions are only provided when requested.
  2. Action menus: Concise responses conclude with an enumerated menu offering options to drill down or progress the task.
  3. Sequential interaction: Exactly one question per turn; the AI must wait until all questions are answered before executing actions.
  4. Literal interpretation: Questions are not invitations to act; 'Why' requires a reason and 'Can you' expects a yes or no.
  5. Verification and scope: Facts require verification via primary sources, files, or commands; unrelated code must remain untouched.
  6. Breach signal 'x': A user typing 'x' signals a protocol breach, prompting the AI to immediately resend the corrected answer without explanations.

Why it matters

AI coding assistants often alternate between overly verbose answers and cryptic shorthand. This protocol provides a systematic framework ensuring developers retain precise control over conversational direction and task execution.

Analysis

The protocol targets common friction points in AI-assisted development sessions. By enforcing a 'fork, don't dump' approach and forbidding conversational filler, it prevents assistants from making unprompted architectural changes. The single-character breach signal ('x') provides a low-overhead steering mechanism that forces prompt compliance without wasting tokens on apologies.

Original post

DEJAN

@dejanseo · 22 August 2026

If you're coding with Claude, you'll appreciate this. I do. t.co/RrmqDqVA1Q

89 likes4 replies198 bookmarks27.8k views

View on X
More posts in the thread (2)
  1. @natzir9 I tried everything including walking to my living room and boxing the couch in unfiltered rage. Try interacting with my demo here t.co/oXIUPlZe6r on a complex topic (needs a sign in with Google to be able to enter a URL). t.co/3lCKlxBQqh
  2. @JamesEpps6 In my experience concise turns to cryptic and I spend as much time trying to understand it as I would with a verbose response. You?
Selected replies (5)
  • @natzir9 @dejanseo I don’t know but maybe it might work better if you add it to the outputStyles path. Did you try it there? t.co/5ZtIt2FMd5
  • @natzir9 @dejanseo xDD ok ok I just put your protocol into an output style and it works there too. The win is that it lands in the actual system prompt (CLAUDE.md is injected as a user message, so it carries less instruction-following weight... token cost is the same either way) and... 1/2
  • @JamesEpps6 @dejanseo /Concise ?
  • @soniatamayodiaz @dejanseo Disrespect is correct t.co/Tw3uOwU9f2
  • @natzir9 @dejanseo ... Claude Code re-injects adherence reminders mid-session so rules stay fresh without the "x" patch, and with the "keep-coding-instructions: true" lets you drop rules 6–8 since built-ins cover them and its "more deterministic". Here it is the md outputStyle version: 2/2 t.co/ntqcdPO35W

Mentioned links

Links and tools from this entry

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Repository:multica-ai/andrej-karpathy-skills

    CLAUDE.md Ruleset Based on Andrej Karpathy's LLM Coding Observations

    This repository provides a single CLAUDE.md configuration file designed to improve Claude Code's performance. It translates Andrej Karpathy's observations on common LLM coding pitfalls into four actionable instructions to reduce overengineering and unintended code edits.

    210.3kstars

    AI & LLMs· Resource

  • X post:Alamin

    8 Structured Prompts for Software Engineering with Claude

    Alamin (@iam_chonchol) advises against treating Claude like a junior intern with simplistic commands and shares eight role-based prompt templates for software development workflows.

    129bookmarks13.5kviews

    AI & LLMs· Resource

  • Article:Anthropic

    How Claude Code Retains Project Context: CLAUDE.md and Auto Memory Explained

    Anthropic outlines the memory architecture of Claude Code. Because every session starts with a clean context window, the CLI tool relies on two complementary systems: user-defined CLAUDE.md files for explicit instructions and an auto memory feature that records learnings from user feedback.

    AI & LLMs· Resource

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.