Five ChatGPT Codes That Change Every Reply

Today I've got five ChatGPT shortcodes, the setup block, a searchable Claude Code decision log, and free API options for when you hit your limit.

Happy Friday,

Today I've got five ChatGPT shortcodes, the setup block, a searchable Claude Code decision log, and free API options for when you hit your limit.

Let's get into it.

In This Email

» Save This: ChatGPT Shortcodes, five codes you set once

» Build This: the 2-minute setup, with the exact block to paste

» Try This: give Claude Code a memory with Gemini Notebook

» Try This Too: keep Claude Code running on free APIs

» Here's My Take: stop retyping the same instructions

Save This: One Word, a Better Answer

ChatGPT Shortcodes

In this lesson you'll set up five shortcodes once in Custom Instructions and try them in your chats.

→ Open the free ChatGPT Shortcodes lesson here

Here's the problem it solves. You type the same requests over and over: "make this sound human," "give me a few options," "be honest, what's wrong with this."

A shortcode is a nickname you define once: "when I type X, do Y." You save it in ChatGPT's Custom Instructions, and they guide replies rather than guarantee them. Project instructions may override them. Type CRITIC after your business idea and you get pushback instead of applause.

What it replaces: retyping the same three-sentence instruction at the bottom of every prompt.

What it costs: nothing. OpenAI's help page says custom instructions "are available on all plans on Web, Desktop, iOS, and Android," Free included.

The trap: these codes are not built into ChatGPT. Type CRITIC into a fresh account and ChatGPT just guesses what you mean. They only work after you save them. And space is tight on the free plans: OpenAI says Free and Go users "can save up to 1,500 characters in custom instructions," so all five codes plus anything you already have there must fit.

One number: 1,500 characters on Free and Go, 5,000 on Plus and higher (OpenAI Help). The block below is about 560 characters, so it fits on every plan.

Build This: Set Up All Five in 2 Minutes

Here's the exact setup. Copy the block as it is.

Open Custom Instructions. On web or desktop, click your profile picture, then Settings, then Personalization, and make sure customization is turned on. On your phone, open Settings, then Customize ChatGPT.

Paste this block into the instructions box. If you already have instructions there, add it below them:

Shortcodes. When my message includes one of these codes, follow the rule:
/human = Rewrite in plain, natural language. Short sentences, no buzzwords, no filler.
ELI10 = Explain it like I'm 10. Simple words, one everyday example.
X10THINK = Think it through carefully before answering. Check your work, then give the answer.
CRITIC = Don't just agree with me. Point out the weak spots, risks and what I'm missing, then say what you'd change.
ALT3 = Give me three clearly different versions, labeled 1, 2 and 3.
Codes can be combined, like ALT3 CRITIC.

Save. OpenAI says changes apply immediately across all your chats, including existing ones.
Test it in a new chat. Paste a stiff email draft and add one code:
Here's my email to a client: [paste your draft] /human
Stack two codes. This is where it pays off. Try:
ALT3 CRITIC subject lines for this email: [paste your email]
You get three versions plus the weak spots in each.
Make your own. Add a line in the same format, CODE = what you want it to do. Be specific. "Make it better" is weak. "Cut it to half the length and keep the main point" works.

The lesson has a test for each code and explains why X10THINK won't make ChatGPT 10x smarter.

→ Follow the step-by-step lesson here

Try This: Give Claude Code a Memory

notebooklm-py With Gemini Notebook

Claude Code already supports CLAUDE.md and auto-memory. A searchable decision log adds dated context: why you chose something, what broke and what remains open. notebooklm-py is a free, MIT-licensed community tool that lets Claude Code save notes to a Gemini Notebook (Google's renamed NotebookLM) and ask it questions at the start of the next session.

How to try it Monday: the lesson's setup starts with two commands (you need uv, a Python tool installer):

uv tool install "notebooklm-py[browser]"
notebooklm login

Those commands only install and sign in. Create and select your notebook, then connect it with the notebooklm skill or MCP server. Add the lesson's wrap-up skill and CLAUDE.md startup instruction. Save notes with /wrap-up before closing; query the notebook next session. It only remembers what you save.

The trap: it's unofficial. The README says it "uses undocumented Google APIs that can change without notice," and the first login downloads Chromium, about 170 MB. Keep client data you can't put in a personal Google account out of it.

→ Get the Gemini Notebook memory lesson here

Try This Too: Keep Coding When You Hit Your Limit

Awesome Free LLM APIs

Awesome Free LLM APIs is a free, CC0 list of "LLM APIs with permanent free tiers for text inference," including Google Gemini, Groq, Mistral, NVIDIA NIM and OpenRouter, each with its free models and rate limits. The lesson shows how to point Claude Code at those free models, as your main model or as a backup.

How to try it Monday: the lesson's quickest route is OpenRouter. Add this to ~/.claude/settings.json with your own OpenRouter key:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://openrouter.ai/api",
    "ANTHROPIC_AUTH_TOKEN": "sk-or-YOUR_KEY",
    "ANTHROPIC_API_KEY": "",
    "ANTHROPIC_MODEL": "openrouter/free"
  }
}

If you were logged in with your Anthropic account, run /logout once. Restart claude, then run /status: check that it shows the OpenRouter base URL and ANTHROPIC_AUTH_TOKEN. Keep the base URL exactly as shown, without /v1, and leave ANTHROPIC_API_KEY empty, not missing.

The trap: free tiers aren't private by default. For Mistral, the list says "Free-mode prompts may be used to train Mistral models unless you opt out." Keep client code on your paid plan.

→ Get the free LLM APIs lesson here

Here's My Take

The best AI habit I know isn't a new model. It's never typing the same instruction twice.

Shortcodes do that for ChatGPT in two minutes, on any plan. The memory setup does it for Claude Code across sessions. Same idea: tell the tool once, then get on with the work.

My advice: paste the shortcode block today and use CRITIC on the next thing you're about to send. You'll keep using it.

Go Deeper

The full walkthrough is in the AI Resource Vault: ChatGPT Shortcodes: The 5 Codes + Setup. For the Claude Code side, see Give Claude a Memory With Gemini Notebook and Keep Claude Code Running on Free LLM APIs. Every Friday pick comes from the free AI Academy, where 5,000+ members share guides, tools and what's working for them.

Join the Free AI Academy →

Free, no card. Guides, tools and the people running them live here.

That's the Debrief. Which code are you using first? Reply and tell me, I read them all.

See you Tuesday for the news.

From Drew and The AI Debrief team

P.S. Made your own shortcode? Send it over and we'll share the best ones. Hit reply, we read every one.

P.P.S. New here, or missed one? Every past edition is in the archive.